Merit AC News: Funding https://usemeritai.com/news/topic/funding AI news for people who use AI at work. Every piece is dated and linked to its sources. en Anthropic's amended S-1 shows $17.0B in third-quarter revenue and an operating loss driven by short-term compute https://usemeritai.com/news/anthropic-amended-s1-q3-revenue-short-term-compute https://usemeritai.com/news/anthropic-amended-s1-q3-revenue-short-term-compute Tue, 03 Nov 2026 12:00:00 GMT Funding The November 2 filing adds the quarter ended September 30: revenue up 1,127% year over year, but a $567 million operating loss that Anthropic attributes mainly to $5.0 billion of compute bought on short-term terms, which it says costs significantly more than its multi-year contracts. What it means for your business: Anthropic says it may change usage or rate limits and expects to buy more short-term compute. If a workflow depends on Claude, check your contract's capacity and deprecation terms and keep a tested fallback.

Anthropic filed an amended S-1 registration statement with the SEC on November 2, a week after its original IPO filing, adding results for the quarter ended September 30. Revenue was $17.0 billion for the quarter, up 1,127% year over year and up from $11.6 billion in the second quarter. Revenue for the first nine months of 2026 was $33.2 billion. The filing is direct about the pace: "We do not expect our revenue will continue to grow at comparable rates." As of September 30, Anthropic had more than 11,900 customers paying it $100,000 or more over the trailing twelve months, up from more than 950 a year earlier, and more than 1,900 paying $1 million or more.

Where the quarter's money went

Anthropic swung from $974 million of operating income in the second quarter to a $567 million operating loss in the third. The filing attributes the loss primarily to $5.0 billion of compute under short-term commitments and $2.0 billion of non-cash charges tied to its donation matching program. Most of its compute is contracted under multi-year commitments procured 12 to 18 months ahead; when customer demand or research needs exceed that capacity, it buys short-term compute that currently costs significantly more. Anthropic expects its use of short-term compute to keep rising in the near term and to compress its margins. Its Adjusted Operating Margin fell from 19% to 15% between the two quarters.

The third-quarter net loss was $23.3 billion, most of it a $21.9 billion change in the fair value of convertible notes, derivative liabilities and SAFEs rather than operating cash. Anthropic also reports purchase commitments and lease agreements totalling about $895.4 billion, substantially all for compute and infrastructure, with about $69.8 billion due in the twelve months after September 30. The filing also says Anthropic has changed, and may change again, usage limits, rate limits, or model and product capabilities.

Why it matters

The filing describes a supplier whose demand is running ahead of the capacity it contracted in advance, and which is paying more to close the gap. For a business spending on Claude, that makes capacity and limits a live question, not a fixed term. Check what your contract says about rate limits and model deprecation, keep a fallback model tested on your most important workflows, and measure what each workload produces, so you know which ones to protect if limits tighten.

Sources

]]> Aramco completes its purchase of 49% of Saudi AI company HUMAIN, putting in about $2.6 billion in cash https://usemeritai.com/news/aramco-closes-49-percent-humain-stake-2-6-billion https://usemeritai.com/news/aramco-closes-49-percent-humain-stake-2-6-billion Sun, 01 Nov 2026 12:00:00 GMT Funding Aramco will contribute roughly SR9.9 billion in cash during 2026 plus its own AI, computing and digital assets, while PIF keeps 51% of HUMAIN. No valuation was disclosed. What it means for your business: Gulf AI data center capacity is being financed at state scale, which should add regional capacity and data-residency options over time. HUMAIN's gigawatt figures are targets, so plan around capacity that is already running.

Saudi Aramco has completed its acquisition of a 49% stake in HUMAIN, the artificial intelligence company owned by the Public Investment Fund, according to a joint statement from PIF and Aramco on November 1 reported by Arab News. Aramco holds the stake through a wholly owned subsidiary and will contribute approximately SR9.9 billion ($2.6 billion) in cash to HUMAIN during 2026, along with its own AI, computing and digital assets. PIF keeps the remaining 51%. The closing formalizes a non-binding term sheet the two signed on October 28, 2025. No valuation was disclosed, and Universal Asset Owners notes that the release names no outside co-investor and describes the cash going into HUMAIN, with no payment to PIF.

What HUMAIN is building

HUMAIN launched in May 2025 and works across data centers, cloud and high-performance computing, AI models including the Arabic language model ALLaM, and AI services for companies and government. Arab News reports it targets 3 gigawatts of AI data center capacity by 2030. Universal Asset Owners lists a 1.7 gigawatt target for 2027 and 6 gigawatts by 2034, and cites CEO Tareq Amin naming xAI and Amazon Web Services as customers. In October, HUMAIN and DataVolt announced a preliminary framework under which Nebius would take up to 50 megawatts at the Oxagon campus in Neom, with operations starting in phases from the fourth quarter of 2027, per Arab News.

Aramco expects to use HUMAIN's infrastructure to deploy AI across its industrial operations. "This partnership with PIF strengthens Aramco's position as a global technology leader and AI powerhouse," Aramco president and CEO Amin Nasser said. Universal Asset Owners points out that the cash is small next to the build-out, which is meant to be funded mainly through debt and outside funds; it cites the publication Enterprise putting the 2027 build alone at around $20 billion. Amin has said: "When you have bankable off-take deals, raising capital for me is not an issue."

Why it matters

Saudi Arabia is assembling an AI compute supplier with its state oil company and sovereign fund behind it, and with US model makers and clouds as named customers. For businesses with operations or customers in the Middle East, that should eventually mean more local capacity and more options for keeping data in the region. The figures to rely on are the ones already built. Three gigawatts is a 2030 target, the Nebius capacity is a preliminary framework for 2027, and most of the money still has to be borrowed or raised, so any plan that depends on this capacity should have a fallback.

Sources

]]> OpenAI's mega-round comes up short as AI investor appetite cools https://usemeritai.com/news/openai-funding-round-falls-short-valuation-cut https://usemeritai.com/news/openai-funding-round-falls-short-valuation-cut Sat, 31 Oct 2026 12:00:00 GMT Funding OpenAI has secured about $11 billion so far against an original $30 billion target, at a $1.25 trillion valuation rather than the $1.4 trillion it aimed at, as a tech selloff cools investor interest in funding the AI buildout. What it means for your business: If your tools depend on a model vendor, note the financing climate is tightening; a vendor raising less than planned at a lower valuation can face more pressure on price, so avoid lock-in and keep a tested alternative.

OpenAI's latest funding round is falling short of its goal. The company has secured roughly $11 billion so far, according to a Wall Street Journal report relayed by The Decoder, against an original target of $30 billion, and the round is now pegged at a $1.25 trillion valuation rather than the $1.4 trillion it had aimed at. The shortfall is attributed to fading interest from major investors, amid a tech selloff driven by weak quarterly results from Google, Amazon and Intel.

The other side of the capex story

This is the companion to a week of record AI capital-spending numbers. The hyperscalers are committing hundreds of billions to data centers and chips, and the money to fund that buildout is raised from the same pool of investors now pulling back. A round that comes in at a lower valuation, with commitments below target, is a signal that the appetite to finance AI at any price has limits. OpenAI plans to go public early next year, and rival Anthropic could list as soon as November, which puts both in front of public markets that are scrutinizing AI spending harder than they were a quarter ago. Exact figures here are reported rather than company-confirmed and vary across outlets, so treat the specific numbers as a developing picture, not a filing.

Why it matters

For a business that runs on a model vendor, the useful read is not the headline number but what a cooling financing climate does to the product you buy. A vendor raising less than it wanted, at a lower valuation, has more pressure to show a path to profit, which over time pushes toward higher prices, tighter free tiers, or faster monetization of what is now cheap. None of that is a reason to switch today. It is a reason to avoid deep lock-in to a single provider, to keep a tested alternative you could move real volume to, and to judge each AI tool on value per dollar now, while the terms are still generous. The spending race and the funding squeeze are the same story from two sides, and both point to the same discipline: know what your AI is worth to you, so you can tell when its price stops matching it.

Sources

]]> Intel's data center sales jump 78% on AI demand, but a weak outlook sends the stock down 28.6% https://usemeritai.com/news/intel-q3-2026-data-center-ai-growth-stock-falls https://usemeritai.com/news/intel-q3-2026-data-center-ai-growth-stock-falls Fri, 30 Oct 2026 12:00:00 GMT Funding Third-quarter revenue rose 21% to $16.5 billion and data center and AI revenue reached $7.3 billion, but Intel guided to $15 billion to $16 billion for the fourth quarter and says server chip supply stays tight into 2027. What it means for your business: Agent workloads are lifting demand for ordinary server CPUs faster than Intel can make them, while memory shortages weigh on PCs. Plan hardware purchases and refreshes early, and expect tight supply and firm prices through the first half of 2027.

Intel reported third-quarter revenue of $16.5 billion on October 29, up 21% from $13.7 billion a year earlier, with non-GAAP earnings of $0.52 per share against $0.23 a year ago. Its Data Center and AI segment grew 78% to $7.3 billion, and Intel Foundry revenue rose 58% to $6.7 billion, while the client computing group fell 5% to $8.1 billion. The fourth-quarter outlook was the problem: Intel guided to revenue of $15 billion to $16 billion, non-GAAP gross margin of 38.0% and non-GAAP earnings of $0.25 per share. The stock fell 28.6% on Friday, October 30, while the S&P 500 lost 1.8%, according to Trefis.

Demand Intel cannot meet

CEO Lip-Bu Tan said in the release that "demand for our CPUs remains robust" and that Intel is working to unlock more supply for its Xeon server family. On the call, per Investing.com, he said the "server CPU TAM, largely driven by agentic workloads, could increase as much as five times from where it is today." Management said shortages of wafers and substrates will likely remain a limiting factor through the first half of 2027, and that Intel's 2027 market share will depend largely on how much it can supply.

The guidance includes about $300 million of lost fourth-quarter revenue and roughly 200 basis points of gross margin pressure from the shutdown of Intel's Fab 28 plant in Israel, according to CFO David Zinsner. Intel expects client computing revenue to fall more than 20% from the third quarter, citing industry-wide memory shortages and softer demand, and Zinsner said: "we see the PC slowdown continuing at least through the first half of next year." Intel spent $4.2 billion on capital equipment in the quarter.

Why it matters

The useful detail for a business is where AI demand is landing. Agents call tools, run code and move data, and that work runs on general-purpose server processors as well as GPUs. Intel is telling investors it cannot build those chips fast enough, and that the memory shortage feeding AI data centers is also cutting into PCs. For a company buying hardware, that points to longer lead times and firmer prices on servers and laptops for at least the next two quarters. If you plan to run agent workloads on your own infrastructure or have a laptop refresh coming, get quotes and delivery dates early, and include hardware in the cost of each AI workflow rather than counting model spend alone.

Sources

]]> AWS growth reaccelerates to 39% as Amazon's free cash flow swings to a $42 billion outflow https://usemeritai.com/news/amazon-aws-q3-2026-capex-free-cash-flow https://usemeritai.com/news/amazon-aws-q3-2026-capex-free-cash-flow Fri, 30 Oct 2026 12:00:00 GMT Funding In Q3 results reported October 29, AWS grew 39.3% to $45.98 billion with a $578 billion backlog, while Amazon's consolidated free cash flow turned to negative $42.4 billion from positive $14.8 billion a year earlier on about $220 billion of planned 2026 capital spending. What it means for your business: Cloud AI demand is real and still growing, so expect capacity and price pressure to continue; the lesson in Amazon's negative cash flow is that heavy AI spend pays back later if at all, so hold your own spend to that same test.

Amazon reported third-quarter results on October 29, and the two numbers that matter sit on opposite sides of the ledger. AWS revenue grew 39.3% to $45.98 billion, reaching a $184 billion annualized run rate, with a contracted backlog of $578 billion disclosed on the earnings call. At the same time, Amazon's consolidated free cash flow was negative $42.4 billion, compared with positive $14.8 billion a year earlier. CEO Andy Jassy told investors to expect AWS revenue growth above 40% in both 2027 and 2028.

