Covers Tuesday, Nov 3 (after the morning Daily Brief) through early Wednesday, Nov 4 (Eastern), plus two Monday rounds our Tuesday edition missed (marked "catch-up"). The theme of the day is ownership. Elad Gil led a $350M round for Applied Compute at $3.25B. The company helps enterprises train and run their own AI models, and it says it reached a $100M annualized run rate in under 18 months. Traversal, which builds AI agents that diagnose production outages, raised $120M at $1B on the same day PagerDuty made its own SRE agent generally available. fal confirmed a Bessemer-led Series F, but neither its announcement post nor Bessemer's gives the size or price. Eleven financings cleared our checks. Each is confirmed by the company, its lead investor or both. Figures companies report about themselves are labeled company-reported.

Applied Compute said on Monday, Nov 2, that it raised $350 million at a $3.25 billion valuation in a round led by Elad Gil (Applied Compute · SiliconANGLE). Seventeen other investors joined, among them Kleiner Perkins, Lux Capital, NVIDIA, Blackstone, Shopify Ventures and HubSpot Ventures. The post gives no round letter. The company helps businesses train, serve and keep improving their own AI models, and its pitch leans on open-weight models as the starting point. It says annualized revenue run rate went from zero to $100 million in less than 18 months, and that the number of enterprises it serves has more than doubled since its previous raise (company-reported). Named users include Microsoft, Shopify, Mercor, Harvey, DoorDash, Cognition, Clay and Decagon. Its AC2 platform, launched about three months ago, combines training, inference and continual improvement. Applied Compute says customers have run more than 1,700 training runs on it since July, that its inference service handles over 1 trillion tokens a day, and that 40% of customers now buy both training and inference (company-reported). Two examples it gives: a model for Harvey's Review Table product that improved answer and citation quality at under half the inference cost of a frontier alternative, and a router for Clay that cut inference costs by up to 47% (company-reported). The money goes to compute capacity, hiring and enterprise deployment. Why it matters for investors: The price is about 32.5 times the stated run rate (our arithmetic). That multiple assumes the business is a recurring infrastructure business more than a research consultancy. The mix is the key question. Training projects behave like services work, and inference revenue recurs but carries GPU costs. It also matters that Harvey and Mercor, two of the customers named here, were also named by Trajectory, which Sequoia valued at $300M on Monday. The same handful of AI-native buyers are underwriting several "own your model" vendors at once, and NVIDIA is on both cap tables. Diligence questions: What share of the $100M run rate is inference rather than project-based training? What is gross margin after compute? How concentrated is revenue among the named logos? What happens to the pitch if frontier API prices keep falling faster than open-weight serving costs?

Historical lens. Selling services around freely available software is an old model. Red Hat's IPO registration statement, filed June 4, 1999, said its Linux products could be downloaded for free but came without technical support or printed documentation in that form. Its revenue for the fiscal year ended Feb. 28, 1999, was about $10.8 million, mostly from sales of Official Red Hat Linux, its paid edition (Red Hat S-1). For fiscal 2012, which ended Feb. 29, 2012, Red Hat reported $1.13 billion in revenue, $965.6 million of it from subscriptions, and its CEO called it the first pure-play open-source company to pass the billion-dollar mark (Red Hat). IBM completed its purchase of the company on July 9, 2019, paying $190 a share in cash, or roughly $34 billion in equity value (IBM). Our interpretation, not established fact: Applied Compute is running a version of that playbook on top of open-weight models. The base is free, and customers pay for fitting it to their work and keeping it running. The big difference is cost. Red Hat's subscriptions were mostly software and support, while every inference token Applied Compute serves needs GPUs.

Traversal, a New York company building AI agents for site reliability engineering, announced $120 million in Series B funding at a $1 billion valuation. Kleiner Perkins, Redpoint Ventures and Khosla Ventures co-led, and Sequoia Capital, Battery Ventures and Amex Ventures participated (Traversal blog · Traversal release via Yahoo Finance · SiliconANGLE). Kleiner, which led the Series A, and Sequoia, which led the seed, both added to their positions. Sequoia used its growth fund. Total funding is $173 million. That includes $48 million when Traversal left stealth in June 2025 and a strategic investment from Amex Ventures that December. Three of the founders came out of causal machine learning research at MIT, Berkeley, Columbia and Cornell. Their product maps a company's telemetry, code and deployment history into what they call a Production World Model, then tests causes rather than guessing from correlations. Customers named include Nubank, Harvey, DigitalOcean and PepsiCo. Traversal says ARR is up 9x so far this year, usage is up nearly 100x and monthly active users nearly 20x, and customers ran more than 150,000 investigations last quarter (company-reported). Why it matters for investors: The thesis is that coding agents produce more code than humans can keep running, so demand moves from writing software to operating it. The competition is arriving just as fast. On the same day, PagerDuty (NYSE: PD) made its own SRE agent, Paige, generally available on its renamed PD Reliability Platform (see Notable launches). Traversal has to win against an incumbent that already holds the on-call workflow and says it serves two-thirds of the Fortune 100. A 9x growth rate with no disclosed starting figure says little about scale. Diligence questions: What is ARR in dollars, and how much comes from the five largest accounts? How often does the agent actually fix an incident, rather than just diagnose it? How exposed is Traversal to observability vendors that control the telemetry it reads?

fal, which runs image, video and audio generation models for developers and studios, announced its Series F on Tuesday. Bessemer Venture Partners led, NVIDIA invested again, and the round was announced together with a Sequoia-led Series E that closed earlier this year (fal · Bessemer · TechCrunch). Neither fal's post nor Bessemer's gives the amount or valuation of either round. The figures in circulation come from press reports, not from either primary, so we list both rounds as undisclosed. What the primaries do say: fal is investing in inference capacity and in its team, and the two rounds fund its work in 2027. Bessemer says it first invested at the Series B two years ago and has joined every round since. It also says companies including Adobe, Canva, Shopify, World Labs and Amazon have moved production volume to fal (Bessemer-reported). Bessemer describes fal as two-sided. Developers reach thousands of models through one API, and model labs use fal for distribution and serverless hosting. Why it matters for investors: Until the size and price appear in a primary document, outside investors should treat secondary-market pricing claims about fal as unverified. The business question is structural. fal sits between model labs and the companies that use their models, so its margin depends on lab terms it does not control and on GPU costs, and the labs could choose to sell direct. Diligence questions: What share of fal's volume runs on its own or post-trained models rather than on third-party labs' models? What is gross margin per generated second of video? How much of the new money is committed to compute, and on what contract terms?

fal's Series E and F are not ranked because neither fal nor Bessemer disclosed an amount. See story 3 and the Pro table.

Round notes

Not investment advice. The AI Cap Table is published for information only and is not investment, legal, or tax advice. Nothing in this newsletter is an offer or solicitation to buy or sell any security. Funding data comes from public sources we cite and verify, but errors are possible; do your own due diligence and consult a qualified adviser before making any investment decision.