Modal's $750m round puts a $15.75bn price on renting out AI compute
Accel led a Series D that more than triples the inference platform's valuation in five months. Investors are paying for growth in a layer that owns no data centres.
Modal, the New York company that rents out graphics processors to developers through its own software layer, said on 14 October that it had raised a $750m Series D at a post-money valuation of $15.75bn. Accel led the round and Kleiner Perkins joined as a new backer, alongside existing investors including Menlo Ventures, General Catalyst, Redpoint, Lux Capital, Amplify Partners and Bain Capital Ventures. The company also disclosed that it had crossed $1bn in annualised run rate.
The numbers are striking less for their size than for their speed. In May, Modal raised $355m at a $4.65bn valuation and reported annualised revenue just above $300m. Five months later the valuation is more than three times higher and the run rate has roughly tripled. For a venture market that has spent 2026 pouring money into a handful of model laboratories, the deal is a clear signal that investors now see the plumbing beneath those laboratories as a category worth backing at frontier prices.
What Modal sells
Modal began as a way to run Python code in the cloud without managing servers. It has since built its own file system, container runtime and scheduler, and sells access to compute for inference, model training, batch jobs and, increasingly, for AI agents. Its fastest growing product, Sandboxes, gives agents isolated environments in which to execute code. Modal says Sandboxes alone is a $300m run rate business, that it has launched close to 3bn sandboxes, and that the system can scale to a million running at once.
Crucially, the company owns no data centres. In its announcement it says its capacity pool spans more than 20 clouds; in May it described capacity pooled across hundreds of data centres. That is the heart of the investment case and also its principal risk. Modal buys scarce graphics processors from others and resells them, wrapped in software that makes them faster to start and easier to scale. Its gross margin therefore depends on the gap between what it pays for capacity and what customers will pay for convenience, and the company does not disclose either figure.
The growth claims are also the company's own. Modal says usage has tripled since the Series C and that it has grown more than 15 times year on year. These are unaudited, self-reported metrics, which is normal for a private company but worth bearing in mind when a valuation moves this far, this fast.
A crowded and richly priced layer
Modal is not alone. When TechCrunch first reported the round on 28 September, it noted that rival Baseten was nearing new capital at a $26bn valuation, double its level in June, and that Fireworks had said in July that its annualised revenue had reached $1bn, five times its level a year earlier. Fal, which focuses on image and video generation, was also said to be talking to investors. The same report cited a source expecting several inference companies to reach $1bn in revenue by the end of the year.
The economics behind this are straightforward. Once a model is trained, every query costs money to serve, and that cost recurs for as long as the product is in use. Training budgets are lumpy; inference bills scale with customers. Companies such as Modal sit where that recurring spending lands, and their revenue rises with the success of their clients rather than with any single model release. Modal's customer list illustrates the breadth: it has named DoorDash, Cognition, Ramp, Decagon, Suno and Physical Intelligence among its users, and says TypeSafe AI's Jev model served 1tn tokens on its platform in the first three days.
The flip side is that the layer is thin. Large cloud providers and the chip makers themselves are able to offer similar services. If capacity loosens and prices fall, intermediaries may gain; if capacity tightens and suppliers raise prices, they may find it hard to pass on the full cost.
The venture backdrop
The round lands in a market that has rarely been more concentrated. The Q3 2026 PitchBook-NVCA Venture Monitor, the latest edition of the report, put US venture deal value at $515.8bn through the third quarter, with AI companies accounting for 82.7% of that total. Third quarter deal value alone was $98.4bn across an estimated 5,012 deals. Exits remain the weak point: the report counted just 18 US venture-backed initial public offerings in the quarter and 992 active unicorns with a combined valuation of $5.7tn.
Fund managers are concentrated too. US venture funds raised $108.5bn through the third quarter, and vehicles of $500m or more took 78.1% of that capital. Large platforms such as Accel and Kleiner Perkins have the capacity to write cheques of this size, and they increasingly need deals large enough to move the needle for funds of that scale. A company that has gone from about $300m to more than $1bn of annualised revenue in five months is one of the few that can absorb such cheques.
Mid-October activity confirmed the pattern. In the week to 19 October, a tally by RecentFunding counted 197 startups announcing rounds with a disclosed amount, and Modal's $750m was among the largest, alongside an $800m raise by ZeroCarbon and a fintech growth deal for ID Finance. Several of the other big tickets that week went to compute adjacent hardware, from photonic chips at Salience Labs ($125m) to AI cloud capacity at NeoAI Cloud ($445m).
How to read the price
At $15.75bn and a run rate above $1bn, Modal is valued at a little under 16 times annualised revenue. That is rich by any historical standard for infrastructure software, and it leaves little room for disappointment. Investors are in effect betting that inference demand will keep compounding and that Modal's software, rather than its access to hardware, is what customers are paying for.
There are reasons for caution. The run rate figure is a snapshot of recent months annualised, not twelve months of booked revenue, and it can fall as quickly as it rose if a few large customers move workloads elsewhere or build their own capacity. TechCrunch has also reported that in late July a Modal customer's data was compromised in a wider hacking campaign; Modal's chief technology officer said the flaw lay in the customer's code rather than in Modal's systems, but security will matter more as agents run untrusted code at scale.
Our view
The Modal round is a rational bet priced at an exuberant level. The company has shown that a software layer over rented chips can grow faster than most model developers, and it has done so with a team of a little more than 200 people. That is an unusually capital efficient profile in a sector defined by enormous spending.
Yet the valuation assumes that today's scarcity of compute persists long enough for Modal to entrench itself, and that its margins survive a market in which suppliers and the largest clouds are also competitors. Founders building on top of such platforms should note that their own costs are now someone else's growth story, and negotiate accordingly. Investors should watch three things: whether Modal or its peers begin to disclose gross margins, whether Baseten's reported round closes at the $26bn mark, and whether any of these companies chooses the public market rather than another private round. With 992 unicorns waiting for liquidity, a credible infrastructure listing would matter well beyond the inference layer.
Sources
- Modal: Modal crosses $1B annualized run rate and raises $750M Series D
- Modal: Modal's Series C, raising $355M at a $4.65B valuation
- Pillsbury: Pillsbury advises Modal Labs on $750 million financing
- TechCrunch: Inference provider Modal Labs closing in on $750M round at $15.75B valuation
- Cloudnews: Modal tops $1 billion in annualized revenue, raises $750M
- NVCA: PitchBook-NVCA Venture Monitor
- RecentFunding: Startup funding this week, Oct 12 to Oct 19, 2026
Filed under Venture Capital. Written by InvestTimes Editorial under our editorial standards. Commentary and analysis for information only, not investment advice.