The spend behind the growth

Amazon reaffirmed roughly $220 billion in company-wide cash capital expenditures for 2026, most of it going into the data centers and chips that run AWS, and signaled more to come. The scale shows up in the balance sheet: trailing-twelve-month net purchases of property and equipment reached $210.7 billion, up 82% from a year earlier. That is the spending that flipped free cash flow from a $14.8 billion inflow to a $42.4 billion outflow in the space of a year. The backlog and the run rate say the demand is contracted and real; the cash flow says the bill is being paid now, against revenue that arrives over the life of those contracts.

Why it matters

This is the third big cloud provider in a week to show the same shape: cloud AI revenue growing fast, and the cost of building for it growing faster. For a business deciding whether its own AI spend is worth it, the useful read is the timing. Amazon is paying out tens of billions ahead of the revenue it expects those investments to earn, and the market is now watching whether that revenue shows up on the promised scale. The same question applies one line down, to your own bill: spending committed today is justified only by work it produces over time, and a growing cost is good news only when the value it buys outruns it. The backlog is the bet. Whether it pays is the part still being decided, for Amazon and for anyone buying what it sells.

Sources

]]> OpenAI-backed Thrive Holdings agrees to buy ADT for $15.3 billion including debt https://usemeritai.com/news/thrive-holdings-to-acquire-adt-15-3-billion https://usemeritai.com/news/thrive-holdings-to-acquire-adt-15-3-billion Thu, 29 Oct 2026 12:00:00 GMT Funding Thrive Holdings will pay $10 a share in cash, about $7.6 billion in equity value and a 45.6% premium, with Apollo-managed funds providing the financing and a go-shop window open until November 28. What it means for your business: AI-aligned capital is now buying whole service businesses to rebuild them with engineers and models. If a provider you depend on changes owners this way, expect product changes and recheck its data commitments after the deal closes.

ADT agreed on October 29 to be acquired by Thrive Holdings for $10.00 per share in cash, a deal that values the home security company at about $7.6 billion in equity and $15.3 billion in enterprise value. According to ADT's announcement, filed with the SEC, the price is a 45.6% premium to the unaffected closing price of $6.87 on October 27 and 52.1% above the 30-day volume-weighted average. ADT's board approved the agreement unanimously. Funds managed by affiliates of Apollo Global Management are the sole source of financing commitments, and the deal has no financing condition. The companies expect it to close in the first quarter of 2027, subject to stockholder approval and US antitrust clearance, after which ADT's stock will stop trading on the New York Stock Exchange.

A holding company built around AI buys a legacy operator

Thrive Holdings is a permanent capital holding company spun out of Thrive Capital, the venture firm led by Josh Kushner. TechCrunch reports that it holds more than 75 companies and aims to keep them for the long term while revitalizing them with AI. OpenAI is a partner and investor in Thrive Holdings and supplies models, technology and forward-deployed engineers to its companies, per TechCrunch.

The announcement itself does not mention AI or OpenAI. It says Thrive's engineering teams will work with ADT staff to find ways technology can improve ADT's products and services, and that ADT "will form the foundation of Thrive Holdings' smart home security business." Kareem Zaki, a partner at Thrive Holdings, said: "We intend to build on that foundation by strengthening ADT's technology and product capabilities." The release also states that ADT's data and privacy policies will remain unchanged and that no new data sharing will happen without customer permission.

The agreement includes a 30-day go-shop period, ending at 11:59 p.m. New York time on November 28, during which ADT and its advisers may seek competing proposals. Parties that bid in that window can qualify for a 10-day extension.

Why it matters

Most AI money reaches ordinary businesses as software they buy. This deal is a large example of the other route: an investor with a model maker behind it buying an established service company outright and putting its own engineers inside. For a business that relies on a provider owned this way, the practical effects are product changes, new data uses and new pricing, usually arriving after the close. ADT's written data commitment is a useful baseline to hold a vendor to. The go-shop also means the buyer and price are not final until late November.

Sources

]]> Microsoft and Alphabet keep pouring billions into AI, and the market starts pricing the payoff https://usemeritai.com/news/microsoft-alphabet-q1-fy2027-ai-capex https://usemeritai.com/news/microsoft-alphabet-q1-fy2027-ai-capex Thu, 29 Oct 2026 12:00:00 GMT Funding In October 28 earnings, Microsoft reported $45.2 billion of quarterly capital spending and roughly $170 billion for 2026, while Alphabet held 2026 guidance at $195 to $205 billion. Microsoft shares rose; Alphabet's fell. What it means for your business: The hyperscalers' AI buildout is still accelerating, so expect capacity limits and price pressure to persist; judge your own AI spend the way the market now judges theirs, on what it returns rather than how large it is.

Microsoft and Alphabet reported earnings on October 28, and both showed the same pattern: cloud revenue growing fast, and capital spending on AI growing faster. Microsoft's revenue rose 19% to $92.3 billion, with its Azure cloud business up 48%, ahead of estimates. Alphabet's Google Cloud unit grew 103%, to $30.7 billion in sales and $11.1 billion in operating income. The question investors pressed on was not demand. It was the bill.

The spend

Microsoft's capital spending reached $45.2 billion for the quarter including finance leases ($39.37 billion in cash), and the company guided to roughly $170 billion for calendar 2026 and more than $50 billion in the current quarter. The figure came in below some analyst expectations, which CFO Amy Hood attributed partly to delays in third-party capacity deliveries rather than any pullback: "We still have demand that exceeds supply," she said. Alphabet held its 2026 capital-expenditure guidance at $195 billion to $205 billion and signaled that 2027 would increase significantly, without giving a number.

The market treated the two differently. Microsoft shares rose after the report while Alphabet's fell, a split that turns less on how much each is spending than on how convinced investors are that the spending will pay.

Why it matters

These are the largest AI bills in the world, and the shift worth noting is that the market has started grading them the way a careful buyer grades any AI spend: not by size, but by return. A capex number that would have been cheered as ambition a year ago now draws the question of what it produces. For a business running its own AI tools, two things follow. First, the supply crunch is real and set to continue, so capacity constraints and price pressure are the operating environment, not a passing phase. Second, the discipline being applied to these companies is the one to apply to your own line: a growing AI bill is only good news if the work it buys is worth more than the bill. The spend is the easy part to see. The value per dollar is the part that decides whether it was wise.

Sources

]]> FuriosaAI raises $620 million at a $2.9 billion valuation to scale its RNGD inference chip https://usemeritai.com/news/furiosaai-620-million-series-d-2-9-billion https://usemeritai.com/news/furiosaai-620-million-series-d-2-9-billion Thu, 29 Oct 2026 12:00:00 GMT Funding The South Korean chipmaker's Series D, led by Infinitum Partners with $280 million from Korea's National Growth Fund, values it at roughly four times its previous $735 million mark. What it means for your business: More credible inference chips outside Nvidia should put pressure on the cost of running models over time. Most buyers will reach this hardware through a cloud or service provider, so ask yours what accelerators it offers and at what price per token.

FuriosaAI, a Seoul-based maker of AI inference chips, announced a $620 million Series D on October 29 at a post-money valuation of about $2.9 billion. Dealroom reports that the valuation is roughly four times the company's previous $735 million mark and that US firm Infinitum Partners led the round. South Korea's National Growth Fund was the largest investor at $280 million, existing shareholders put in $125 million and international investors $110 million, per Pulse 2.0. Other new backers include Jericho Capital, QUDA Capital, Cisco Investments, Worldline Capital and Abu Dhabi's NeoFort Holding, and the Export-Import Bank of Korea made its first direct investment in a venture company. Digital Today reports the round in won as 820 billion at a valuation of about 3.8 trillion, enlarged on investor demand from an original target of 450 billion to 750 billion won.

The chip and its customers

The money goes to increasing production of RNGD, FuriosaAI's inference accelerator, which entered mass production in January 2026, along with expanding its software ecosystem and developing a third-generation accelerator with Broadcom. The company describes RNGD as a 180W air-cooled card with throughput comparable to GPUs that use more than twice the power, which is its own claim. Dealroom reports TSMC will produce about 20,000 RNGD chips this year and more than 40,000 in 2027, and that the Broadcom chip is not expected to sample until 2028.

Named customers are Korean: LG runs its EXAONE models on RNGD, Upstage uses it for search summarization, and Samsung SDS has launched an NPU-as-a-service offering built on it. Dealroom notes FuriosaAI turned down an $800 million acquisition offer from Meta in 2025. "We are entering a new phase of bringing our technology to the global AI infrastructure market at scale," CEO June Paik said.

Why it matters

Inference, the cost of running a model each time someone uses it, is where most business AI spending ends up, and it is the part of the market where specialized chips have the best chance against Nvidia. A $620 million round backed heavily by the Korean state gives one of those challengers the money to ship in volume, though tens of thousands of cards a year is still a small supply. Few companies will buy these chips directly. The effect for most buyers arrives through cloud and service providers, such as Samsung SDS's service, offering cheaper inference capacity. It is worth asking your providers which hardware sits behind their pricing, since the per-token rate depends on it.

Sources

]]> &Truth emerges from stealth with a $70 million seed to mathematically verify code https://usemeritai.com/news/andtruth-70-million-seed-formal-verification-code https://usemeritai.com/news/andtruth-70-million-seed-formal-verification-code Thu, 29 Oct 2026 12:00:00 GMT Funding The public benefit corporation, co-led by Stanford's Clark Barrett and former Kilometre Capital CEO Chris Hsu, is building a Lean-based system to prove code correct. It did not name its investors or disclose a valuation. What it means for your business: Investors are funding ways to check AI-written code, which signals that review is becoming the bottleneck in AI-assisted development. There is no product to buy yet, so treat the 100x speed claims as unproven and keep testing and human review as your controls.

&Truth came out of stealth on October 29 with $70 million in seed funding to build what it calls a "universal verification machine" for software. The company, a public benefit corporation founded in early 2026, has its US headquarters in Palo Alto and its international headquarters in Singapore. Its announcement describes the backers only as a group of AI safety and venture capital leaders; it did not name them or disclose a valuation.

The co-founders are Clark Barrett, co-CEO, a Stanford computer science professor, founding co-director of the Stanford Center for AI Safety and project lead for the cvc5 reasoning engine; Ilya Sergey, chief science officer, who founded the VERSE Lab at the National University of Singapore; and Chris Hsu, co-CEO and executive chairman, a former CEO of Kilometre Capital and former managing director at Citadel Investment Group. The team also includes Sebastian Ullrich, a co-architect of Lean 4.

What it says it is building

&Truth says its system will read code in any programming language and produce mathematical proofs that the code is safe, secure and correct, proofs that anyone can check independently. It builds on Lean, an open-source proof system, along with the Mathlib library and CSLib, and says it will release its foundational libraries as open source while working with the Lean Focused Research Organization. A second piece, which it calls a "specification factory," is meant to turn business requirements into formal specifications at scale.

The company claims AI already speeds up proof generation by 100x or more and that its own automation adds another 100x, cutting the work from months or years to minutes or hours. Those figures are &Truth's own and come without a public benchmark. The announcement names no customers and no generally available product. "In the AI era, even 99% safe is certainly broken," Hsu said. Sergey said: "Trust in software should rest on independently checkable proof, not any company's word, including ours."

Why it matters

As coding agents write more of a company's software, the expensive step moves from writing code to confirming it does what it should. A $70 million seed for formal verification is a sign that investors expect that review cost to become a market of its own. For teams buying AI coding tools today, nothing here changes the controls you need now: tests, code review and limits on what agents can deploy. It is worth tracking whether &Truth ships something that works on ordinary business code, and at what price, because a cheap proof of correctness would change how much of an agent's output you can accept without a person checking it.

Sources

]]> Anthropic files to go public, showing $16.3B in first-half revenue and a mostly non-cash $80.7B loss https://usemeritai.com/news/anthropic-s1-ipo-filing-revenue-loss https://usemeritai.com/news/anthropic-s1-ipo-filing-revenue-loss Tue, 27 Oct 2026 12:00:00 GMT Funding Anthropic's S-1, filed October 26, discloses $16.3 billion of revenue in the first half of 2026 — more than triple its full-year 2025 figure — alongside a headline $80.7 billion net loss that the filing attributes predominantly to non-cash accounting adjustments from earlier financing. What it means for your business: If your stack depends on Anthropic, read the filing for vendor durability: revenue is scaling fast and the $80.7B loss is mostly a non-cash charge, so weigh the operating economics, not the headline number.

