Nvidia has announced a significant, multiyear partnership with the AI startup Thinking Machines Lab.
At its core, this deal is straightforward: Nvidia will invest a substantial amount into Thinking Machines and provide them with at least a gigawatt of its upcoming Vera Rubin platform systems. For Thinking Machines, this guarantees access to the immense computing power needed to train advanced AI models. For Nvidia, it locks in a major customer for its next-generation hardware, which is crucial given that its data center business now accounts for nearly 90% of its total revenue.
However, this partnership is much more than a simple sales agreement. It's a prime example of Nvidia's highly effective 'capital-plus-compute' flywheel strategy. First, Nvidia identifies promising AI startups and invests in them, providing the capital they need to grow. Second, as part of the deal, these startups commit to using Nvidia's hardware and software ecosystem, like CUDA. This creates a powerful, self-reinforcing cycle: Nvidia's investments fuel the AI industry's growth, and that growth, in turn, drives demand for Nvidia's products, securing its revenue for years to come.
There's another critical layer to this story: geopolitical risk. The U.S. government has been tightening export controls on advanced AI chips, creating uncertainty for companies worldwide that rely on this technology. By partnering with a U.S.-based company like Nvidia for a domestic supply of chips, Thinking Machines is making a savvy strategic move. They are hedging against the risk that future regulations could disrupt their access to essential hardware, thereby safeguarding their long-term research and development roadmap.
In essence, this deal is a win-win, shaped by both Nvidia's brilliant business strategy and the complex global landscape of AI. It showcases how securing computing power has become as critical as securing capital for the next wave of AI innovation.
- Glossary
- Gigawatt: A unit of power equal to one billion watts. In this context, it's used to describe the enormous electrical power required to run the massive data centers needed for AI model training.
- Export Controls: Government regulations that restrict the sale of certain technologies, like advanced AI chips, to other countries, often for national security reasons.
- CUDA: A parallel computing platform and programming model created by Nvidia. It allows software developers to use Nvidia GPUs for general-purpose processing, which is fundamental for accelerating AI and machine learning tasks.