Who's Paying for America's Fabs?

Chip Alliance

The basic contours of modern trade deals are becoming familiar. One country threatens another with tariffs, and in exchange for avoiding those tariffs, the second country agrees to buy more goods or, increasingly, to invest vast sums of money in the first country. That pattern was on full display this week, when Japan, facing the threat of new U.S. tariffs, agreed to a sweeping $550 billion investment package directed at the American economy.

But the arrangement comes with a peculiar wrinkle. It turns out the money isn't just for building Japanese car factories in Ohio. It's also for building Taiwanese semiconductor factories in Arizona. This raises the obvious question: why would Japan pay for Taiwan to build factories in America. Reuters reported on the details, quoting Japan’s top trade negotiator, Ryosei Akazawa:

"Japan, the United States, and like-minded countries are working together to build supply chains in sectors critical to economic security," Akazawa told public broadcaster NHK.

To that end, he said projects eligible for financing under the package are not limited to U.S. or Japanese firms.

"For example, if a Taiwanese chipmaker builds a plant in the U.S. and uses Japanese components or tailors its products to meet Japanese needs, that's fine too," he said, without specifying companies.

The unnamed "Taiwanese chipmaker" is, of course, Taiwan Semiconductor Manufacturing Co., the world's sole mass-producer of the advanced chips that power the AI revolution. The logic of this trilateral arrangement is both simple and profoundly revealing of the new economic era. The U.S. desperately wants TSMC’s advanced manufacturing on its own soil to secure its supply chain from potential disruption in Asia. TSMC, in turn, needs hundreds of billions of dollars to fund these astronomically expensive new fabs. And Japan, facing the economic threat of tariffs, happens to have a very large checkbook.

This is what the new era of "friend-shoring" looks like: a messy, expensive, multinational scramble where avoiding tariffs might mean financing your ally’s most important supplier on another continent’s soil. Each party brings something essential to the table—American land, Taiwanese technology, and Japanese capital—all in the service of building a semiconductor supply chain that exists outside of China’s shadow. The deal is less a straightforward investment and more the financial architecture of a new strategic alliance, forged by the shared anxieties of a tech-driven cold war.

Promise Fatigue

There is a type of CEO whose most valuable asset is not their operational skill or their financial acumen, but their ability to tell a really good story about the future. For years, Elon Musk has been the undisputed master of this genre. His story—that Tesla is not merely a car company but a world-changing AI and robotics juggernaut on the cusp of solving full autonomy—has been so compelling that it has powered the company to a valuation that often seems completely detached from the mundane business of selling cars for a profit.

The story has been, for a long time, a bankable asset. But assets can depreciate. Last week, on Tesla's quarterly earnings call, Musk tried to tell the story again, promising a vast robotaxi network and a future $20 trillion valuation.¹ The market's reaction was, let's say, unenthusiastic. The stock dropped 8%. It seems investors may be suffering from promise fatigue.

Even the company's biggest supporters are starting to sound impatient. Here are the analysts at Canaccord Genuity, in a note quoted by CNBC, who are still bullish on the stock:

"Look, we love robotaxis. And robots. Over time, Tesla is well positioned to benefit from these future-forward opportunities."

The analysts, however, said that they're focused on the profit and loss statement, writing: "But we love growth too, in the here and now. We need the P&L dynamics to turn."

This is a polite way of saying, "We are tired of the story; please show us the spreadsheet." The problem is that the story and the spreadsheet are telling two very different tales. The story is about a company on the verge of deploying autonomous vehicles to half the U.S. population. The spreadsheet shows a company whose car sales are down 16% and which still requires a human "safety supervisor" in its robotaxis.

And then there is the competition. While Musk was talking about what Tesla will do, Alphabet disclosed what its Waymo unit has done: driven over 100 million fully autonomous miles on public roads. Tesla's Austin trial, by comparison, has logged about 7,000 miles. When Musk teases a robotaxi launch in San Francisco, it's useful to know that, according to state regulators, Tesla hasn't even applied for the necessary permits yet.

For a long time, you could ignore these details because the story was so good. The narrative was that Tesla's millions of cars were collecting more "real world" data than anyone else, and that this data was the key to unlocking a general-purpose AI that would leave Waymo's geofenced approach in the dust. The market seems to be concluding that maybe 100 million miles of a car actually driving itself is a more useful data point than billions of miles of a car that needs a human to keep it from hitting things.

What is the correct discount rate for a promise that has been repeatedly broken? How do you model the terminal value of a story that people are starting to find a little dull? These are now, apparently, the central questions in valuing Tesla. It's a bit of a problem when your main intangible asset—the CEO's credibility—starts looking like a depreciating one.

¹ That is more than the combined market capitalization of Apple, Microsoft, Alphabet and Nvidia. So, you know, an ambitious target.

The Scoreboard

• AI: Bytedance’s AI Robot System Can Fold Clothes and Do Housework (Euronews)
• Semiconductor: Intel Drops 8% as Chipmaker’s Foundry Business Axes Projects, Struggles to Find Customers (CNBC)
• Data Center: Equinix to Boost Japan Urban Data Centers on 'Edge' AI Demand (Nikkei Asia)
• Internet: Is AI Killing Google Search? It Might Be Doing the Opposite (WSJ)