AI infrastructure startups are entering a new phase of the generative AI boom, with investor attention rapidly shifting from model training to inference. At the centre of this transition is San Francisco-based Modal Labs, which is reportedly in discussions to raise fresh capital at a valuation of around $2.5 billion.
If completed at that level, the new round would more than double Modal Labs’ $1.1 billion valuation secured less than five months ago following its $87 million Series B. General Catalyst is said to be in talks to lead the investment, according to industry reports.
The reported valuation surge reflects a broader realisation across the AI ecosystem: inference is becoming core digital infrastructure.
Why AI inference is the new battleground
In the early stages of the generative AI wave, most capital flowed into model training. Building ever-larger foundation models required vast compute resources, specialist hardware and significant upfront investment.
Now, as enterprises deploy AI into products and workflows at scale, the economics of inference have become more critical. AI inference refers to running trained models in production to generate responses, recommendations or automated actions in real time.
For businesses serving millions of queries per day, inference costs can quickly outpace training expenses. Even small improvements in latency or compute efficiency can translate into substantial margin gains.
This shift has turned inference optimisation into one of the most competitive segments in AI infrastructure.
What Modal Labs does?
Founded in 2021 by Erik Bernhardsson, former head of data at Spotify and former CTO of Better.com, Modal Labs provides infrastructure designed to make it easier and more efficient for developers to run AI workloads in production.
The company focuses on enabling scalable, on-demand compute for machine learning models, helping teams deploy inference workloads without managing complex infrastructure themselves. By abstracting away infrastructure complexity, Modal aims to allow developers to focus on building AI applications while optimising cost and performance behind the scenes.
Modal Labs has attracted backing from prominent venture firms including Lux Capital and Redpoint Ventures. Its annualised revenue run rate is reportedly around $50 million, signalling strong commercial traction for a company founded just four years ago.
Inference valuations are accelerating
Modal Labs is not alone in riding the inference wave. The sector has seen a sharp uptick in large funding rounds and valuation jumps.
Baseten recently raised $300 million, pushing its valuation from $2.1 billion to $5 billion within months. Fireworks AI secured $250 million at a reported $4 billion valuation. Together, these rounds signal that investors increasingly view inference platforms as long-term infrastructure bets rather than short-term tooling plays.
If Modal Labs reaches a $2.5 billion valuation, it would further cement inference as one of the most aggressively funded areas in AI.
From tools to infrastructure layer
The strategic question for investors is whether inference platforms like Modal Labs can evolve into foundational infrastructure providers, similar to cloud platforms in earlier computing cycles.
As AI adoption expands across enterprises, inference optimisation touches nearly every AI-powered product. Companies that can consistently deliver lower latency, reduced compute costs and seamless scalability may become deeply embedded in the software stack.
However, competition remains intense. Cloud hyperscalers are also investing heavily in AI inference optimisation, while startups are racing to differentiate through developer experience, pricing models and hardware integrations.
Can Modal justify a $2.5B valuation?
With a reported $50 million ARR and strong growth momentum, Modal Labs appears well positioned within the inference gold rush. A $2.5 billion valuation would imply significant confidence in continued revenue expansion and long-term infrastructure defensibility.
As generative AI moves from experimentation to mission-critical deployment, inference is emerging as the economic backbone of the ecosystem. If Modal Labs can scale alongside enterprise AI adoption, its valuation trajectory may reflect a broader truth: in the AI era, running models efficiently may prove just as valuable as building them.
Also Read: Trener Robotics Raises €26M Series A to Build the “Physical AI” Layer for Industrial Automation