After Jensen Huang, Who Gets to Build AI?
Nvidia's next CEO inherits a business that helped create customers for its chips and competitors to the world's biggest clouds. Where Nvidia puts its money next could change who gets to build at scale.
Jensen Huang has resigned as Nvidia's CEO, president, and a member of its board. Nvidia named longtime CFO Colette Kress interim CEO on Thursday, November 12. The board says it expects the product roadmap and customer commitments to continue. The next CEO's biggest decisions may be about which new projects Nvidia is willing to help get off the ground.
Imagine you're running an AI company and need a lot more computing power. You can buy it from a big cloud that might also sell your customers a competing AI service. Or you can rent it from a specialist that makes its living selling compute. Under Huang, Nvidia put money and engineering into making that second option viable at scale. How far Nvidia keeps pushing it is one of the most consequential questions of the post-Jensen era.
Nvidia built another route to its chips
Look at Nebius. When Nvidia announced a $2 billion investment in March 2026, the deal also included engineering work on data centers and deploying new Nvidia systems. Its CoreWeave partnership, announced in January 2026, pairs another $2 billion equity investment with work on infrastructure and software. These companies give customers another place to rent Nvidia chips. They also give Nvidia a way to reach buyers beyond the biggest cloud platforms.
That matters even to a company that never rents a single GPU from a specialist. A credible second supplier gives a buyer room to negotiate on price, delivery dates, and the size of a long-term commitment. It gives a startup somewhere to go when its usual cloud is booked. If Nvidia becomes more selective about financing those suppliers, the first thing customers might notice is fewer credible bids when they go shopping. That can happen while Nvidia is still selling every chip it can ship.
There's a trade-off for Nvidia's board. Funding an independent cloud can broaden the market for Nvidia hardware. It also puts Nvidia's own money at risk if the data center takes longer to fill or its customers struggle to pay. A deal can look expensive on its own and still be worth doing because of who it keeps in the market. That's the judgment Kress inherits.
Nvidia's biggest buyers are building their own options
Anthropic says it trains and runs Claude on Nvidia GPUs, Google TPUs, and Amazon Trainium. OpenAI announced plans to deploy AMD GPUs and is working with Broadcom on accelerators of its own. A lab can keep using Nvidia where it gets the best results and move a well-understood, high-volume workload to another system when the economics work. Every credible alternative gives it more leverage the next time it sits down with Nvidia.
Nvidia has an answer to that, too. Amazon is integrating Nvidia's NVLink Fusion into infrastructure for its next-generation Trainium4 chip. NVLink connects chips so they can work together inside a system. Amazon can design the accelerator itself and still use Nvidia technology to connect the rack. Nvidia's next CEO has to decide how much of that rack the company wants to sell and where it's willing to let customers bring their own hardware.
The two strategies can coexist. Supporting independent clouds gives Nvidia more customers for its full systems. Working with Amazon on custom chips lets Nvidia stay in the design as a major customer builds more for itself. If Nvidia leans harder into the Amazon model and puts less money behind the specialists, it could have a thriving business inside a cloud market with fewer powerful suppliers. That would feel very different to a startup trying to get its first big cluster.
The first deals will tell us more than the first forecast
When Nvidia reports on November 17, the sales outlook will tell us what management expects to ship. I'll be listening for where Kress expects the next wave of capacity to go. Is Nvidia still willing to put capital behind independent clouds? How much of the newest hardware will those partners be able to offer? The clearest answers may come after the call, when a project gets Nvidia's money or a cloud provider tells customers where they can actually rent the newest systems.
AI can keep growing while the number of companies able to build at scale shrinks. The answer to what changes after Jensen Huang may be sitting in the next AI founder's inbox. How many companies can quote the job, deliver the chips on time, and afford to compete for the business?
Sources
NVIDIA — NVIDIA and Nebius Partner to Scale Full-Stack AI Cloud
CoreWeave — NVIDIA and CoreWeave Strengthen Collaboration to Accelerate Buildout of AI Factories
NVIDIA Technical Blog — AWS Integrates AI Infrastructure with NVIDIA NVLink Fusion for Trainium4
Reuters — Nvidia invests $2 billion in CoreWeave to boost data center build-out
Data Center Frontier — Inside Anthropic's Multi-Cloud AI Factory
Rich Washburn is a technologist, strategist, and Founder & Chief AI Architect of ARIA AI Labs, working at the intersection of AI, infrastructure, communications, and capital. He also serves as Managing Partner and Chief AI Officer at Eliakim Capital.
Comments