Don’t Blame Nvidia for this China Crisis

You’ll be relieved to learn that I do not plan to expound on the Cracker Barrel affair, variously described as an omnishambles or a marketing masterstroke. My only comment on that particular cause célèbre was posted on LinkedIn, where I remarked that any corporate rebranding should have retained the Cracker Barrel codger, but had him peer distractedly at a smartphone cradled in his left hand. That was probably too much modernity for the traditionalists, who ultimately won the day. What’s old is new again, or at least old again.

Let’s get to the gist of today’s post, which necessarily involves AI, currently sweeping everything in its path. AI, however, is merely our context. Specifically, we’re going to revisit the challenges Nvidia confronts as it is pulled away and pushed away from the Chines market. The pulling is being done by the U.S. government, and the pushing, in reaction to the pulling, is being done by the Chinese government. Nvidia is in the middle, which will not hold. It’s the geopolitical football, getting thrown, punted, and kicked around by two implacable foes.

Anyway, without further prologue, let’s get to it.

Power of Empathy

Empathy has many benefits. When we empathize, we gain perspective, an understanding of how others perceive our actions and how they experience and perceive the world. Empathy allows us to break through our own egotism and solipsism, conferring valuable insights that not only help us become better people, but also help us make better decisions, with fewer unintended consequences.

Empathy, however, is not sympathy. Empathy is the capacity to share and understand another person’s emotional experience. It can also be extended, with suitable discretion, to coherent and identifiable groups or collectives.

Conversely, sympathy is a feeling of concern or compassion for someone who is experiencing difficulties. Compassion, closely associated with sympathy, involves recognizing and wanting to alleviate the suffering experienced by another.

In short, empathy involves putting yourself in another person’s place, attempting to share their perspective, while sympathy and compassion spur some sort of action that offers succor or support to the afflicted. We can empathize without sympathizing; we can appreciate and come to understand another’s circumstances or position without necessarily endorsing or supporting it. You can empathize with anybody, even adversaries.

When Fourth-Best Isn’t Good Enough

I’ve taken you into this digression on semantics to make a point about Nvidia’s increasingly tenuous relationship with Chinese customers. To understand what’s happening, you have to empathize with China’s position. I’ve written on this subject before, most recently little more than a week ago. If you care to do so, you can consult last week’s post for more detail.

Suffice it to say, Nvidia didn’t contribute to its worsening relationship with China. Geopolitics is to blame for the schism. Regardless of blame, Nvidia has to deal with the consequences, principally decreasing demand from China for Nvidia’s chips.

You don’t to be a tasseographer to read Nvidia’s Chinese tea leaves or coffee grinds. If you empathize with China’s position — again, you don’t have to support the position to empathize with it — you will understand that the situation is untenable.

By the U.S. government’s own admission, China is being offered Nvidia’s technological castoffs. Nvidia would like to offer China better chips, more like the ones it sells to U.S. and other customers, but the U.S. government won’t allow it. Worse, for China and for Nvidia, the Trump Administration might peremptorily demand that Nvidia sell even worse technology to China, or perhaps no technology at all. It’s a slippery slope, and China knows it only ends at the bottom of the hill.

Yes, China firmly grasps its conundrum, and it has done so for a long time. Even if China had missed the clues, if China had been willfully obtuse to U.S. intentions, Howard Lutnick, the U.S. commerce secretary, helpfully clarified matters in an interview with CNBC:

"We don't sell them our best stuff, not our second best stuff, not even our third best. The fourth one down, we want to keep China using it... The idea is the Chinese are more than capable of building their own. You want to keep one step ahead of what they can build, so they keep buying our chips. You want to sell the Chinese enough that their developers get addicted to the American technology stack. That's the thinking."

Diversification Imperative

There’s no mystery here, folks. Everybody knows the game. It’s a recursive bait-and-switch, and it’s meant to keep China “addicted” and well behind the AI-processing performance curve.

No wonder, then, that China is making strenuous, sustained efforts to reduce dependence on U.S. technology, including Nvidia GPUs and AI accelerators. Eventually, Nvidia’s China revenue — usually at about 15% of the total in any given quarter, down to 5.9% in the latest quarter — will go to zero. That won’t happen overnight, of course, but the trend line is clear.

