The Lutnick Insult

Saying the Quiet Part Loud

The US-China tech war entered a truly baffling phase. For weeks, Chinese authorities have been cautioning their own tech giants—Tencent, Baidu, ByteDance—against buying Nvidia's newly approved H20 AI chips. The warnings, which cited vague security concerns, made little sense. Why would Beijing lobby against a chip it had just fought hard to get access to?

The answer, it now transpires, lies in what the US Commerce Secretary decided to say on television.

The basic situation was a transactional detente. The US agreed to let Nvidia sell its nerfed chips in China, and Beijing agreed to resume the export of some critical rare earth minerals. Then, US Commerce Secretary Howard Lutnick went on CNBC and bragged about the government's thinking behind the deal. Here is the Financial Times:

“We don’t sell them our best stuff, not our second-best stuff, not even our third-best,” Lutnick told CNBC on July 15, the day after the Trump administration lifted export controls, implemented in April, on H20 sales.

“You want to sell the Chinese enough that their developers get addicted to the American technology stack, that’s the thinking,” he added.

Some of China’s senior leaders found the comments “insulting”, leading the policymakers to seek ways to restrict Chinese tech groups from buying the processors, according to two people with knowledge of the latest regulatory decision-making.

The problem, of course, is that this kind of cynical, dependency-based strategy operates on an unspoken understanding. Chinese tech giants and officials understood perfectly well that the H20 was not Nvidia's top-tier product; that was the implicit price of market access. But there is a world of difference between quietly accepting a less-powerful chip and having the US Commerce Secretary very publicly boast about selling your country "not our second-best stuff, not even our third-best." The act transforms a quiet, geoeconomic reality into a public shaming, forcing Beijing to react not to the chip's specifications, but to the insult.

And the US government had a direct financial stake in this now-sabotaged dependency, having negotiated a 15% cut of all of Nvidia's China revenue as its price for granting the export license. The whole point was to keep Chinese developers hooked on Nvidia's CUDA software platform, thereby slowing the development of a fully independent Chinese AI ecosystem.

The result is that Beijing is now actively discouraging its companies from buying the very chips the US just approved, pushing them instead toward Huawei's offerings. The irony, of course, is that the American plan was designed to make it difficult for Chinese tech giants to choose homegrown chips. Now, their own government is making it difficult for them to choose anything else.

Your Phone Company Sells You a Car

The basic playbook for a price war is pretty simple: the biggest company with the biggest factories and the lowest costs cuts prices until all the smaller companies start bleeding money and quietly disappear. You’d think this would be probably true for making cars, especially in China, where competition is brutal and businesses are famous for economies of scale.

Except, in this case, that isn't quite what's happening. The electric vehicle giant, BYD, has been cutting prices for two years, and yet a handful of smaller, nimbler competitors are doing surprisingly well. Stellantis-backed Leapmotor just reported its first-ever profit. Smartphone-maker Xiaomi, which just started selling cars last year, thinks its EV unit could be profitable by the end of this year. Their stocks have wildly outperformed BYD’s since the war began.

So what is going on? One theory is that these smaller players are winning not by playing the old industrial game, but by playing a new tech game. They are approaching this not like car companies, but like software companies that happen to make hardware. One way to see this is in the R&D budget. Reuters columnist Katrina Hamlin points out:

They have proportionately higher research and development budgets than larger rivals, with Xpeng's hitting 12 percent of revenue and Leapmotor's almost 8 percent in the first half of 2025. By contrast, BYD is forecast to spend 6.6 percent of its top line on R&D this year, per LSEG data.

This R&D spending goes into things like proprietary self-driving systems, which are a big deal for consumers. And it’s not just for their own cars. In a rather interesting turn of events, both Leapmotor and Xpeng are now licensing their advanced EV platforms and software to Western giants like Stellantis and Volkswagen. For decades, the deal was that Western car companies would teach their Chinese partners how to build cars. Now, the Chinese upstarts are selling the brains for new EVs back to the old masters.

The other part of the story is branding, or what you might call the ecosystem. Xiaomi was already a household name; it spent a decade getting its phones and electronics into millions of homes. Now it can just sell those people a car, too. Another carmaker, Seres, did something similar by partnering with Huawei. You can go see its cars in the same stores where you buy your phone. If it works, this might be an advantage? Your customer acquisition cost is much lower because you already acquired the customer when you sold them a smartphone five years ago.

So the old playbook was to win by having the biggest, most efficient factory. The new playbook, maybe, is to also win by having the stickiest smartphone and the smartest software. The game isn't just about building a better car. It’s about building a better computer that you can sit in.

The Private AI Factory

Ordinarily, if you are a giant, publicly traded company awash in cash, and you want to build a factory, you use your own cash. Or you issue some very cheap, investment-grade public bonds. What you probably don't do is go to the expensive, opaque private credit market and ask for a loan, and then watch as the biggest names on Wall Street engage in a months-long battle to be the one to give it to you.

And yet, this is exactly what Meta is doing. This month, the company just orchestrated a $29 billion private financing for a new AI data center, with Pimco and Blue Owl beating out a team of Apollo and KKR. The whole affair is a fascinating window into the strange new economics of the AI revolution.

The answer, it seems, has to do with the strange logic of the AI arms race. First, there is secrecy. By financing the project privately, Meta avoids broadcasting sensitive details about its AI infrastructure strategy—its capacity, its technology, its cost structure—to competitors.

Second, there is financial engineering. The deal is structured so the debt is tied directly to the data center as a physical asset, keeping the massive liability off Meta's main corporate balance sheet. It is a classic real estate developer's move. Here is Bloomberg on the off-balance-sheet mechanics:

Rather than paying for the center with Meta’s own cash, which would be more expensive, or with straight corporate debt, which would add leverage to the corporate parent, Morgan Stanley proposed a special purpose vehicle that would be tied to the assets themselves.

For the financiers, the deal is even simpler. It is an opportunity to lend money at a premium—reportedly 1.5 percentage points above Meta's public debt—secured by the most valuable collateral on Earth. If you are going to make a secured loan, a loan backed by the data center's core assets—which in the age of AI means a pile of Nvidia GPUs—is a pretty good one.

This is not just one deal; it is a signpost for a new era. A JPMorgan analysis estimates that AI data centers will require some $150 billion in financing in just the next two years, creating a massive new asset class for Wall Street to fight over. The construction of these digital factories will be fueled not just by the profits of tech giants, but by the deep, discreet, and very expensive pools of private capital that are now reshaping finance.

The Scoreboard

• Self-driving: Nvidia is latest investor to back AV startup Nuro in $203M funding round (TechCrunch)
• Consumer: Google Is Beating Apple on Smartphone AI (WSJ)
• Semiconductor: Chinese cities target 70% AI chip self-sufficiency to counter Nvidia (Nikkei Asia)