AI inference startup Baseten in talks to raise $1 billion at $11 billion valuation

The AI infrastructure gold rush is shifting from training giant models to something far less visible but far more expensive: running them at scale. That shift is turning inference startups into some of the hottest companies in Silicon Valley. One of the biggest winners right now appears to be Baseten.

The San Francisco-based AI startup is reportedly in talks to raise $1 billion at an $11 billion valuation, according to a person familiar with the discussions. If completed, the deal would more than double Baseten’s valuation from just three months ago, when the company raised capital at a $5 billion valuation.

“AI startup Baseten has recently been in talks with investors to raise $1 billion at an $11 billion valuation, including the money, according to a person with knowledge of the fundraise. That would more than double the company’s $5 billion valuation from its last round, which was announced just three months ago,” The Information reported.

The speed of the company’s rise says a lot about where the AI market is heading. Training frontier models still grabs headlines. Serving those models to millions of users is becoming the real business.

Founded in 2019 by Amir Haghighat, Tuhin Srivastava, Philip Howes, and Pankaj Gupta, Baseten focuses on AI inference infrastructure. Put simply, the company helps enterprises deploy and run AI models in production without the latency, reliability, and cost headaches that often come with large-scale AI systems.

Its platform supports open-source, custom, and fine-tuned models across cloud, hybrid, and self-hosted environments. Baseten has leaned heavily into developer tooling and performance optimization, positioning itself as infrastructure for companies building AI-native products.

The company’s Truss framework has become one of its best-known tools among developers building compound AI systems and multi-model workflows. Baseten has spent the past year pitching itself as an “AWS for inference,” betting that inference will become one of the largest spending categories in AI.

That bet increasingly looks right.

Why Investors Are Betting on Baseten

Analysts expect inference to account for roughly two-thirds of AI compute demand by the end of 2026, up sharply from just one-third in 2023. The shift accelerated after the Chinese AI lab DeepSeek released its R1 reasoning model earlier this year, prompting the industry to focus more on performance efficiency and lower serving costs.

Baseten moved quickly to support DeepSeek’s R1 model and has promoted its ability to run advanced reasoning systems at lower cost than OpenAI’s proprietary alternatives.

The startup has quietly built an impressive customer list that includes Notion, Cursor, Writer, Gamma, Patreon, Descript, and HeyGen. The company says it serves more than 100 enterprises and hundreds of smaller businesses.

Baseten’s rise has been fueled in part by close relationships with hardware and cloud giants including NVIDIA, Google Cloud, and Amazon Web Services. The company recently highlighted performance gains on NVIDIA’s Blackwell GPUs as competition intensifies around AI serving costs.

Investors have poured money into the company at a staggering pace.

Inside Baseten’s rise

In September 2025, Baseten raised $150 million in a Series D round at a $2.15 billion valuation led by BOND with participation from CapitalG and Premji Invest.

Just four months later, Baseten announced a $300 million Series E round at a $5 billion valuation led by IVP and CapitalG. That financing included a reported $150 million anchor investment from NVIDIA, pushing the company’s total funding to roughly $585 million.

The latest funding talks suggest investors believe the AI infrastructure boom is still in its early innings.

The company has continued shipping new products during its growth sprint. In May, Baseten partnered with Benchling to support biotech AI workloads with on-demand GPU infrastructure for scientific models. It recently launched Baseten Loops, a reinforcement learning training SDK aimed at frontier AI systems.

Competition in inference infrastructure has intensified over the past year as startups race to become the default layer between foundation models and enterprise applications. Rivals including Together AI are chasing the same opportunity, alongside offerings from major cloud providers.

Baseten’s pitch has centered on ease of deployment, performance optimization, uptime reliability, and flexibility across multiple model types. That combination has resonated with companies trying to move AI systems from demos into production environments that cannot afford downtime or unpredictable costs.

The reported deal has not been finalized, and terms could still change. Still, an $11 billion valuation would cement Baseten as one of the most valuable independent AI infrastructure startups in the market today.

For venture investors, inference is no longer the overlooked plumbing layer of AI.

It is quickly becoming one of the most valuable parts of the stack.