Anthropic filed its S-1 registration statement on October 26, taking the first formal step toward a Nasdaq listing under the ticker "ANT." The filing puts hard numbers on a company that has mostly reported them in press briefings: $16.3 billion in revenue in the first half of 2026, against approximately $4.6 billion for all of 2025, with second-quarter revenue alone between $11.5 billion and $11.6 billion.

The number behind the number

The headline that will travel is the net loss: $80.7 billion for the first half. The filing attributes it predominantly to non-cash accounting adjustments connected to earlier financing rounds, rather than to cash spent running the business. The distinction matters, because a loss of that shape is an accounting event — a remeasurement of instruments issued when the company was worth far less — not money going out the door. For scale on the operating side, Anthropic reported an $8.06 billion operating loss and $7.33 billion in compute and infrastructure spending for 2025. The S-1 names only two direct competitors; Yahoo Finance identifies OpenAI as Anthropic's chief rival.

Why it matters

For any business whose products run on Claude, this filing is the clearest read yet on the durability of a core vendor and the pricing behind it. Revenue tripling into a single half-year says the demand is real; a GAAP loss dominated by a non-cash remeasurement says the scary figure is largely an artifact of how past fundraising is accounted for, not the rate at which cash is burning. The signal worth tracking as this listing proceeds is the operating economics and the revenue trajectory, which say whether today's model prices and support can hold. It is the same discipline that separates real value from slop in your own AI spend: judge by what the work is worth and what it actually costs to run, not by the biggest number in the headline.

Sources

]]> Pigment raises $250 million at a $2.3 billion valuation, led by the EU's Scaleup Europe Fund https://usemeritai.com/news/pigment-250-million-series-e-scaleup-europe-fund https://usemeritai.com/news/pigment-250-million-series-e-scaleup-europe-fund Wed, 21 Oct 2026 12:00:00 GMT Funding The Paris planning-software company says it passed $100 million in annual recurring revenue this year and that 85% of new metrics in its product are now created by an agent.

Pigment, a Paris company that sells business planning and performance-management software, announced on October 21 a "$250 million Series E funding round" at a "$2.3 billion valuation." The round was led by the "Scaleup Europe Fund, managed by EQT," with World Innovation Lab joining and existing investors Greenoaks, ICONIQ, IVP and Meritech taking part. The announcement also mentions "a planned investment from NVIDIA."

What the company says it has

Pigment says it "surpassed $100 million in Annual Recurring Revenue (ARR) earlier this year, once again doubling year-over-year." It names Danone, Siemens, Snowflake, ServiceNow and Unilever as customers, along with Anthropic, ElevenLabs and Uber, and says L'Oréal replaced its salary planning tool with Pigment for more than 95,000 employees. Its headline AI figure is that "85% of new metrics in Pigment are created by an Agent." The money is going to R&D, engineering hires, its Graphite compute engine and agents that run planning workflows end to end. "Enterprise planning will change more in the next few years than it has in the last thirty," said co-founders Eléonore Crespo and Romain Niccoli.

Why it matters

The 85% figure counts how many metrics an agent creates, not whether those metrics are right or get used, and every number here is company-reported. If your finance team is evaluating agent-built plans, ask the vendor how often people correct or delete what the agent creates, and test it on last quarter's numbers before trusting it with next quarter's.

Sources

]]> Meta's AI build-out outran its cash flow in the third quarter https://usemeritai.com/news/meta-q3-2026-capex-negative-free-cash-flow https://usemeritai.com/news/meta-q3-2026-capex-negative-free-cash-flow Wed, 21 Oct 2026 12:00:00 GMT Funding Meta spent $34.91 billion on capital expenditures against $32.79 billion of operating cash flow, leaving free cash flow negative, while saying its AI investments already deliver extra ad revenue.

Meta reported third-quarter 2026 results on October 21. "Revenue was $64.60 billion, an increase of 26% year-over-year on both a reported and constant currency basis." In the same quarter, "Capital expenditures, including principal payments on finance leases, were $34.91 billion," and "Cash flow from operating activities was $32.79 billion, and free cash flow was negative $2.12 billion."

The spending outlook

Meta narrowed its full-year 2026 capital expenditure range to $135-145 billion, from a prior $130-145 billion, and guided fourth-quarter revenue "to be in the range of $73-76 billion." Its outlook commentary says "Our AI investments are already delivering substantial incremental advertising revenue," crediting better engagement, recommendations and monetization. Third-quarter costs also include a $9.89 billion charge for the state attorneys general settlement Merit AC covered in August.

Why it matters

This is the largest public test of whether AI spending pays back, and the answer Meta gives is a revenue claim rather than a payback figure. Revenue grew 26%; capital spending grew fast enough to push free cash flow below zero. If you are making the same argument inside your company, that AI spend shows up in revenue, write down which revenue, measured how, before the next budget cycle asks.

Sources

]]> Salesforce agrees to buy Braze for about $5.66 billion to add it to Marketing Cloud https://usemeritai.com/news/salesforce-to-acquire-braze-5-66-billion https://usemeritai.com/news/salesforce-to-acquire-braze-5-66-billion Tue, 20 Oct 2026 12:00:00 GMT Funding Braze shareholders would get $47.50 a share in cash. Salesforce pitches the deal as giving its Agentforce agents real-time customer engagement across apps, web, email, SMS and WhatsApp.

Salesforce and Braze announced on October 20 that Salesforce has signed a definitive agreement to acquire Braze for "approximately $5.66 billion in enterprise value." Holders of Braze Class A common stock would receive "$47.50 in cash per share." The release, filed with the SEC as an exhibit to Braze's Form 8-K, says the deal is "expected to close in Salesforce's fiscal year 2028," subject to "approval by Braze's shareholders and receipt of required regulatory approvals."

Where Braze fits

Braze runs customer messaging across mobile apps, websites, email, SMS and WhatsApp. Salesforce says it will join Marketing Cloud alongside its CRM and Data 360 products, and that combining Braze with customer context "and Agentforce would help businesses" act on what customers are doing in real time. Patrick Stokes, Salesforce's president of applications and marketing, said Braze "was built to turn live customer behavior into personalized engagement at scale." Braze chief executive Bill Magnuson said Salesforce's platform "will help us innovate faster."

Why it matters

If you pay for both Braze and Salesforce today, expect the two contracts to be pitched as one once the deal closes, likely with agent features bundled in. Before that conversation, know what each tool costs you per message or per contact and what campaigns it produces, so a bundled price can be compared with what you pay now.

Sources

]]> ASI, formerly Air Space Intelligence, raises $325 million at a $3 billion valuation https://usemeritai.com/news/asi-325-million-series-c-3-billion-valuation https://usemeritai.com/news/asi-325-million-series-c-3-billion-valuation Tue, 20 Oct 2026 12:00:00 GMT Funding Spark Capital led the Series C. The company's AI plans and runs air traffic, military command-and-control and grid operations, and it holds an $875 million FAA program and a $99 million Army contract.

ASI, the company formerly called Air Space Intelligence, announced on October 20 a "$325 million" Series C at a "$3 billion" valuation. Spark Capital led the round, with affiliates of Citadel Securities, Meritech Capital, Craft Ventures, Energy Impact Partners and Andreessen Horowitz taking part.

What it sells

ASI builds AI that simulates physical systems so operators can plan and run them in real time. Its announcement cites an "$875 million program of record" from the FAA and Department of Transportation to modernize the National Airspace System, and a "$99 million" Army contract covering command-and-control work, including the Next Generation Command and Control program. SiliconANGLE reports that United Airlines, Delta Air Lines and National Grid also use its software, and that the money will fund expansion into energy and logistics. Founder and chief executive Phillip Buckendorf said ASI "is deeply focused on AI systems that can manage and control physical assets."

Why it matters

Most AI spending goes to tools that produce text and code, which a person reviews. ASI is paid to make decisions about aircraft and power grids, where a wrong answer costs more than a rewrite. Here the proof of value is a multi-year government contract, not a seat count, and that is a useful bar to hold any AI vendor to when it asks for a bigger commitment.

Sources

]]> Anthropic commits 5% of its shares to a new Public Benefit Fund, subject to stockholder approval https://usemeritai.com/news/anthropic-public-benefit-fund-five-percent-shares https://usemeritai.com/news/anthropic-public-benefit-fund-five-percent-shares Sun, 18 Oct 2026 12:00:00 GMT Funding Another 2% can be added at the board's discretion for every 50% rise in the company's value. With earlier pledges, Anthropic says about 20% of its shares are now set aside for the public good.

Anthropic announced on October 18 that, "subject to stockholder approval, 5% of Anthropic's shares will be committed today" to a new Anthropic Public Benefit Fund, "with an additional 2% added at the board's discretion for every 50% increase in the company's value above a set baseline." The company did not say what the baseline is.

How the fund is set up

The fund is a wholly owned LLC, and Anthropic says none of the equity it holds "will be used for lobbying or political giving, in the US or abroad." It may make large donations to nonprofits or direct distributions to specific groups of people; Anthropic names labor-market effects of AI as one example. Counting its co-founders' pledges to donate 80% of their equity and an employee donation match, Anthropic says "roughly 20% of Anthropic shares will be dedicated to the public good." It says it will share the fund's structure and governance in the coming months. The Associated Press reports Anthropic is planning an initial public offering as soon as November; Anthropic's post does not mention one.

Why it matters

A fund aimed partly at AI's effect on jobs is a signal of how the company that sells the tools expects them to change work. For a buyer, the more practical takeaway is that the providers you depend on are about to change shape, and a public listing brings quarterly pressure on pricing and revenue. Know what you spend with each provider and what it produces, so a price or plan change is something you can measure.

Sources

]]> Modal raises $750M at a $15.75B valuation after passing $1B in annualized run rate https://usemeritai.com/news/modal-750-million-series-d-15-75-billion https://usemeritai.com/news/modal-750-million-series-d-15-75-billion Wed, 14 Oct 2026 12:00:00 GMT Funding The AI inference infrastructure company's Series D was led by Accel. Its valuation has more than tripled since a $355M round in May.

Modal announced a $750M Series D on October 14 at a $15.75B post-money valuation. Accel led the round, Kleiner Perkins joined, and all existing investors took part, including Menlo Ventures, General Catalyst, Redpoint, Lux Capital, Amplify Partners and Bain Capital Ventures. The Next Web reports Modal raised $355M at a $4.65B valuation in May.

The numbers behind it

Modal says it "recently crossed $1B in annualized run rate," that usage "has tripled" since the May round, and that it has "grown more than 15x year-over-year." Its sandboxes product is a $300M run-rate business on its own, and Modal says it has spun up "close to 3 billion sandboxes." The New York company rents GPU capacity for running and training models, sourced from more than 20 cloud providers, and The Next Web reports it plans more products for training and fine-tuning.

Why it matters

A $300M business in sandboxes shows how much of AI spending now goes to running agents' code, not just to model tokens. If your teams run agents, that execution cost belongs in the same ledger as model spend; otherwise the cost per task looks cheaper than it is.

Sources

]]> Labelbox raises more than $100M in a Series E for reinforcement-learning data and agents https://usemeritai.com/news/labelbox-100-million-series-e https://usemeritai.com/news/labelbox-100-million-series-e Wed, 14 Oct 2026 12:00:00 GMT Funding Infinitum Partners led the round, which mixes equity and a credit facility. Labelbox says it turned profitable in 2026 and grew more than fivefold year over year.

Labelbox announced on October 14 that it has raised more than $100 million in Series E financing, "including equity and a credit facility." Infinitum Partners led the round, with Wipro Ventures, Andreessen Horowitz, B Capital, Tau Ventures and Snowpoint Ventures taking part. The release does not give a valuation or say how the total splits between equity and debt.

