Baseten raises $1.5 billion at a $13 billion valuation, betting on cheap AI models

US startup Baseten has raised $1.5 billion in a Series F round at a $13 billion valuation. It is one of the largest rounds of the year in AI infrastructure, and a clear signal of where the market is heading.

What Baseten is

Baseten is not another language model, nor an LLM startup in the usual sense. The company works at the infrastructure layer: it provides software and compute to organizations that want to deploy and scale AI models themselves, especially cheaper open ones.

Put simply, Baseten solves a problem more and more companies face: how to take an open model such as Mistral, Llama or Gemma and quickly run it in production without building infrastructure from scratch.

The platform provides:

• A ready-made environment for deploying AI models
• Compute optimized for inference
• Tools for scaling under variable load
• An API layer for integration into customers' products

Who invested and why it matters

Altimeter led the Series F. It is one of the most active tech funds in AI. Its involvement is telling: the fund is known for cautious but large bets on infrastructure companies with a clear, proven business model.

The $13 billion valuation makes Baseten one of the most expensive startups in AI infrastructure, alongside names such as Together AI, Modal and Replicate. The market is clearly consolidating around a few large platforms.

Why cheap models became a strategic asset

Baseten's main bet is the trend toward economical models. Over the past 12–18 months the market has shifted clearly: companies that initially built products on GPT-4 or Claude are starting to migrate to more affordable alternatives. There are several reasons:

• Inference on open models costs several times less than on proprietary ones
• The quality of Llama 3, Mistral Large and similar models has reached a level sufficient for most business tasks
• Local deployment lets companies keep data inside their own perimeter, which is essential for corporations and regulated industries

"Companies are starting to realize they don't need the most expensive models for most tasks,"

Baseten CEO Tuhin Srivastava said in an interview with Bloomberg.

That is exactly where Baseten sits: between model developers and the companies that want to use their models. While AWS, Google Cloud and Azure promote their own AI services, Baseten offers a vendor-neutral platform: run any model, scale it to your requirements.

What it means

Baseten's round reflects a broader market logic: the center of gravity in AI is shifting from building models to running them efficiently. The infrastructure layer, meaning hosting, inference optimization and compute management, is becoming the key competitive advantage.

Judging by Altimeter's bet, investors are ready to pay $13 billion for whoever builds this infrastructure best.