Modal and Baseten in Funding Talks as AI Inference Demand Grows

Modal and Baseten, two startups that help businesses run artificial intelligence models, are in talks to raise new funding, Bloomberg reports.

Both companies operate in a segment of the AI market known as inference — the business of running trained AI models in production, rather than building them. That layer of the stack has drawn growing attention from investors as companies move AI systems out of experiments and into daily operations.

Bloomberg's report, based on sources familiar with the discussions, did not specify the size of the rounds, the valuations under discussion, or the investors involved. Neither Modal nor Baseten has publicly confirmed the talks, and terms may change or the negotiations may not result in completed deals.

The report lands amid a broader shift in where investors see value in the AI supply chain. The first wave of AI venture capital concentrated on model developers — companies such as OpenAI and Anthropic that build foundation models. Attention has since spread to the infrastructure that sits between those models and the businesses that want to use them.

That is where Modal and Baseten position themselves. Rather than selling AI models of their own, they sell the plumbing: services that let engineering teams deploy, run, and scale AI workloads without building and managing the underlying computing infrastructure themselves.

The pitch addresses a real operational problem. Companies adopting AI face rising costs for the computing power needed to serve models to users, and they need reliable ways to handle unpredictable demand. Startups that can lower the cost and complexity of that work — or make it faster — have become acquisition and investment targets.

Bloomberg's reporting on the funding talks signals that investors continue to put money into this layer even as they grow more selective about AI investments overall. Earlier rounds across the sector rewarded raw model development and headline valuations. Current conversations, by contrast, increasingly focus on companies with clear enterprise use cases and infrastructure that businesses will pay to use continuously.

Inference is also a market with structural tailwinds. Every application built on top of a large language model generates ongoing compute demand each time a user queries it. Unlike training, which happens once, inference recurs — making it a recurring-revenue business for the infrastructure providers that handle it well.

For Modal and Baseten, new capital would fund expansion in a market where scale matters. Running AI workloads efficiently requires access to computing capacity, typically graphics processors supplied by providers such as Nvidia, and competing against larger cloud platforms — Amazon Web Services, Microsoft Azure, and Google Cloud — demands sustained investment.

The competitive pressure cuts both ways. The hyperscalers offer their own AI-serving tools and bundle them with their cloud businesses. Specialized startups counter with developer experience, simpler pricing, and infrastructure tuned specifically for AI workloads rather than general-purpose computing.

Investor appetite for that specialization has been visible across the sector. Companies building tools for running, serving, and optimizing AI models have raised significant sums over the past two years, as enterprises that moved quickly to adopt AI discovered that operationalizing it was harder and more expensive than expected.

Whether Modal and Baseten close their rounds, and on what terms, will add another data point to how the market values inference infrastructure. Bloomberg's report of active talks at both companies at the same time suggests investors see the category as one worth backing now, before the winners consolidate.

The next signal to watch is whether either company confirms a deal — and at what valuation. Those numbers will indicate whether the market for AI-running infrastructure is still climbing toward peak enthusiasm or already settling into a more measured phase.