The AI Bubble Just Keeps Growing - Thanx NVIDIA

The AI ecosystem continues to accelerate in ways that feel less like market evolution and more like engineered inevitability. NVIDIA’s newly announced one‑gigawatt partnership with Thinking Machines Lab is the clearest signal yet that the company is no longer just selling computing, it is architecting the entire upstream and downstream flow of capital, infrastructure, and influence. In my last article, I broke down the circular investment loop where NVIDIA funds AI labs, those labs immediately spend that capital on NVIDIA hardware, and NVIDIA reports record profits. This new deal doesn’t just reinforce that pattern; it industrializes it.

Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, will deploy next‑generation Vera Rubin systems at a scale that rivals national infrastructure projects. A gigawatt of AI compute is not a research cluster , it is a sovereign‑grade capability. And NVIDIA’s “significant investment” in the lab ensures that the demand signal for Rubin hardware is locked in before a single rack is powered on. This is not a coincidence. It is strategy.

OpenAI continues to dominate the narrative but is increasingly constrained by its own governance contradictions. Anthropic, meanwhile, is navigating a very public collision with the Department of Defense after being designated a supply‑chain risk. Their principled refusal to support certain military applications has placed them on the outside of federal procurement pipelines. Whether one agrees with their stance or not, the consequence is clear: the U.S. government is now looking for frontier‑scale partners who will not challenge the mission. NVIDIA understands this and the partnership positions both to fill the vacuum created by Anthropic’s standoff.

The result is a realignment of the frontier AI landscape. NVIDIA is consolidating influence not by choosing winners, but by ensuring that every winner is built on its stack. Thinking Machines gains capital, hardware, and legitimacy. NVIDIA gains guaranteed demand, strategic compliance, and deeper control of the AI supply chain. And the broader ecosystem is left to reconcile with a future where infrastructure, investment, and national‑security alignment are no longer separate conversations but a single, integrated operating model.

In the end, deals like this don’t resolve the AI bubble, they reinforce it. Capital keeps cycling, demand keeps inflating, and the industry keeps mistaking engineered growth for real adoption. As long as infrastructure is built on circular economics rather than genuine productivity gains, the bubble doesn’t correct, it compounds.