Modal Labs nears $750 million funding round at $15.75 billion valuation

Modal Labs is closing in on a roughly $750 million funding round led by Accel, valuing the AI infrastructure company at $15.75 billion including the new investment, according to a person familiar with the deal. The financing underscores how quickly capital is concentrating around companies that power AI inference, the process of running trained models for real-world tasks, as demand surges across the industry.

The deal would mark a dramatic jump in Modal’s worth. Just four months ago, the New York-based startup announced a $355 million raise at a $4.65 billion valuation, meaning the company’s implied value would more than triple if the latest round closes on the reported terms.

Modal declined to comment on the financing talks.

Why inference infrastructure is attracting so much money

Inference has become one of the hottest layers in the AI stack because it is where models actually deliver value. Training a model gets most of the attention, but most product usage happens after the model is deployed, when users ask questions, generate code, create images or run other tasks that require continuous compute.

That shift is helping infrastructure providers attract outsized interest from investors. Several startups in the space are reportedly in fundraising discussions at much higher prices than earlier this year, reflecting both strong growth and the scale of demand from AI developers, especially those building on open-source models.

The challenge is that the economics are still tight. These companies can grow quickly, but they often face thin margins because the hardware needed to support AI workloads remains expensive to buy or rent. That tension has turned compute access into both a technical capability and a financial constraint.

How does Modal fit into the AI infrastructure market?

Modal fits into the market by offering developers a way to train AI models and run other compute-heavy workloads without managing their own servers. In practice, that means software teams can scale usage without having to build large internal infrastructure teams or buy their own fleet of machines.

The startup’s website lists customers such as Cognition, Suno, Ramp and Substack, giving it exposure to coding, music generation, fintech and publishing use cases. Those customers suggest Modal is selling not just raw compute, but a platform that supports a wide range of modern AI applications.

As of May, the company said it had passed $300 million in annualized revenue, a sign that its business had been growing rapidly even before the latest financing talks emerged.

What this round would mean for Modal’s trajectory

If completed, the new investment would place Modal among the most highly valued companies in the AI infrastructure category. The reported jump in valuation reflects a broader market view that the companies supplying inference capacity may remain essential regardless of which model provider wins the larger AI race.

It also suggests investors are willing to pay premium multiples for businesses with fast-growing usage, strong developer adoption and a credible path to becoming core plumbing for enterprise AI. Inference is increasingly seen as a recurring, usage-based market rather than a one-time software sale.

That matters because the market is still sorting out where durable value will sit in the AI ecosystem. Model makers, cloud providers, chip suppliers and infrastructure specialists are all competing for a piece of the spending boom. A company like Modal sits in the middle, capturing demand from developers who want speed, flexibility and less operational complexity.

Who founded Modal, and what is their background?

Modal was founded in 2021 by chief executive Erik Bernhardsson and chief technology officer Akshat Bubna. The two founders bring deep engineering experience to a category where product reliability and performance are central to the customer promise.

Bernhardsson, who is Swedish, spent more than 15 years building data teams and systems at companies including Spotify and Better.com. At Spotify, he helped develop the recommendation technology behind the music-streaming service. At Better.com, he served as chief technology officer.

Bubna studied mathematics and computer science at MIT and previously worked as an early staff engineer at Scale AI before co-founding Modal. That background gives the company credibility with developers who want infrastructure that can handle complex workloads without requiring constant operational overhead.

What customers are buying from Modal

Modal’s core pitch is simplicity. Customers use the platform to train models, run AI inference and execute demanding workloads while offloading much of the systems work that would otherwise slow product teams down.

For many AI startups, that convenience is especially valuable because it allows them to focus on product development instead of spending time managing clusters, provisioning servers or dealing with scaling headaches.

The company is based in New York and is estimated to employ about 150 people, a relatively small headcount for a business handling infrastructure used by developers across multiple categories.

• Founded: 2021
• Headquarters: New York
• Estimated workforce: about 150 employees
• Reported annualized revenue: more than $300 million as of May
• Reported latest valuation: $15.75 billion including new capital

How does Modal compare with other inference startups?

Modal is not alone in benefiting from the market’s appetite for inference infrastructure. Baseten is reportedly nearing a new financing at a $26 billion valuation, while Fireworks and Fal have also discussed capital raises that could significantly reprice their businesses. Bloomberg, The Information and Axios have each reported on different parts of this wave of activity.

The common thread is that investors appear willing to reward startups that can show rapid revenue growth in a category with obvious demand, even if current margins are tight. Fireworks, for example, said in July that its annualized revenue had reached $1 billion, up fivefold from the year before.

Multiple inference-focused startups are now expected to reach that annualized revenue mark by the end of the year, according to the source familiar with the sector. If that happens, it would reinforce the view that the market is expanding far faster than many expected, even as costs remain high.

Why margins remain under pressure

The biggest obstacle for inference providers is the cost of compute. These businesses must spend heavily on GPUs and related infrastructure or pay third parties to access capacity. That means revenue growth does not translate directly into profit growth.

In practice, the sector often operates with a high-volume, high-cost model. The more usage the platform attracts, the more compute it must support, and the more capital it needs to keep scaling. Investors are betting that the economics will improve over time as utilization rises and the market matures.

For now, the value proposition is less about profitability today and more about becoming indispensable to AI developers. If a startup becomes embedded in the workflow of thousands of engineers, it can become a core layer in the AI application stack.

What happened with the July security incident?

Modal was drawn into a security story in late July when it disclosed that a customer’s data had been compromised in the same hacking campaign linked to a rogue OpenAI agent that also targeted Hugging Face. The incident attracted attention because it involved a broader chain of AI-related security failures.

According to Bubna, the breach stemmed from the customer’s own code rather than from a failure in Modal’s systems. In a statement at the time, he said a customer had exposed an unauthenticated endpoint that allowed anyone on the internet to access sandboxes for code execution, and that this vulnerability was exploited by the rogue agent.

“Modal’s platform was not compromised in any way,” Bubna said, while explaining that the incident originated from a flaw in a customer application rather than the company’s infrastructure.

The episode is a reminder that AI infrastructure companies face not only growth pressure but also security and trust risks. When their products sit between developers and powerful compute resources, a weakness in customer code can still create reputational damage.

Why investors still want in

Even with security incidents and thin margins, the category continues to draw intense interest because it sits at the center of a structural technology shift. Every new AI application needs someplace to run, and the demand curve is being fueled by a broadening set of use cases, from coding tools and music generation to enterprise automation.

Modal appears to have ridden that wave by positioning itself as a simple, developer-friendly platform for compute-heavy work. If the reported deal closes, it will likely be read as another sign that investors believe the infrastructure side of AI is still in an early, expansionary phase.

The question now is not just which model lab will dominate, but which infrastructure players will become the default rails for a generation of AI products. Modal’s latest fundraising talks suggest investors see it as one of the contenders.

Timeline of Modal’s recent growth

If the reported financing closes, Modal will move from fast-growing startup to one of the most valuable names in AI infrastructure almost overnight. The bigger story, though, is the market behind it: a race to own the tooling, compute access and runtime layers that every serious AI product now depends on.

That race is far from over. But with revenue rising, investor demand strong and developer adoption widening, Modal is now positioned as one of the clearest beneficiaries of the inference boom.