Thinking Machines Lab signs multi-billion-dollar deal with Google Cloud

In short: Thinking Machines Lab, led by former OpenAI executive Mira Murati, has signed a multi-billion-dollar agreement to expand its use of Google Cloud for building and running AI.

What happened

Thinking Machines Lab has signed a new deal with Google Cloud that is valued in the single-digit billions of dollars, according to TechCrunch. The agreement covers the computing systems and related services the startup uses to train and deploy its AI models.

A key part of the deal is access to Google’s newest AI computing systems that use Nvidia’s GB300 chips. These are specialized chips used to do the heavy math behind AI, similar to having a much faster engine for training and running models. Google says these GB300-based systems can be about twice as fast for training and serving AI compared with the previous generation.

This is the first time Thinking Machines Lab has signed a major contract with a cloud provider. The deal is not exclusive, meaning the startup can still use other cloud companies over time.

The news also sheds some light on what the company is working on. Google said it can support the startup’s “reinforcement learning” work, which is a way to train AI through practice and feedback (like teaching a dog tricks by rewarding good behavior). TechCrunch previously reported that Thinking Machines launched a product called Tinker, which helps automate making custom, high-end AI models.

Why it matters

Big AI systems require huge amounts of computing power, which is expensive and hard to secure. Deals like this show how cloud companies such as Google are competing to sign fast-growing AI labs early, and it can shape which tools and services end up being built and how quickly they arrive.

Source: TechCrunch AI