The AI Arms Race Isn’t About Models. It’s About Poured Concrete.
By: Reginald Vance
(SeaPRwire) – Hans Tung just reframed everything we think we know about AI competition. And he did it by talking about ships.
The Notable Capital managing partner has spent decades tracking where money actually lands when empires rise. His answer is blunt and uncomfortable for Silicon Valley’s current narrative. The British didn’t dominate the globe through better naval tactics or smarter admirals. They built steel ships powered by steam engines. Everything else followed infrastructure.
Tung sees a mirror image today. He argues the real question isn’t which AI model performs best on benchmarks. It’s whether a nation has enough physical compute infrastructure to deploy those models at scale when it matters. The models that win are the ones that are available. Period.
Let me put this where it belongs. In the hardware and infrastructure stack, not in the model room. For years the market has priced AI winners as software companies with great algorithms. Tung is saying that logic is backwards. The winner will be whoever finishes building first.
Consider what that actually means. The United States currently controls an estimated 47 gigawatts of working data center capacity. China sits at roughly 24 gigawatts. On paper that looks comfortable. But paper doesn’t win wars or build economies.
China spent nearly two decades executing infrastructure at a pace the world has never seen. They now possess more high-speed rail than the rest of the world combined. Over 80 percent of new commercial shipbuilding orders worldwide run through Chinese yards. One state-owned conglomerate builds more ships annually than the entire U.S. shipbuilding industry has constructed since 1945. That is not a metaphor. That is a track record measured in steel and concrete.
The American buildout faces a fundamentally different reality. A Columbia Business School professor estimates the AI infrastructure expansion could absorb 3.6 percent of U.S. GDP every year until 2032. That kind of capital deployment runs into grid constraints, community opposition, transformer lead times, and a permitting system that moves at a fraction of the speed required. None of these problems show up in a pitch deck.
Think about the Civil War the way Tung does. The North didn’t win because it had better generals or richer politicians. It won because its railroad network let it move resources faster than the South could respond. Most major battles happened within 24 miles of a rail line. Infrastructure determines outcomes long before technology gets a say.
This is the hard truth hanging over every AI infrastructure buildout in America right now. We have world-class engineers. We have deep capital. We have companies like Anthropic, Coinbase, and Airbnb on our roll call. But none of that matters if the grid cannot deliver the power, if the construction pipeline cannot keep pace, or if the permitting process devours three years for a project that should take eighteen months.
Tung’s final line matters more than any of his historical references. He said we are not going to just win. We have to work hard and make sure we win. That is a warning, not a prediction. It acknowledges that infrastructure advantage is earned through execution, not inherited through talent.
The data center buildout is a capital war. It is fought in gigawatts, in transformer shipments, in grid interconnection queues, and in municipal approval timelines. The company with the best model will lose to the company with the most available compute if that model cannot reach users at scale. That lesson is written into every industrial transformation in human history.
Britain proved it in 1890 when 80 percent of the world’s ships came from British yards. China proved it over the last two decades. The question now is whether American execution can close the gap before the architecture of AI dominance is locked in for another generation.
Author bio: Reginald Vance is a venture partner specializing in semiconductor valuation and advanced materials, with over two decades tracking infrastructure capital cycles and hardware supply chain consolidation globally.