Nvidia Backs Mira Murati's Startup With Gigawatt Chip Deal

Nvidia Commits Gigawatt-Scale Compute to Mira Murati's Vision

Nvidia and Thinking Machines Lab, the AI startup co-founded by former OpenAI chief technology officer Mira Murati, have announced a multi-year strategic partnership that will deploy at least 1 gigawatt of computing hardware powered by Nvidia's forthcoming Vera Rubin chips. The deal also includes a significant investment from Nvidia, though neither company disclosed the specific amount.

At the scale announced, the compute commitment is staggering. Jensen Huang estimated in 2025 that building 1 gigawatt of AI computing capacity costs approximately $50 billion, with GPUs accounting for roughly two-thirds of that expense. The partnership makes Thinking Machines one of the largest single buyers of Nvidia's next-generation hardware, alongside hyperscalers like Microsoft, Meta, and Google.

Next-Gen Vera Rubin GPUs at the Core

Thinking Machines will deploy Nvidia's latest-generation Rubin GPUs, which come in two variants: the Rubin CPX optimized for inference calculations, and the standard Rubin with 336 billion transistors for broader AI training and research workloads. The hardware package also includes Nvidia's Vera CPUs, featuring 88 Arm-based cores supporting 176 threads. Deployment is expected to begin early 2027.

The deal gives Thinking Machines an enormous compute runway as it races to build enterprise AI systems. The startup launched its first product, Tinker, in October 2025 -- a cloud API that enables researchers and developers to create customized versions of open-source large language models using LoRA fine-tuning technology.

From OpenAI CTO to $2B Startup in Under Two Years

Murati founded Thinking Machines in February 2024, less than six months after departing OpenAI where she had overseen the development of ChatGPT, the Sora video generator, and several other flagship products. The startup previously raised $2 billion in seed funding at a $12 billion valuation from investors including Nvidia, AMD, and ServiceNow.

The Nvidia partnership represents a significant escalation of that relationship, moving from financial investment to a deep hardware commitment. For Nvidia, the deal secures a major customer for its Rubin architecture ahead of the platform's commercial launch and ahead of GTC 2026 (March 16-19), where CEO Jensen Huang is expected to showcase the Rubin lineup in detail.

What This Means for Engineers and Job Seekers

The partnership signals that independent AI labs -- not just Big Tech hyperscalers -- are becoming major buyers of cutting-edge GPU infrastructure. For engineers, this creates expanding opportunities in GPU cluster management, distributed training systems, and model optimization at startups that can now compete with the largest companies for compute resources. Thinking Machines' focus on making open-source models customizable also points to growing demand for fine-tuning and model adaptation expertise.