What the money is for

Labelbox says it "reached profitability in 2026 while growing more than fivefold year over year" and works with more than 400 customers, "including over 90% of leading U.S. AI labs." The funding goes to two products. Horizon supplies reinforcement-learning environments, expert feedback and evaluation infrastructure that labs use to train frontier models. Recursion Managed Agents, opened to enterprise customers earlier in October, runs fleets of agents inside the applications a company's teams already use.

Why it matters

Much of what frontier labs now buy is evaluation and training environments, not raw labeled data. The same idea applies to your own agents at a smaller scale: an agent that is measured against a fixed set of real tasks after every change is one you can tell is getting better, and one that isn't measured is a guess.

Sources

]]> Character.AI files for Chapter 11, with Disney bidding $20M for a license to some of its IP https://usemeritai.com/news/character-ai-chapter-11-disney-stalking-horse https://usemeritai.com/news/character-ai-chapter-11-disney-stalking-horse Mon, 12 Oct 2026 12:00:00 GMT Funding The chatbot company is running a court-supervised sale due to close by early December. The app stays up for now, but if no buyer will keep it running, it will be wound down.

Character Technologies, the company behind Character.AI, filed for voluntary Chapter 11 bankruptcy on October 12 in the US Bankruptcy Court for the District of Delaware (case 26-11603), to run a court-supervised sale of all or substantially all of its assets. The company blamed "significant legal, regulatory, and operational headwinds" and says it expects the sale to finish by the end of November or early December 2026. Reuters, citing a court filing, reports the company listed estimated assets and liabilities of between $50 million and $100 million. According to The Hollywood Reporter, the docket lists Google as the largest creditor, owed more than $4 million.

Disney's bid, and what it covers

A Disney subsidiary, Disney Worldwide Services, is the stalking-horse bidder at $20 million. Per the filing quoted by The Hollywood Reporter, the bid is for a non-exclusive perpetual license to certain company intellectual property and certain company claims. As stalking horse, Disney sets the floor and other bidders can top it. The deal follows Disney's hiring last month of Character.AI's then-CEO Karandeep Anand as its chief technology officer. The Hollywood Reporter says the filing names former Character.AI employees who joined him as providers of transition services so the company can keep operating its platform for a limited time while the sale runs.

Why it matters

Character.AI says the platform remains up, subscriptions stay active and billing is unchanged for now, and users can still export their data. But it's explicit about the downside: if no buyer will keep the service running, the platform will be wound down. This is a consumer app, but the pattern applies to any AI vendor: key people and a license to the technology can go to someone else while the product's own future is still undecided, and a paid subscription lasts only as long as the platform. For each AI tool your company pays for, know where the data export is and what happens to your workflows if the service stops.

Sources

]]> Manus raises more than $500M, its first round since China forced the end of its $2B Meta deal https://usemeritai.com/news/manus-500-million-round-after-meta-split https://usemeritai.com/news/manus-500-million-round-after-meta-split Thu, 08 Oct 2026 12:00:00 GMT Funding Boyu Capital and IDG Capital led the round for the AI agent startup's parent, Butterfly Effect, with Tencent, HSG and ZhenFund returning, six months after Chinese authorities ordered the Meta acquisition unwound.

Butterfly Effect, the parent company of AI agent startup Manus, said in a WeChat post on October 8 that it has raised "more than $500 million," TechCrunch reported. It is the company's first funding round since Meta was forced to call off its $2 billion acquisition. Boyu Capital and IDG Capital led, with existing shareholders Tencent, HSG (formerly Sequoia China) and ZhenFund participating. Manus didn't disclose a valuation. It was reported last month to be in talks to raise $500 million at a $4 billion valuation, and the company said it "will continue hiring both at home and abroad."

How it got here

Manus went viral with an agent demo, moved its staff to Singapore in mid-2025, and announced the $2 billion Meta deal that December, when it was reportedly bringing in over $100 million in annual recurring revenue. In April, Chinese authorities ordered the deal unwound amid concern about AI talent leaving for the West. Manus resumed independent operations in August and said it had to delete some user data as part of the split. It is now reportedly considering a Hong Kong IPO.

Why it matters

The funding is the headline. The forced unwind, including deleted user data, is what buyers should note. If you're standardizing on a general-purpose agent vendor, ownership and jurisdiction are operational risks: a product can change hands, then change back, with data obligations shifting each time. Price that into any agent contract alongside per-task cost.

Sources

]]> Sequoia leads a $30M seed for Catalyst, which builds AI agents that turn a stated goal into retail trades https://usemeritai.com/news/catalyst-30-million-seed-ai-trading-agents https://usemeritai.com/news/catalyst-30-million-seed-ai-trading-agents Thu, 08 Oct 2026 12:00:00 GMT Funding Jump Trading, PeakXV, Lux Capital, Coinbase and Premji Invest joined the round; a pilot with power traders generated what the company says was hundreds of millions in trading volume over a couple of weeks.

Catalyst, a startup building AI agents for retail trading, has raised a $30 million seed round led by Sequoia, Fortune reported October 8, with Jump Trading, PeakXV, Lux Capital, AntiFund, Coinbase and Premji Invest also participating. No valuation was disclosed. Cofounders Justin Zheng, 25, who joined Worldcoin in 2019, and Dylan Iskandar, 21, started the company in 2025.

What the agents do

Catalyst's agents turn a natural-language goal into a trading strategy and then execute it; per Fortune, "a human in the loop has to accept" before execution. Iskandar's framing: "A farmer shouldn't need a hedge fund to be able to hedge." Sequoia partner George Robson called the agents "a suit of armor" for navigating financial markets, and Zheng said, "We're not here to build a casino." The company says a pilot focused on power traders generated hundreds of millions in trading volume over a couple of weeks. Users are being let in from a waitlist, with wider availability expected later.

Why it matters

Trading volume is an activity metric, not an outcome: it says the agents were used, not that users ended up better off. That's the same distinction any company has to draw about its own AI spend -- usage up is easy to show, value produced is harder. Catalyst's required human approval step is the right shape for an agent that moves money; the evidence worth waiting for is how its users' results compare with what they'd have done without it.

Sources

]]> Anthropic commits $150 million over three years to the federal Genesis Mission https://usemeritai.com/news/anthropic-genesis-mission-150-million https://usemeritai.com/news/anthropic-genesis-mission-150-million Thu, 08 Oct 2026 12:00:00 GMT Funding Claude access for more than 15 federal science agencies, free API credits for "several hundred" research projects, and a focus on fusion energy and quantum computing, announced at a White House science summit and building on a DOE partnership Anthropic started last December.

Anthropic announced a $150 million, three-year commitment on October 8 to the Genesis Mission, a federal initiative to accelerate scientific and technological discovery through AI. The announcement came at the White House Office of Science and Technology Policy's "Science: A New Golden Age" summit in Washington, DC, where Anthropic said it shares "the administration's view" that AI can have a profound and positive impact on scientific discovery.

What the money funds

The commitment makes Claude available to more than 15 Genesis Mission agencies, with Claude, Claude Code and API credits set aside for "several hundred" research projects. NASA, the National Institutes of Health and the National Science Foundation are the agencies named so far. Anthropic says it will work with agencies and national laboratories on fusion energy and quantum computing, which the administration has named as priorities, and provide training and onboarding for scientists and newly joining agencies. The announcement lists no compute allocation beyond the credits, so the commitment is access and support. It does not include dedicated infrastructure.

An expansion of an existing deal

Anthropic first partnered with the Department of Energy and the Genesis Mission last December and has since brought Claude to scientists at national laboratories. This commitment widens that footprint. It follows other science moves this year: Claude Science, 10,000 free and discounted seats for academic scientists, an expanded AI for Science program, and a research preview of a hardware standard for running AI agents against lab instruments safely. The number to watch is whether "several hundred research projects" produce published results from NASA, NIH or NSF labs that credit Claude. That is checkable, and this announcement doesn't include it yet.

Sources

]]> Three ex-Ramp engineers deleted their first startup's codebase. The pivot just raised $20M. https://usemeritai.com/news/melius-series-a-ad-agents https://usemeritai.com/news/melius-series-a-ad-agents Tue, 06 Oct 2026 12:00:00 GMT Funding Melius, built after the founders' first product "didn't have legs," crossed $1M in annualized revenue two months after launch, in a market where the category leader already does $700M.

Melius, a New York startup founded by three former Ramp engineers, raised a $20 million Series A led by CRV on top of an earlier $5 million seed led by General Catalyst, TechCrunch reported October 6. That is $25 million total for a company whose current product has existed for about three months. Founders Joowon Kim, Young Kim, and Arnav Ramu describe a hard pivot. Their first product, an AI-powered performance-marketing tool, spent six months in market before the team concluded it "didn't have legs." They didn't adjust it. "We scrapped the entire codebase; we burned all of it," Joowon Kim told TechCrunch.

What they built next

Melius now calls itself an "agents lab for creative work": a system that generates ad campaigns, images, and video for marketing teams. It launched publicly in July 2026 and passed $1 million in annualized revenue two months in. That is a fast ramp in a category with a much larger leader: Higgsfield, valued at $5.4 billion as of August 2026, reportedly runs north of $700 million in annualized revenue. Krea and Flora AI round out the competitive set TechCrunch names.

The willingness to delete

Scrapping a product after six months, when investors had already funded it, deserves more attention than the funding number. A team that deletes its own codebase on that timeline when the market isn't responding is making a falsifiable claim about its own judgment, and the $20M Series A prices that judgment as much as the new product. Against a $700M-ARR incumbent, the open question is whether Melius's early revenue reflects product-market fit or early-mover froth in a hot niche. Retention and revenue numbers over the next few quarters will answer it; the funding headline won't.

Sources

]]> Lambda is raising $4B ahead of a planned IPO on a backlog that tripled in three months https://usemeritai.com/news/lambda-4-billion-raise-ipo https://usemeritai.com/news/lambda-4-billion-raise-ipo Tue, 06 Oct 2026 12:00:00 GMT Funding The Nvidia-backed AI-cloud startup's order backlog went from $15B to $50B between June and September, with $35B of it tracing to a single Anthropic compute deal. Its IPO is now pushed to 2027.

Lambda is raising up to $4 billion at a $14.5 billion pre-money valuation, TechCrunch reported October 6, with Coatue Management and Blackstone leading the round. That comes on top of $1 billion in debt financing the Nvidia-backed AI-cloud company raised the prior week. Lambda had targeted an IPO for 2026; that has slipped to 2027, a delay TechCrunch attributes to market uncertainty and nothing company-specific.

The backlog does the talking

Lambda's pitch to investors is backlog growth: from $15 billion in June to $50 billion by September, more than tripling in three months. One customer accounts for $35 billion of that: Anthropic, which signed a compute commitment with Lambda in late August. The same pattern runs across AI infrastructure. A buildout this capital-intensive depends on a few enormous, multi-year compute commitments from frontier-model labs, so Lambda's growth and Anthropic's compute needs now hold each other up.

The delayed IPO is the more interesting data point

A tripling backlog looks like the moment to go public. Lambda is pushing its IPO to 2027 anyway while raising $4B privately at a premium valuation. That suggests Lambda and its new lead investors see more value in staying private through further growth than in locking in today's valuation, even with demand this strong. Two other neoclouds, CoreWeave and Nebius, are already public, and a third, Nscale, filed last month. Track Lambda's decision to sit this wave out alongside the backlog number. If you buy compute from a neocloud, note how much of its backlog sits with a single customer.

Sources

]]> A 19-year-old raises $11M from a16z to sell a $3,499 "personal AI computer" https://usemeritai.com/news/ghost-core-personal-ai-computer https://usemeritai.com/news/ghost-core-personal-ai-computer Mon, 05 Oct 2026 12:00:00 GMT Funding Ghost's Core ships with three open-weight models preinstalled, an Nvidia Blackwell GPU, and no screen -- a bet that running AI locally is a hardware problem most people's laptops can't solve well.

Ghost, a startup founded by 19-year-old Zain Javaid along with co-founders Nicholas Chua (19), Yifei Chen (24), and Gautam Sharda (23), raised an $11 million seed round led by Andreessen Horowitz, with Abstract, Audacious Ventures, SV Angel, and Nova also participating. The company opened preorders on October 5 for Core, a $3,499 dedicated computer for running AI models locally, built around an Nvidia RTX Pro 4000 SFF Blackwell GPU, with no built-in screen -- it's controlled through a phone app or voice. Shipping is scheduled for the last week of October.