Nvidia gets a disproportionate share of its overall revenue from a small number of customers, and they’re not in China. The company is making a sustained effort to diversify its customer base — always a good idea. To the extent that Nvidia can diversify and expand its business elsewhere before Chinese demand evaporates, investors will remain sanguine. It’s a race of sorts, and I wouldn’t bet against the Nvidia team, even though the company faces more risks than the loss of business in China.

Many observers like to compare and contrast Nvidia with Cisco, the networking behemoth that grew into adulthood during the dot-com boom. Nevertheless, perhaps a better comparison, in certain key respects, is Arista Networks, which built and sustained its impressive datacenter-networking business on the loyal, long-term patronage of a few hyperscalers (cloud titans, as Arista calls them). Arista built its fortune and reputation in the cloud, and Nvidia is doing likewise, using the AI as the catapult. Some of the same challenges and risks that pertain to Arista’s cloud patronage also apply to Nvidia.

Anyway, there’s sound logic underlying China’s quest for technological self-sufficiency. You might not like it, but it’s hard to argue with the reasoning. Nobody wants to be in a situation where a supplier’s key technology can be arbitrarily degraded or cut off entirely.

Delusions and Fantasies

Wall Street analysts are indulging in fantasy when they predict that Nvidia will retain a meaningful long-term presence in the Chinese market. The die has been cast. Howard Lutnick has called out the last bingo numbers. The game is all but over.

Again, I want to reiterate that none of this is Nvidia’s fault. As Nvidia CEO Jensen Huang continually points out, what Nvidia can or cannot sell to China is entirely up to the U.S. government.

Still, Nvidia must live with the fallout. For Nvidia investors, that means watching how the company compensates for the gradual loss of Chinese trade through the cultivation of other customers and markets. If the Chinese business is going away — more a matter of when, not if — Nvidia must find substitutes or lose a source of revenue growth and profitability. I’m not saying Nvidia can’t find new sources of revenue — it’s a well-run company, as its past performance attests — but we have to consider the need for further diversification as Chinese demand recedes.

Not to beat this topic like a Victorian carpet, but what would do? You’d probably do something similar to what China is doing, seeking to accelerate self-sufficiency by cultivating a homegrown AI ecosystem to support domestic demand.

We learned today that DeepSeek is using Huawei AI processors to train some of its artificial intelligence models. Before plumping for Huawei’s Ascend processors, DeepSeek had also tested accelerators from Baidu and Cambricon.

Huawei says it has already built an ecosystem that is entirely independent of the United States, giving the company and its customers self-sufficiency. Perhaps that’s hyperbole — a universal marketing trait, after all — but the direction and intent are clear.

As for Cambricon, perhaps it is enjoying too much of a good thing. The AI-chip company’s stock surged so stratospherically that Cambricon, as reported in the Wall Street Journal, felt compelled to flash a yield sign. An excerpt from the WSJ article:

There is a risk that the stock price may have deviated from current fundamentals, as its current price-to-earnings ratio is over 5000 times, far exceeding the industry level, the company said.

That’s nosebleed P/E territory, no doubt. While Cambricon’s high-flying growth catches the eye (a 4000% revenue jump!), investor exuberance has been known to jump the turnstiles and sprint toward the first-class compartment on the gravy train. Still, what can we take from such a heady episode on the Chinese markets? One reasonable conclusion is that Cambricon is China’s answer (one of them, anyway) to Nvidia. There will be others.

China’s calculation goes as follows: AI chips from domestic suppliers might not be as good as Nvidia’s Blackwell, or whatever comes next from Nvidia. But China isn’t getting Blackwell. They’re getting the H20, a demonstrably worse offering. At any given time, the Trump Administration might decide to give China something worse — maybe nothing at all. Lutnick has admitted that the objective is to lead China down a garden path, preferably one that doesn’t go anywhere.

Send in the Clowns

I am going to close with a non-sequitur, if only because I can. It’s a Friday, after all, and I think we can all benefit from wry historical perspective.

We live in strange times, but perhaps people have always lived in strange times. More to the point, people have probably always made times strange. At the Toronto Star, we read a historical account of the improbable “clown riots” of 1855. It’s an outlandish tale, yet apparently true.

Let’s give the clowns credit; they’re come a long way. They might have resorted to rioting in 1855, but some of them own and operate a globe-girding traveling circus in 2025. In these times, when performance art has been eclipsed by performative living, the big top is bigger than ever.