The world's most powerful chip maker just became a diplomat—and that tells you everything about who holds leverage in the AI arms race.
The Summary
- Nvidia CEO Jensen Huang joined Trump on Air Force One to Beijing as a last-minute addition, putting AI front and center at the Trump-Xi summit
- Separately, Anthropic is in early talks to raise at least $30 billion in what would be its largest funding round yet
- The timing isn't coincidence: AI infrastructure and compute access are now bargaining chips in great power negotiations
The Signal
Jensen Huang doesn't get pulled onto Air Force One for photo ops. When the CEO of a $2 trillion company joins a head of state for sensitive bilateral talks, it means the thing he controls—compute—matters more than diplomacy. China needs advanced chips. America controls the supply chain. Nvidia is the chokepoint.
Trump said the summit would focus on trade, not the Iran War. But when you bring Jensen Huang, you're talking about the most strategic trade of all: access to the infrastructure that powers AI. China has been racing to build domestic alternatives to Nvidia's H100 and newer Blackwell chips since export controls tightened. They're years behind. That gap is leverage.
"The CEO of compute just became part of the negotiating team."
Here's what that means in practice:
- China likely wants relaxed export restrictions on AI chips for "civilian" applications
- The US wants concessions on something else—rare earth minerals, manufacturing, or geopolitical positioning
- Nvidia gets clarity on whether it can sell into the world's second-largest AI market
Meanwhile, back in Silicon Valley, Anthropic is shopping a $30 billion raise. That's not just a funding round. That's a war chest. For context, Anthropic raised $7.3 billion across multiple rounds through 2024. A $30 billion round would be 4x that cumulative total—in one shot.
The message: foundation model labs believe the next phase requires capital at nation-state scale. Training runs for frontier models now cost hundreds of millions. The compute clusters to serve them cost billions. Anthropic is preparing for a world where AI infrastructure spending looks less like software companies and more like telecom or energy buildouts.
Two data points, same story: AI is infrastructure now, not software. Infrastructure requires massive capital. Infrastructure gets negotiated at the state level. And the companies that control the picks and shovels—Nvidia for chips, Anthropic and OpenAI for models—are the ones setting terms.
The Implication
Watch what happens after this summit. If export restrictions ease even slightly, Nvidia's stock will move and China's domestic chip companies will take a hit. If restrictions hold, expect Beijing to double down on homegrown alternatives and tighter capital controls on Chinese AI investments flowing to the US.
For founders: if you're building anything that requires serious compute, your supply chain just became a geopolitical risk. Plan accordingly. For investors: the next decade of AI won't be won by the best models alone. It'll be won by whoever controls the infrastructure to train and deploy them at scale. That's why Anthropic wants $30 billion. That's why Jensen Huang is on Air Force One.