The pitch is that your laptop is the wrong box for this

Core ships preloaded with three open-weight models (Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B) and supports pulling additional models from Hugging Face, with processing kept local and personal data encrypted on-device. Javaid's framing for why this needs to be its own device rather than software on an existing machine: "You can run models on existing hardware, but it's a terrible experience on many levels." That's the bet under the $3,499 price tag: the bottleneck for running capable AI privately is consumer hardware that wasn't built for the job, not model quality.

A hardware bet in a market full of software bets

Most well-funded AI startups right now are selling access to someone else's model over an API. Ghost is selling a box, at consumer-hardware prices that are easy to compare against a MacBook or a gaming PC, for a narrower promise: AI that runs entirely on your own machine instead of a vendor's cloud. Whether that promise is worth a $3,499 premium over buying a capable GPU and running the same open-weight models yourself is the question buyers, not seed investors, are about to answer once Core starts shipping this month.

Sources

]]> Ande raises $52M to put corporate entertainment spend on autopilot https://usemeritai.com/news/ande-52-million-seed-series-a https://usemeritai.com/news/ande-52-million-seed-series-a Thu, 24 Sep 2026 12:00:00 GMT Funding An agentic platform for the $325B enterprises spend yearly on client dinners and events: booking, approvals, contracts and reconciliation, with Cloudflare and Salesforce already customers.

Ande raised $52 million in combined seed and Series A funding from Lightspeed Venture Partners, Redpoint Ventures, Duration Ventures and Sierra Ventures, with Bain Capital Ventures also participating. The company pitches itself as an "entertainment operating system" for enterprises: an agentic platform that automates the client dinners, sporting events and corporate experiences on which, Ande says, enterprises spend an estimated $325 billion a year.

What the agent does

Ande's workflow covers venue availability, booking, approval routing, contract execution and expense reconciliation, the full loop from picking a client outing to a closed-out expense report, without manual coordination at each step. It automates one recurring back-office workflow previously handled by admins and expense systems bolted together. That is a narrower, more operational use of agentic AI than most funding headlines in this category describe.

Traction claims

Ande says more than 60 enterprises use the platform, including Cloudflare, Salesforce and McGraw Hill, with customers reporting 12-15% savings and more than $400 million in entertainment spend flowing through the platform annually. These are the company's own numbers, not audited. Treat them as directional.

Sources

]]> Mistral buys Paris adtech startup Pimento for its team https://usemeritai.com/news/mistral-acquires-pimento https://usemeritai.com/news/mistral-acquires-pimento Wed, 23 Sep 2026 12:00:00 GMT Funding Mistral's third acquisition of 2026 folds Pimento's engineers into Vibe, its conversational assistant. Reports of the price disagree, and Mistral confirms only "several million euros."

Mistral AI confirmed on September 23 that it has acquired Pimento, a Paris startup whose software turns campaign briefs, brand identities, and campaign data into ad visuals and short social-media animations. It's Mistral's third acquisition of 2026. Florent Facq and Tomás Yany founded Pimento in October 2022, and it raised a €3 million seed round in December 2023 led by Partech and Cygni Capital.

The price is unsettled

Mistral told AFP only that the price was "several million euros," and the reporting that has filled the gap doesn't agree. French outlet L'Informé put the deal at around €3.8 million; Sifted, citing filings, reported €12.7 million in cash and shares. Both are small next to Mistral's €3 billion raise, but the gap between them is large enough that neither should be treated as settled.

An acquihire

Mistral says Pimento's staff are joining its engineering and product teams to work on Vibe, Mistral's conversational assistant (formerly Le Chat). Pimento's ad-creative product won't continue as a separate line. That fits a deal priced as a team buy. It's Mistral's third such move this year, building out consumer-product capability by acquiring small teams.

Sources

]]> Go.AI raises $85M for on-prem AI hardware built for bank examiners, not developers https://usemeritai.com/news/go-ai-85-million-series-a https://usemeritai.com/news/go-ai-85-million-series-a Tue, 22 Sep 2026 12:00:00 GMT Funding The Chicago startup's pitch is auditable, air-gapped AI appliances for regulated industries -- 200+ customers, 8x revenue growth in a year, and already profitable going into the round.

Go.AI, a Chicago-based provider of on-premises AI infrastructure for banks, healthcare providers, and other regulated organizations, raised an $85 million Series A led by Updata Partners, with existing investors GFT Ventures and LAUNCH participating. The round brings the company's total funding to $90 million.

The product is the hardware

Go.AI's core offering, The Go1, is described as on-prem, examiner-ready AI hardware and software built specifically so regulated institutions can run AI without sending data to a third-party cloud -- a direct answer to the compliance objection that stalls a lot of AI adoption inside banks and healthcare systems. The company rebranded from Go Abacus earlier this year alongside The Go1's launch.

Growth numbers, with the usual caveat

Go.AI reports more than 200 customers, annual recurring revenue up more than eightfold year-over-year, and says the business is already profitable. Those are the company's own disclosed figures rather than independently audited ones, which is standard for an early-stage funding announcement. The capital is earmarked for engineering headcount, product development, and expanding go-to-market beyond regulated industries specifically, which is the more interesting signal here: a company built for compliance-heavy verticals betting there's a broader on-prem AI market to sell into next.

Sources

]]> Cohere and Aleph Alpha sign a definitive merger, betting on sovereignty over scale https://usemeritai.com/news/cohere-aleph-alpha-merger-sovereign-ai https://usemeritai.com/news/cohere-aleph-alpha-merger-sovereign-ai Wed, 16 Sep 2026 12:00:00 GMT Funding The combined company, dual-headquartered in Toronto and Berlin and reportedly valued around $20 billion, pitches itself as the enterprise and government alternative to routing everything through a US frontier lab.

Cohere and Aleph Alpha signed a definitive business combination agreement on September 16, formalizing a deal the two companies first disclosed back in April. The combined business will operate globally under the Cohere name, dual-headquartered in Toronto and Berlin, with Aleph Alpha's Heidelberg site kept on as a dedicated research center and a combined headcount above 1,000. SiliconANGLE reports the deal reportedly values the merged company around $20 billion, up sharply from Cohere's own $7 billion valuation a year earlier; neither company's own announcement discloses a figure.

Why "sovereign" is the whole pitch

Cohere co-founder and CEO Aidan Gomez framed the combination around a specific customer anxiety, not a capability gap: "No government or enterprise should have to choose between capable AI and control over their technology." Aleph Alpha built its business largely around European government and regulated-industry customers who need models they can run on-premises or in domestically controlled infrastructure, rather than call out to a US-hosted API. Aleph Alpha co-CEO Ilhan Scheer, who becomes the combined company's COO, put it as bringing "what our team has built to more organizations worldwide." The deal also folds in a STACKIT integration with German retailer Schwarz Group's sovereign cloud, and per SiliconANGLE's reporting, Schwarz has separately committed roughly €500 million in financing alongside a potential $3 billion from a Canadian government-backed consortium.

What it means for buyers already tracking AI spend

For a company deciding where to route enterprise AI workloads, this is a bet that "which model" is now a jurisdiction question as well as a capability-and-price question, especially for public-sector and regulated customers who can't have their data or their model weights subject to a foreign company's terms of service. Whether the combined Cohere can deliver frontier-competitive quality on top of that sovereignty pitch, rather than trading capability for control, is what to watch once the deal closes later this year and enterprise contracts start testing the claim against real workloads.

Sources

]]> Canada and Germany put CAD 150M and EUR 100M behind Bengio's non-goal-seeking AI https://usemeritai.com/news/canada-germany-lawzero-scientist-ai-funding https://usemeritai.com/news/canada-germany-lawzero-scientist-ai-funding Wed, 16 Sep 2026 12:00:00 GMT Funding LawZero's "Scientist AI" is explicitly built not to act autonomously or pursue its own objectives -- a monitoring layer, not an agent. Two governments backed that bet over the industry's default architecture.

The Government of Canada and the Federal Government of Germany announced on September 16, at Montreal's All In AI conference, that they're investing CAD 150 million and EUR 100 million respectively in LawZero, the Montreal-based AI-safety non-profit founded and scientifically led by Turing laureate Yoshua Bengio. Per the official announcement, the money is meant to fund LawZero's build-out of "Scientist AI" and comes bundled with a partnership with Canadian compute providers Hypertec and 5C, plus a commitment to create 360 full-time jobs in Canada. Germany's contribution still requires European Commission notification.

A different architecture

What the money funds matters more than the dollar amount. Scientist AI is deliberately built to not act autonomously or pursue objectives of its own; instead of an agent that takes actions in the world, it's designed to reason transparently and produce evidence-based outputs, positioned as a monitoring and guardrail layer that can flag misaligned behavior in other, more agentic systems without itself being trained via reinforcement learning toward some goal. Germany's Federal Minister for Digital Transformation, Karsten Wildberger, called it "a novel approach to develop inherently safe AI without pursuing its own goals." Canada's AI minister, Evan Solomon, tied the funding explicitly to sovereignty: "Canada's future in AI depends on developing advanced systems that reflect our values and protect our interests" -- language that places this alongside the broader Sovereign Technology Alliance the two countries launched in February.

Why this matters beyond the safety framing

Every dollar a government puts into a non-goal-seeking monitoring model is a dollar not going toward a bigger, more autonomous frontier system. It's also a live test of whether "safety as a core capability," as Bengio put it, can be commercially and technically viable rather than only a research agenda. For buyers weighing whether to deploy increasingly autonomous agents against their own systems, LawZero's bet is a useful counter-model to watch: a purpose-built overseer that never becomes an agent itself is a different risk profile than adding another layer of prompting on top of a general-purpose one.

Sources

]]> Factory triples to a $5B valuation five months after its last round https://usemeritai.com/news/factory-200-million-series-d-5-billion-valuation https://usemeritai.com/news/factory-200-million-series-d-5-billion-valuation Tue, 15 Sep 2026 12:00:00 GMT Funding The maker of the "Droid" coding agent raised $200M led by Blackstone, Khosla Ventures and Sequoia, up from $1.5B in April, as coding-agent valuations keep compounding across the market.

Factory, the company behind the "Droid" coding agent, announced on September 15 that it has raised $200 million at a $5 billion valuation, according to its own blog post. That's more than triple the $1.5 billion valuation the company carried after its $150 million Series C round five months ago, in April, and brings Factory's total funding past $400 million. The round was backed by Blackstone, Khosla Ventures, Sequoia Capital, Insight Partners, Evantic Capital, Sound Ventures, NEA, Mantis VC and Clearlake, with angel checks from Nico Rosberg, Brad Gerstner and Marc Benioff.

What Factory sells

Factory was founded in 2023 by Matan Grinberg and Eno Reyes, and its pitch is autonomous agents ("Droids") that handle software tasks beyond writing code (refactors, migrations and deployments), deployable in the cloud, on-premises or fully air-gapped. Since Factory 2.0 shipped in April, the company says it has added Factory Router, which assigns models per task and claims to cut token spend by more than 60%, plus an "Agent Effectiveness" feature aimed at letting enterprises measure the value of their AI spend rather than only usage volume. Its blog describes the goal as giving enterprises "control over how their software factory learns, which models it uses, and where it runs."

Context: a fast-compounding category

The jump is one data point in a broader repricing of coding-agent startups this year -- Cursor's maker was acquired by SpaceX for $60 billion in August, and rival Cognition raised $2 billion at a $48 billion valuation on September 9, tripling its own valuation in four months. Whether that pace of repricing reflects proportional enterprise adoption, or is running ahead of it, is the question a spend/value tracker is built to answer once these tools show up in seat counts and usage logs rather than funding announcements.

Sources

]]> AIUC raises $40M Series A to build a UL-style certification layer for AI agents https://usemeritai.com/news/aiuc-40-million-series-a-ribbit-capital https://usemeritai.com/news/aiuc-40-million-series-a-ribbit-capital Tue, 15 Sep 2026 12:00:00 GMT Funding Founded by early Anthropic employee Rune Kvist and former METR COO Rajiv Dattani, the startup now has $55M total to expand AIUC-1, a standard that tests agents against 5,000 risk scenarios before enterprises will trust them.

