Baseten, an artificial intelligence startup that operates AI models for clients using cloud infrastructure, has successfully raised $75 million in funding, which values the company at $825 million.
This funding round highlights the confidence venture capitalists have in the ongoing AI boom, which is benefiting various startups beyond those solely focused on large language models. Recently, major players like OpenAI, Anthropic, and xAI have secured billions in funding, primarily for Nvidia graphics processing units (GPUs) to support their AI operations.
Once companies complete training AI models with extensive data, they need to deploy these models to generate outputs in response to user requests. This is where Baseten steps in. Instead of maintaining its own data centers, Baseten utilizes cloud providers like Amazon and Google to run its software. Clients can also provide their own infrastructure through an enterprise tier. By leveraging multiple providers, Baseten ensures access to more GPUs than any single cloud service can offer.
According to co-founder and CEO Tuhin Srivastava, the key advantage for customers is speed. “You can go to production without worrying about reliability, security, and performance,” he stated. Although companies can manage model deployment independently, securing enough Nvidia chips can be challenging, especially when cloud providers suddenly take GPUs offline for maintenance. Baseten helps clients navigate these challenges seamlessly.
The recent emergence of Chinese AI lab DeepSeek, which claims to train models at significantly lower costs than its U.S. counterparts, has intensified the focus on efficiency in AI. Baseten quickly integrated support for DeepSeek’s R1 reasoning model, which competes with OpenAI’s offerings.
Many organizations are exploring cost-effective alternatives, with Baseten clients often experiencing a reduction in inference costs by over 40%. Founded in 2019 and based in San Francisco, Baseten has grown rapidly, with its revenue increasing sixfold over the past year.
