In 2026, Hans Tung and I are talking about 19th-century ships.
“The Dutch, the Spanish, the Portuguese—they sailed around the world with beautiful large ships, but they were wooden wind ships,” said Tung, Notable Capital managing partner and early backer of companies like Airbnb, Coinbase, and Anthropic. “The British were the first to scale ships with the steam engine. Adopting steel technology and applying it to the steam engine—that made the British Navy stronger than the rest of the world combined. Without that buildout—steel ships, the steam engine—we wouldn’t even have the American concept of free trade.”
Tung loves history (as do I, which anyone subjected to whaling ventures-centric writing knows), and I came to him with something that had been irking me: I’ve often heard VCs talk about AI as an “Industrial Revolution-level change,” in a way I’ve frequently found hand-wavey. My question to Tung: What if we took the historical analogue seriously? What does it mean to take that parallel to its full conclusion?
And so Tung started with ships, suggesting a throughline from Portuguese spice routes to Britain’s steel ship global dominance (in 1890, 80% of the world’s ships came from British shipyards). To Tung, this is his first mental model for the technological leap we’re living through as our global infrastructure morphs to meet the age of AI. His case: There’s perhaps too much focus on any one app or company, and that major technological change only creates geopolitical dominance if the infrastructure is there to back it up.
“Wood ships don’t last as well in the long run,” said Tung. “The need for steel ships with steam engines was important to be able to go around the world. The British end up doing this because they invest in technology that allows them to leverage the benefits of the Industrial Revolution. They end up dominating the world for more than 100 years.”
I also asked Tung a question I’ve been asking all sorts of people lately, my own strange version of a Rorschach test perhaps: Are data centers the railroads of the 21st century? His personal answer is, essentially, yes.
“If you don’t have enough compute to meet demand, it doesn’t matter if you have a better model,” Tung told Fortune. “The models that win are the ones that are available.”
Consider, for example, the Civil War, won partially because the railroad infrastructure of the North was far more developed than the South’s. It wasn’t just about having food or troops; it was about getting around.
“It doesn’t matter if you have the resources if you don’t have the [means] to move them,” said Tung. “Most major Civil War battles happened within 24 miles of a railroad… If you don’t have a railroad—or compute—it doesn’t matter if you have better engineers, better politicians, or better thinkers.”
This is, vitally, where geopolitical concerns enter the picture: China has been largely outbuilding the U.S. for decades.
“Over the last almost 20 years, the [amount of] subway and high-speed rail built in China was unprecedented,” said Tung, who’s Taiwanese-American. “Now, China has more high-speed rail than the rest of the world combined, kind of like the North in the Civil War. They executed very, very well. It’s mind-boggling.”
This goes beyond modern railroads: Looking at ships, some stats suggest that over 80% of new shipbuilding orders worldwide go through China. For perspective: Each year, one state-owned Chinese conglomerate (the China State Shipbuilding Corporation) builds more commercial ships than the aggregate U.S. shipbuilding industry has built over the 81 years since World War II ended.
In short, China is uniquely adept at building, at a moment when that especially matters.
For the data center buildout, the picture’s more layered: China currently controls about 24 gigawatts of working data center compute capacity, while the U.S. has at minimum 47 gigawatts by conservative estimates. However, the U.S. is staring down the challenge of executing a historic building boom. (A Columbia Business School professor has estimated that the AI infrastructure buildout could account for an average of 3.6% of U.S. GDP every year until 2032. This buildout, of course, comes with all kinds of concerns, from community pushback to questions about the electric grid’s capacities.)
Tung’s case, ultimately, is this: The best ideas and talent don’t count, not on their own. Amid a sweeping socioeconomic shift driven by technology, it’s not invention that matters most. It’s having the infrastructure to use the AI that you’ve built, effectively and at scale.
“We’re not going to just win,” said Tung. “We have to work hard and make sure we win.”
See you tomorrow,
Allie Garfinkle
X: @agarfinks
Email: alexandra.garfinkle@fortune.com
Submit a deal for the Term Sheet newsletter here.
Joey Abrams curated the deals section of today’s newsletter.
This story was originally featured on Fortune.com