AIUC (Artificial Intelligence Underwriting Company) announced a $40 million Series A on September 15, led by Ribbit Capital with First Harmonic and Terrain participating. With a $15 million seed round from NFDG, total funding is $55 million. The company was founded by Rune Kvist, an early Anthropic employee, and Rajiv Dattani, former COO of AI evaluation nonprofit METR.

Certify agents the way UL certified appliances

AIUC draws the historical parallel itself: "electricity burned down houses until insurers funded Underwriters Laboratories (UL) to certify products." Its AIUC-1 standard tests AI agents against roughly 5,000 risk-and-attack combinations, including jailbreaks, hallucinations and data leaks, built with input from more than 250 security and risk leaders. Customers already certifying against it include Cursor, ElevenLabs, Harvey, KPMG, Lovable, UiPath and Fin.

Kvist told TechCrunch the problem is trust. "AI is getting smarter at an increasingly rapid rate," he said, but that doesn't make adoption easier for regulated industries. He said "banks, hospitals, governments and militaries" hold back because "nobody can currently guarantee" a model will stay inside the commitments those institutions have made to their own customers.

What it means for agent spend

AIUC-1 is a third-party answer to a question every AI buyer asks internally: can this agent's output be trusted without a human re-checking it? If AIUC-1 or something like it becomes a requirement before enterprises expand agent deployments, certification status becomes part of the buying decision. Ask your agent vendors whether they've certified, and against what.

Sources

]]> Tandem Health raises $100M to grow from AI scribe into a full clinic operating system https://usemeritai.com/news/tandem-health-100-million-series-b-clinic-os https://usemeritai.com/news/tandem-health-100-million-series-b-clinic-os Mon, 14 Sep 2026 12:00:00 GMT Funding The Stockholm startup's Series B, led by the EU-backed Scaleup Europe Fund, is aimed at expanding beyond AI-generated medical notes into scheduling, triage and patient communications for European clinics.

Tandem Health, a Stockholm-based healthtech startup, has raised $100 million in a Series B led by the Scaleup Europe Fund, an EU-backed vehicle managed by EQT, with existing investors Kinnevik, Northzone, Amino Collective and Visionaries also participating. The round follows a $50 million Series A in 2025 and brings the company's total funding to $160 million.

Tandem's core product is an LLM-powered co-pilot that generates medical notes during patient consultations -- the company says it's now used by 10,000 care organizations across 14 European markets, with the UK, where NHS clinicians use it, as its largest market. The roughly 170-person company says the new capital goes toward two things: deepening its position with European care providers, and expanding from an AI medical assistant into what it calls an AI-native "Clinic Operating System" that would also handle patient flow, triage, scheduling and patient communications.

CEO and co-founder Lukas Saari framed the raise as a sovereignty argument as much as a product one, saying Tandem was built to give European clinics an alternative: "Europe should be setting the standard for how AI enters healthcare, not adopting someone else's."

Tandem's bet, and what decides whether this round turns into value rather than another AI tool bought and left half-used, is whether an AI co-pilot that saves clinicians documentation time gets adopted deeply enough, and widely enough across a clinic's other workflows, to show up as measurable time or cost saved rather than a line item nobody uses.

Sources

]]> OpenAI buys smartphone-camera AI startup Glass Imaging for over $300 million https://usemeritai.com/news/openai-acquires-glass-imaging-camera-startup https://usemeritai.com/news/openai-acquires-glass-imaging-camera-startup Mon, 14 Sep 2026 12:00:00 GMT Funding The Wall Street Journal reports OpenAI has acquired Glass Imaging, a computational-photography startup founded by former Apple engineers, adding another piece to its Jony Ive-led hardware push.

OpenAI has acquired Glass Imaging, a Los Altos-based computational-photography startup, in a deal worth more than $300 million, according to a Wall Street Journal report cited by both TechCrunch and SiliconANGLE. Neither company has publicly confirmed the purchase (TechCrunch notes "OpenAI did not immediately respond to a request for comment"), but the price and structure are consistent across both outlets' reporting.

Ziv Attar and Tom Bishop, two former Apple engineers who worked on the iPhone's Portrait Mode, founded Glass Imaging in 2019. Attar joined Apple in 2015 when it acquired his earlier startup, LinX Imaging, for roughly $20 million. Glass Imaging's product, GlassAI, is a neural image signal processor that runs raw camera-sensor output through networks trained on the physical characteristics of specific camera modules. It handles demosaicing, noise reduction, sharpening and frame fusion in a single pass, replacing traditional post-processing. The technology already ships in Honor's 600-series phones.

A cheap bet next to OpenAI's other hardware moves

The company had raised only about $29 million before the acquisition: a $9.3 million extended seed in 2024 and a $20 million Series A led by Insight Venture Management in May 2025 that valued it at $100 million. A $300 million sale is roughly 3x that valuation, modest next to OpenAI's $6.5 billion purchase of Jony Ive's io Products in 2025. It fits the same pattern of OpenAI buying specific hardware capabilities (industrial design with io, camera imaging with Glass Imaging) as it works toward its still-unannounced consumer device.

Sources

]]> Cornelis Networks raises $205M to make AI networking do compute as well as carry it https://usemeritai.com/news/cornelis-networks-205-million-active-compute-fabric https://usemeritai.com/news/cornelis-networks-205-million-active-compute-fabric Mon, 14 Sep 2026 12:00:00 GMT Funding The Intel networking spinoff's new Active Compute Fabric offloads collective operations into the network itself, and its 400 Gbps CN5000 switch is shipping now as an open, GPU-agnostic alternative to Nvidia's fabric.

Cornelis Networks, the Intel networking-technology spinoff, announced a $205 million funding round on September 14, led by IAG Capital Partners, alongside a new product architecture it's calling Active Compute Fabric and a strategy collaboration with Qualcomm. The capital will go toward scaling production of the company's switch line and funding its next-generation roadmap.

What Active Compute Fabric does

Rather than treating the network as a passive pipe between GPUs, Active Compute Fabric pushes programmable compute into the fabric itself -- offloading collective operations (the synchronization traffic that keeps distributed training and inference jobs in step) so accelerators spend less time idle waiting on data. It's built to industry standards (UALink, ESUN, Ultra Ethernet) rather than a single vendor's stack, which is the company's explicit pitch against Nvidia's proprietary networking: an open, GPU-agnostic layer that works across accelerator brands. CEO Lisa Spelman put the underlying bet plainly: "the fabric has to become an active part of the compute system."

The 400 Gbps CN5000 switch is shipping today; an 800 Gbps CN6000 is sampling with customers ahead of broader availability in Q4 2026.

The bigger number behind it

IAG Capital partner Joel Whitley framed the round around a market-sizing claim ("more than $55 billion of opportunity by 2030" for this category), which is a projection, not a settled fact. What is verifiable is the immediate competitive framing: as GPU clusters scale into the hundreds of thousands of chips, the network connecting them increasingly determines whether that hardware spend translates into usable training and inference throughput, rather than idle accelerators waiting on a slow fabric.

Sources

]]> Buildots raises $130M to make AI progress-tracking a default layer for data-center construction https://usemeritai.com/news/buildots-130-million-construction-ai-data-centers https://usemeritai.com/news/buildots-130-million-construction-ai-data-centers Mon, 14 Sep 2026 12:00:00 GMT Funding The Tel Aviv startup's late-stage round, led by Eyal Ofer's O.G. Venture Partners, leans into a specific bet: the AI data-center buildout is itself generating demand for AI tools that manage the buildout.

Buildots, a Tel Aviv-based construction-AI startup, has raised $130 million in late-stage funding led by billionaire Eyal Ofer's O.G. Venture Partners, with participation from Lightspeed Venture Partners, Intel Capital, Mohari Ventures, Human Capital, Qumra Capital, Viola Growth, Poalim Equity and individual investor Avigdor Willenz. The round brings Buildots' total funding to $297 million and follows a $45 million Series D roughly 16 months earlier.

Buildots' platform compares real-time site activity (captured largely through construction-site cameras) against project designs and schedules, flagging bottlenecks and predicting how a delay in one phase will ripple into others. The company says its customer base has grown from about 50 construction firms at its Series D to more than 100 large firms today, including Intel, Digital Realty, STO Building Group, JE Dunn, Mortenson, Bouygues and HOCHTIEF, and that it has sustained threefold annual revenue growth for several years, though it did not disclose specific revenue or valuation figures. The company also recently acquired Genda Inc., a construction workforce and safety-management provider, to fold labor-activity data into its progress-monitoring product.

CEO and co-founder Roy Danon pitched the round around the specific category of project Buildots is chasing: "It will soon be inconceivable that anyone managed a construction portfolio without Buildots. The blind spot we solve for a data center is the same one that's been costing a school or a hospital for decades." O.G. Venture Partners managing partner Ziv Kop put the bet in more direct terms: "Buildots is that layer for construction, trained on a volume and quality of site data that nobody else has."

The round is a specific case of a broader pattern: the same AI capital that's funding gigawatt-scale data-center construction is also funding the software layer that manages how fast those data centers get built, a secondary market that exists only because the primary one is moving as fast as it is.

Sources

]]> China's Z.AI raises $5 billion in its second multibillion-dollar raise in two months https://usemeritai.com/news/zai-zhipu-5-billion-share-placement-convertible-bonds https://usemeritai.com/news/zai-zhipu-5-billion-share-placement-convertible-bonds Sun, 13 Sep 2026 12:00:00 GMT Funding The Hong Kong-listed lab, formerly Zhipu AI, combined a discounted share placement with zero-coupon convertible bonds -- and says 60% of the proceeds go straight to model development and compute.

Z.AI, the Beijing-based AI lab formerly known as Zhipu AI, raised a combined $5 billion this week through a Hong Kong share placement and a concurrent convertible-bond sale, according to Reuters reporting by Kane Wu, independently confirmed through two separate syndications of the same wire report.

The share placement raised roughly $2 billion through 21.97 million new Hong Kong shares priced at HK$714 (about $91.05) each -- a 10% discount to the prior close of HK$793. Alongside it, Z.AI sold 20.14 billion yuan (about $3 billion) of zero-coupon convertible bonds maturing September 2027, priced at 100.5% of face value with an initial conversion price of HK$892.50, a 25% premium over the share-placement price; the company can call the bonds from February 2027 if its shares trade at or above 130% of that conversion price for 20 of 30 trading days.

This is Z.AI's second multibillion-dollar raise in two months: the company listed in Hong Kong in January 2026 and completed a roughly $4 billion follow-on share sale in July. Per Reuters, the company says 60% of this round's proceeds will go toward research and development of next-generation models and a "fully self-training system," 15% toward expansion, and the remainder toward capital-structure optimization and general working capital.

The use-of-proceeds breakdown is the number to note: three-fifths of a $5 billion raise earmarked for model development and the compute to run it is a direct, dollar-denominated answer to how much it now costs a frontier-adjacent Chinese lab to stay in the race -- and how much of that cost investors are still willing to underwrite twice in one quarter.

Sources

]]> Fortaegis raises $50M to root AI infrastructure security in silicon itself https://usemeritai.com/news/fortaegis-50-million-series-a-secure-compute https://usemeritai.com/news/fortaegis-50-million-series-a-secure-compute Sun, 13 Sep 2026 12:00:00 GMT Funding The Amsterdam startup's oversubscribed Series A, led by Serendipity Capital, will fund commercial manufacturing of chips that generate encryption keys from a device's own physical properties instead of storing them.

Fortaegis, an Amsterdam-based startup founded in October 2023, has closed an oversubscribed $50 million Series A led by returning investor Serendipity Capital, with participation from TEL Venture Capital (Tokyo Electron's venture arm), TNO, Prodrive Technologies, NP-Hard Ventures, NovaCapital, Access Ventures, Coalition Capital, and individual investors including Ilyas Khan, Ian Fujiyama and historian Peter Frankopan.

The company's pitch is "Secure Compute": a full-stack architecture spanning hardware, firmware, cryptography and software that uses the unique physical variations baked into a silicon chip during manufacturing to generate encryption keys on demand, rather than storing them on the device where they can be extracted. Fortaegis says the approach is designed to be quantum-safe and says internal and customer testing showed its communications running more than 200 times faster than conventional cryptographic protocols, with the Dutch research organization TNO unable to penetrate the system in its own testing. CEO Boudewijn Wijnands framed the bet in the company's own announcement: "Security cannot remain something we add on top of digital systems -- it must become part of the foundations of compute itself."

The company, which holds 17 filed patents with 14 more in its pipeline and already has more than 25 partnerships across government and industry in Europe, the US and Asia, plans to use the funding to begin commercial manufacturing in 2027, expanding FPGA production while developing application-specific chips. Fortaegis pitches the technology across AI, defense, telecommunications and autonomous systems -- a bet that as AI workloads get distributed across more devices and data centers, the security layer underneath them becomes its own investable category rather than a feature bolted onto someone else's chip.

Sources

]]> TAR raises $120M Series A at $1B valuation to build off-grid power for AI data centers https://usemeritai.com/news/tar-120-million-series-a-off-grid-ai-power https://usemeritai.com/news/tar-120-million-series-a-off-grid-ai-power Sat, 12 Sep 2026 12:00:00 GMT Funding The Austin startup builds modular, self-contained power systems so AI data centers can add generation capacity without competing with local communities for grid capacity.

TAR announced a $120 million Series A at a $1 billion post-money valuation on September 12, 2026, led by Spark Capital. The Austin-based company, founded in 2026 by Pat Becker and Leonhard Soenke, builds self-contained modular systems that combine renewable power generation and battery storage to run independently of the electric grid, letting AI data center projects add capacity without competing with surrounding communities for grid power. Spark Capital general partner Will Reed said "power is becoming the main bottleneck to scaling compute."

Owning the whole build

TAR's pitch goes beyond the hardware to controlling the entire project lifecycle in-house, from site selection and engineering through procurement, construction, and ongoing operations, rather than coordinating multiple outside contractors the way conventional power projects typically get built. The company says it has also built a purpose-designed deployment automation stack meant to install generation capacity faster and with less field labor than standard construction. Its team draws on energy-industry veterans from Hut 8, AES, and Vistra alongside robotics engineers from Zipline, GrayMatter Robotics, and Lucid Motors -- a hybrid staffing bet that treats power buildout as a manufacturing and robotics problem as much as a utility one.

The power constraint is now a funding thesis of its own

A $1 billion valuation for a company with no product history before 2026 is only legible in the context of how binding the power bottleneck has become for AI infrastructure -- data center buildouts increasingly wait on interconnection queues and grid capacity, not chip supply. TAR is currently executing a utility-scale deployment for one of the largest neocloud operators and completing its first manufacturing and logistics facility in West Texas; the round is a bet that off-grid, vertically integrated power delivery becomes standard practice for gigawatt-scale AI campuses rather than a stopgap.

Sources

]]> Mecka AI nears $500M valuation in Sequoia-led round for robot training data https://usemeritai.com/news/mecka-ai-500-million-valuation-sequoia https://usemeritai.com/news/mecka-ai-500-million-valuation-sequoia Fri, 11 Sep 2026 12:00:00 GMT Funding The startup pays people to record themselves doing everyday tasks like making coffee or fixing a car, three months after a $60M round put it at a fraction of that value.

Mecka AI is nearing a roughly $500 million valuation in a new round led by Sequoia Capital, TechCrunch reported September 11, 2026, citing two people with knowledge of the deal. Neither Mecka nor Sequoia would comment, and the size of the new round wasn't disclosed. It follows a $60 million round led by Framework Ventures, with Menlo Ventures, SV Angel, and Kindred Ventures also participating. The jump would be one of the steeper markups among the current wave of robotics-data startups.

Paying people to be training data

Mecka collects human motion data using body sensors and smartphones, then sells it to train humanoid robots and other robotics systems. People strap on sensors and record themselves making coffee or fixing a car. It is an unglamorous answer to the industry's most basic bottleneck: humanoid robots need enormous volumes of real-world physical-task data that no dataset holds yet. Co-founder Josh Gao previously told Fortune the company was projecting a $100 million annual run rate by the end of 2026.

Why data-collection startups command lab-scale valuations

A five-fold-plus valuation jump in roughly three months for a company that pays humans to perform chores says more about where robotics capital is flowing than any humanoid demo video. Model architectures for embodied AI are converging faster than the data to train them, so whoever controls a scalable pipeline of labeled physical-task data has leverage out of proportion to headcount or product maturity. If you're evaluating robotics vendors, scrutinize their data-supply claims as hard as their compute claims. Data is what gates deployment timelines right now.

Sources

]]> Cohere in advanced talks to raise $2-3B at a $20B valuation https://usemeritai.com/news/cohere-2-3-billion-raise-20-billion-valuation-talks https://usemeritai.com/news/cohere-2-3-billion-raise-20-billion-valuation-talks Fri, 11 Sep 2026 12:00:00 GMT Funding The Canadian sovereign-AI lab is reportedly courting the Canadian and German governments alongside existing backers, positioning itself as a non-U.S., non-Chinese alternative for enterprise and government buyers.

Cohere is in advanced talks to raise between $2 billion and $3 billion at a roughly $20 billion valuation, The Globe and Mail reported September 11, 2026, citing four people familiar with the matter. Cohere did not respond to a request for comment. Existing backers reportedly involved in the round include Radical Ventures, Inovia Capital, the Healthcare of Ontario Pension Plan, PSP Investments, the Business Development Bank of Canada, and Nvidia; the Canadian government is also reportedly a participant, with talks underway to bring in the German government as well.

Selling sovereignty, not scale

Cohere builds large language models for enterprise and government customers rather than consumers, and has built its pitch specifically around not being OpenAI, Anthropic, or a Chinese lab -- positioning itself as a "sovereign alternative" for governments and regulated industries wary of routing sensitive workloads through a U.S. mega-cap or a geopolitical rival. Government participation in the round, if it closes as reported, would be an unusually direct expression of that pitch: state capital betting on domestic AI infrastructure independence rather than only on the company's ability to outcompete the labs it's positioned against.

The valuation math is still speculative

A $20 billion valuation would roughly double what Cohere was last reported to be worth, but every figure here comes from anonymous sourcing rather than a signed term sheet -- the deal isn't closed, and both the amount and the valuation could move before it does. For a company whose entire commercial thesis rests on government and enterprise buyers who move slowly and demand long diligence cycles, whether a national government follows through as an investor is a more durable signal than the headline number.

Sources

]]> UAE pledges $46.5 billion to Germany, with a 1-gigawatt data center inside the package https://usemeritai.com/news/uae-germany-46-5-billion-ai-investment https://usemeritai.com/news/uae-germany-46-5-billion-ai-investment Thu, 10 Sep 2026 12:00:00 GMT Funding Announced during the UAE president's state visit to Berlin, the pledge bundles AI data-center capacity with energy and industrial deals -- a sign Gulf capital is now underwriting European AI infrastructure directly.

The United Arab Emirates pledged to invest 40 billion euros (about $46.5 billion) in Germany during President Sheikh Mohamed bin Zayed Al Nahyan's state visit to Berlin on September 10, 2026, with a data center of up to 1 gigawatt of capacity named as part of the package alongside energy and industrial commitments. The visit also produced 29 separate business agreements worth 9.4 billion euros, and builds on a February 2026 memorandum between Germany's RWE and Abu Dhabi's ADNOC for a decade of LNG imports. Current bilateral trade between the two countries runs about $15 billion annually. UAE Industry Minister Sultan Ahmed Al Jaber put it plainly: "The UAE invests for the long term and builds partnerships that endure."

Gulf capital moving from chips to the buildings that house them

Most of the AI-infrastructure money flowing out of the Gulf this year has gone into compute deals -- GPU purchases, cloud joint ventures, chip-adjacent financing. A sovereign-level pledge that names a specific gigawatt-scale data center inside a broader bilateral investment package is a different category: it's Emirati capital underwriting the physical infrastructure layer in a G7 economy directly, at head-of-state level, rather than through a corporate joint venture.

What's still unconfirmed

Neither government has named the site, the operator, or which UAE entity will structure the data-center piece specifically. Pledges announced during state visits routinely take months or years to convert into signed, sited projects, so the number to watch next is whatever specific data-center agreement, if any, eventually gets named, rather than the $46.5 billion topline.

Sources

]]> AI-inference chip startup Positron raises $875M at a $5B valuation https://usemeritai.com/news/positron-ai-875-million-series-c https://usemeritai.com/news/positron-ai-875-million-series-c Thu, 10 Sep 2026 12:00:00 GMT Funding A memory-first architecture, not more raw GPU scaling, is the bet -- Positron's Atlas inference chips are already deployed at hyperscaler scale, with next-gen custom silicon "Asimov" taping out later this year.

Positron AI raised $875 million on September 10, 2026: a $375 million Series C at a $3.5 billion pre-money valuation, plus up to $500 million in a Series C-1, reaching a $5 billion post-money valuation. The round was co-led by NEA, Atreides Management, Valor Equity Partners, Andra Capital, and SemiAnalysis Capital (the investment arm of chip analyst Dylan Patel's firm), with Silicon Graphics/Netscape founder Jim Clark, Qatar Investment Authority, Cisco Investments, and Naver Ventures also participating.

Betting against pure GPU scaling

NEA's Forest Baskett called it "the boldest memory-first bet in AI hardware". The architecture is the story here. Positron's Atlas inference chips are built around consumer-grade memory rather than the high-bandwidth memory driving most AI-chip cost inflation, already deployed at hyperscaler scale; a next-generation custom chip, "Asimov," is set to tape out later in 2026.

Why this matters beyond the valuation number

Once a model is in production, most enterprise AI spend goes to inference, not training, and a $5 billion valuation on a memory-first inference architecture is a direct market signal about where buyers expect that cost curve to bend next. For finance and infrastructure leaders modeling AI unit economics, a viable alternative to raw GPU-scaling on the inference side is worth tracking regardless of whether Positron specifically wins.

Sources

]]> Pentagon in talks to lend AI cloud startup Fluidstack $5 billion https://usemeritai.com/news/pentagon-fluidstack-5-billion-loan https://usemeritai.com/news/pentagon-fluidstack-5-billion-loan Thu, 10 Sep 2026 12:00:00 GMT Funding The Office of Strategic Capital's proposed loan targets manufacturing and supply-chain capacity for data-center components, not a new facility -- a government bet on AI infrastructure supply directly, beyond AI models.

The Pentagon's Office of Strategic Capital is negotiating a roughly $5 billion loan to Fluidstack, an AI cloud-computing startup, according to Wall Street Journal reporting carried on the Reuters wire on September 10, 2026. The loan is structured to fund manufacturing and supply-chain capacity for data-center components, shoring up the broader US supply chain that feeds AI infrastructure buildouts, rather than financing a specific new facility.

A government office built for this

The Office of Strategic Capital exists specifically to make this kind of investment: government capital underwriting the industrial base an emerging technology depends on, the same logic historically applied to semiconductors and critical minerals now extended directly to AI data-center supply chains. A loan at this scale, to a company most outside the industry haven't heard of, says more about how seriously the Pentagon takes AI infrastructure supply risk than a headline chip-export policy would.

Why the supply-chain framing matters

Most AI infrastructure financing stories are about compute capacity directly -- gigawatts, chips, data centers. This one is upstream of that: the components and manufacturing capacity that data-center buildouts depend on. For anyone modeling how exposed the broader AI infrastructure buildout is to supply bottlenecks, a government loan aimed specifically at that layer is worth tracking as its own category, separate from compute-capacity announcements.

Sources

]]> Oracle's cloud backlog hits $664B -- and nearly half of it is tied to one OpenAI contract https://usemeritai.com/news/oracle-q1-fy2027-earnings-ai-backlog https://usemeritai.com/news/oracle-q1-fy2027-earnings-ai-backlog Thu, 10 Sep 2026 12:00:00 GMT Funding Q1 FY2027 revenue beat estimates and OCI revenue rose 121%, but the concentration behind the headline backlog number is the detail worth reading past the stock pop.

Oracle reported Q1 FY2027 results on September 10, 2026: revenue of $19.35 billion (up 30% year over year, beating $19.14 billion consensus), adjusted earnings per share of $1.92 (beating $1.74 consensus), and net income of $4.68 billion, up from $2.93 billion a year earlier. Oracle Cloud Infrastructure revenue rose 121% to $7.4 billion. Remaining performance obligations (Oracle's booked-but-not-yet-recognized backlog) reached $664 billion, with capital expenditure jumping to $28.5 billion from $8.5 billion a year prior. Shares rose 4-7% in after-hours trading.

The concentration the headline number hides

Nearly half of that $664 billion backlog is tied to a single contract with OpenAI. Valoir analyst Rebecca Wettemann called the RPO figure itself "strong", and on its face it is, but a backlog this large concentrated in one customer relationship is a different risk profile than the same dollar figure spread across a diversified base of enterprise buyers.

Why this matters for reading any AI infrastructure number

Big backlog and capex numbers are increasingly the headline metric labs and cloud providers lead with, but Oracle's own disclosure is a reminder to check concentration before treating a backlog figure as evidence of broad-based enterprise AI demand. For anyone benchmarking how much diversified demand is behind the AI infrastructure buildout, "how much of this is one customer" is the question to ask before "how big is the number."

Sources

]]> Nvidia backs a 2-gigawatt AI infrastructure buildout across Australia with eight local partners https://usemeritai.com/news/nvidia-australia-2-gigawatt-ai-buildout https://usemeritai.com/news/nvidia-australia-2-gigawatt-ai-buildout Thu, 10 Sep 2026 12:00:00 GMT Funding Sharon AI alone plans to deploy up to 68,000 Nvidia GPUs, and IREN is building an 800MW campus in South Australia. Hyperscale AI capacity is expanding into a region with its own energy and grid constraints.

Nvidia announced a partnership with eight Australian data-center firms on September 9, 2026 (Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC, and AirTrunk) to expand the country's AI infrastructure capacity toward roughly 2 gigawatts by 2027, hosting multiple generations of Nvidia's DSX AI-factory compute platforms. Sharon AI alone plans to deploy up to 68,000 Nvidia GPUs with Quantum InfiniBand and Spectrum-X networking. IREN is building an 800-megawatt campus in South Australia. Partner Firmus has branded the effort "Project Southgate," tied to roughly $10 billion in financing per subsequent coverage.

A new region

Nvidia VP Raj Mirpuri described the logic: "AI factories turn energy into intelligence -- the essential resource of the AI economy." Putting multi-gigawatt capacity in Australia, in place of further growth at existing US or Asian hubs, brings hyperscale AI infrastructure into a region with its own energy market, grid constraints, and data-residency rules. The operating conditions are new, beyond the extra square footage.

Regional latency and data-residency planning

If you run AI workloads that need to stay close to Asia-Pacific users or meet regional data-residency requirements, a 2-gigawatt buildout across eight operators means meaningful new capacity is coming online in the region. With multiple named local providers competing to host the same platform generation, you also get more choice of provider.

Sources

]]> NEXTDC taps capital markets for the third time in four months to fund AI data centers https://usemeritai.com/news/nextdc-1-1-billion-convertible-notes https://usemeritai.com/news/nextdc-1-1-billion-convertible-notes Thu, 10 Sep 2026 12:00:00 GMT Funding The Australian operator priced A$1.1 billion in convertible notes due 2031, its third raise since early 2026. The pace says as much about the capital intensity of AI-ready data centers as any single deal size.

NEXTDC priced an A$1.1 billion (roughly $795 million) convertible notes offering on September 10, 2026, maturing in September 2031 with a 1.75% coupon and a conversion price of $16.695 per share, a 32.5% premium to the $12.60 reference price. Net proceeds after related capped-call transactions are expected to be about $1.006 billion, which NEXTDC says will fund its Australian data-center development pipeline. The company entered capped call transactions with a $21.42 cap price to limit dilution, and plans to list the notes on the Vienna Multilateral Trading Facility instead of the ASX. Pro forma available liquidity stood at roughly $9.8 billion as of June 30, 2026.

The third raise since early 2026

This is NEXTDC's third capital markets trip in four months, following an A$1.8 billion senior debt facility and a $750 million subordinated notes offering earlier in the year. CEO Craig Scroggie said the deal "provides NEXTDC with efficient, committed funding for our development pipeline and diversifies NEXTDC's sources of capital." Three raises in that window is the signal: AI-ready data-center capacity is capital-intensive enough that a well-funded incumbent keeps returning to the market.

What it funds

NEXTDC hasn't disclosed a new capex figure tied to this raise, but it has previously flagged sharply rising capital expenditure to meet AI-driven demand for contracted, power-ready capacity. If you track how much construction sits behind data-center announcements, a steady cadence of convertible-note and debt raises is a more reliable signal of sustained build-out than one headline figure.

Sources

]]> Microsoft plans to more than triple data-center capacity to 38GW by 2032, after turning away customers https://usemeritai.com/news/microsoft-38-gigawatt-data-center-plan https://usemeritai.com/news/microsoft-38-gigawatt-data-center-plan Thu, 10 Sep 2026 12:00:00 GMT Funding A capacity crunch reportedly forced Microsoft to decline cloud and AI clients and contributed to an 8-hour GitHub outage tied to server constraints. Meanwhile Microsoft is shifting lease terms to manage how the buildout hits its own books.

Microsoft plans to grow its global compute footprint from roughly 12 gigawatts today to more than 38 gigawatts by 2032, according to Bloomberg reporting carried on the Reuters wire on September 10, 2026, citing people familiar with the matter (Microsoft declined to comment). AI-dedicated capacity grows from about 2 gigawatts now to roughly a third of the 38-gigawatt total. Reported drivers include capacity shortages that forced Microsoft to turn away cloud and AI clients (one retailer reportedly went to Oracle) and an August outage tied to server constraints that took GitHub down for eight hours.

The lease terms change how to read the number

Microsoft is reportedly shifting new data-center leases from 15-year to 25-year terms, which lowers how much of the build cost is recognized as capex in any given year. The 38-gigawatt figure is physical buildout. The pace at which it shows up on Microsoft's financial statements is being managed.

Turning away paying customers is the signal

A hyperscaler declining business because it physically can't serve it is a more concrete data point on the AI capacity crunch than any capex guidance number. Microsoft's FY2027 Q1 capex guidance sits at $50 billion, with roughly $175 billion projected for calendar 2026. If you're a finance leader modeling AI infrastructure costs, weight the turned-away customers most heavily: they are evidence of constrained supply today, where capex guidance only projects demand.

Sources

]]> Senior-living AI company Inspiren raises $70M, crossing a $500M valuation https://usemeritai.com/news/inspiren-70-million-series-c-senior-living-ai https://usemeritai.com/news/inspiren-70-million-series-c-senior-living-ai Thu, 10 Sep 2026 12:00:00 GMT Funding A vertical AI product, sensor and AI monitoring across roughly 500 facilities and 40,000 residents, growing on deployment scale.

Inspiren sells sensor-and-AI monitoring for senior-living facilities: fall detection, behavioral alerts, and emergency alerts. It raised a $70 million Series C on September 10, 2026, led by NewView Capital, with Insight Partners, Primary Venture Partners, Scale Venture Partners, Vintage Investment Partners, Lightbank, Avenir Growth Capital, Story Ventures, and Camber Creek participating. The round pushes Inspiren's valuation past $500 million and its total funding to $225 million. The technology is deployed across roughly 500 facilities, monitoring about 40,000 residents.

A smaller round with a concrete proof point

Next to the week's multi-billion-dollar frontier-lab rounds, a $70 million round for a vertical AI company rests on deployment at scale (40,000 monitored residents) more than on future potential. NewView Capital partner Nick Bunick made the case: "few technologies can improve both clinical outcomes and community economics at the same time."

Why this belongs next to the bigger stories

If you're a finance leader trying to separate AI spend that produces measurable outcomes from spend that chases a category, mid-size vertical rounds like this are the useful comparison. They show operational proof at a fraction of the capital, in a sector (senior care) outside the frontier-model hype cycle.

Sources

]]> Fluidstack breaks ground on a $4 billion Texas data-center campus, no tax breaks attached https://usemeritai.com/news/fluidstack-cameron-county-texas-data-center https://usemeritai.com/news/fluidstack-cameron-county-texas-data-center Thu, 10 Sep 2026 12:00:00 GMT Funding The AI cloud startup started site work on 1,100+ acres near Rio Hondo under a closed-loop cooling mandate the county imposed after local pushback on water use -- and is paying full property taxes to get it.

Fluidstack began site work on September 10, 2026 on the first phase of a $4 billion AI data-center campus on more than 1,100 acres near Rio Hondo in Cameron County, Texas, after the county issued a worksite permit the day before. The company projects more than 2,500 construction jobs and several hundred permanent operations roles once the facility is running. Cooling water comes from the East Rio Hondo Water Supply Corporation through a closed-loop system that recycles water rather than losing it to evaporation -- a design Fluidstack says uses less water than one Olympic-sized pool, refilled repeatedly rather than drawn continuously.

Built to a standard the county wrote after a fight

Cameron County passed a resolution in April 2026 opposing open-loop evaporative cooling systems for new data centers, after water consumption became the central objection to AI infrastructure projects in the water-stressed Rio Grande Valley. Fluidstack spokesperson Kate Franko said the company is paying its own way on "our energy, our infrastructure, our water" and full local property taxes, with no tax abatement attached to the project. U.S. Rep. Vicente Gonzalez had separately raised concerns about the region's water supply given how water-intensive data centers can be -- underscoring why the cooling design, not the dollar figure, was the condition local officials cared about.

The financing behind it, and what it isn't

Fluidstack raised an $830 million Series A in July 2026 at a $7.5 billion valuation, and separately is reportedly in talks with the Pentagon's Office of Strategic Capital for a roughly $5 billion loan aimed at data-center supply-chain manufacturing capacity -- not at financing a specific facility like this one. The Cameron County campus is being funded and built as its own project, a distinction worth keeping straight as Fluidstack's name shows up across multiple, separately structured deals in the same stretch of weeks.

Sources

]]> Ayar Labs adds $150M more, bringing 2026 funding to $650M for optical AI links https://usemeritai.com/news/ayar-labs-150-million-series-e-extension https://usemeritai.com/news/ayar-labs-150-million-series-e-extension Thu, 10 Sep 2026 12:00:00 GMT Funding The co-packaged optics startup is racing to move from prototype to high-volume manufacturing as copper interconnects become the limiting factor for scaling GPU clusters.

Ayar Labs announced on September 10, 2026 that it has raised an additional $150 million, bringing its total primary capital raised in 2026 to $650 million. The extension adds to the $500 million Series E the co-packaged optics startup closed in March, which was led by Neuberger Berman with participation from Nvidia, AMD, MediaTek, and Wiwynn. Ayar Labs' TeraPHY technology replaces copper interconnects with fiber optics inside AI data centers, aiming to let GPU clusters scale past the power and thermal limits copper runs into.

Why copper is the bottleneck now

"Copper interconnect is becoming the limiting factor for AI scale-up," said co-founder and CEO Mark Wade. As GPU clusters grow to tens of thousands of chips, the distance and power budget copper wiring can support inside a rack becomes a hard ceiling on cluster size; Ayar Labs sells its optical chip as a drop-in component that hardware partners integrate into their own products rather than a rack-scale system of its own. Wiwynn CEO William Lin, whose company took a strategic stake in the March round, called co-packaged optics "a foundational technology for the next generation of AI and cloud data centers."

From prototype to production line

The new capital is earmarked specifically for the harder, less glamorous half of the CPO pitch: high-volume manufacturing. Ayar Labs says it will fund product validation and scale-up of its manufacturing ecosystem as it moves TeraPHY from limited deployments toward mass production for hyperscale customers. Chip-adjacent infrastructure startups routinely raise large rounds on the strength of a working demo; this money is meant to get Ayar Labs through the real test: shipping optical interconnects at the volume Nvidia- and AMD-scale GPU clusters require.

Sources

]]>