In the shadow of a White House summit between two superpowers, Nvidia has quietly written China out of its most ambitious revenue forecast yet — a $108 billion quarter built entirely on the strength of every market except the world's second-largest economy. What was once a fifth to a quarter of the company's data center revenue has been reduced, through eighteen months of tightening export controls and geopolitical friction, to a deliberate zero. The meeting between Trump and Xi on Thursday does not need to produce anything for Nvidia's business case to stand, but it carries the rare quality o
Nvidia's $108B forecast zeros out China; Trump-Xi summit could change calculus
The forecast stands on its own. Whatever happens is upside.
Mark
So Nvidia is saying it will make $108 billion next quarter, but none of that comes from China. How is that even possible when China used to be such a big part of their business?
Mimi
It's possible because the rest of the world is buying AI chips at a pace that's almost hard to fathom. Data center revenue alone is growing 117% year over year. The U.S., Europe, and other markets are absorbing everything Nvidia can make. China was maybe 20 to 25% of data center revenue before the export controls kicked in. That's a lot, but it's not everything.
Luke
But let's be precise about what "zero" means here. Nvidia did bill $7.9 billion to China last quarter. It's just that almost none of it was AI compute chips. It was gaming chips, workstation chips—products the export rules don't restrict. So the zero is specifically about data center compute, not China as a whole.
Mark
Right. And the export controls—those came from the U.S. government, not just market forces?
Mimi
Both, actually. The U.S. imposed a license requirement in April 2025 for the H20, a chip Nvidia designed specifically for China. That cost them $4.5 billion in charges. But Beijing also limited what Chinese companies could buy. Nvidia's own market share in China went from 95% to essentially zero. Then in February this year, the U.S. started issuing licenses for small quantities of H200 chips to go to China. ByteDance and Tencent got about 10,000 units each over the summer.
Luke
But here's the thing—Nvidia says it's only shipped a fraction of what those licenses actually permit. And they took another $400 million charge on excess H200 inventory. So even with permission, the market isn't there yet.
Mark
And now there's this summit on Thursday with Trump and Xi. Could that change things?
Mimi
It could. Chip export rules are being watched closely by investors. If the two governments agreed to ease restrictions, Nvidia could potentially sell more. But the summit's focus is on AI cooperation and trade broadly. Easing chip restrictions is seen as the harder ask. And after preparatory meetings, the U.S. Trade Representative said advanced AI chip export controls weren't even on the agenda for the AI safety talks.
Luke
Even if something did give, the upside is probably modest. If Nvidia returned to selling H20 chips at pre-restriction volumes, that's maybe $7 billion in additional quarterly revenue—about 7% on top of the current forecast. The company once pulled almost $20 billion a quarter from China. That's not coming back anytime soon because Chinese competitors have filled the gap and Beijing has pushed its own companies toward domestic chips.
Mark
So the forecast of $108 billion doesn't need China to work?
Mimi
Exactly. The forecast stands on its own. Whatever happens at the summit is upside, not the foundation.
Luke
And the stock is priced at about 14 times fiscal 2028 earnings. That valuation doesn't seem to assume a reopened China market. So even if the summit produces a breakthrough, it could surprise the market positively.
The Pulse
- Nvidia is forecasting its largest quarter ever — $108 billion, up 89% year-over-year — while assuming it will sell zero data center chips to China, a market that once generated nearly a quarter of that revenue.
- Export controls have collapsed Nvidia's share of China's AI chip market from roughly 95% to near-nothing, with each chip that does ship requiring U.S. inspection, a 25% tariff the company cannot pass on, and individual government licenses.
- A partial thaw emerged in 2026 — ByteDance and Tencent each received around 10,000 H200 units — but Nvidia has shipped only a fraction of what its licenses permit, and still took a $400 million charge on unsold inventory.
- President Trump hosts President Xi at the White House on Thursday, with AI on the agenda and Jensen Huang expected at the state dinner — raising the possibility, however uncertain, of a policy shift that could add roughly 7% upside to current forecasts.
- Even a full reopening would not restore the old order: Chinese competitors have spent years filling the gap, Beijing has pushed domestic chip development, and customers there have been steered away from foreign vendors.
In the shadow of a White House summit between two superpowers, Nvidia has quietly written China out of its most ambitious revenue forecast yet — a $108 billion quarter built entirely on the strength of every market except the world's second-largest economy. What was once a fifth to a quarter of the company's data center revenue has been reduced, through eighteen months of tightening export controls and geopolitical friction, to a deliberate zero. The meeting between Trump and Xi on Thursday does not need to produce anything for Nvidia's business case to stand, but it carries the rare quality of a moment when policy could move faster than any market model anticipates.
Nvidia's latest quarterly forecast contains an unusual admission buried in its assumptions: the company expects to sell nothing — not a single dollar of data center compute chips — to China. The $108 billion projection, an 89% jump from the prior year, is built entirely on markets outside the world's second-largest economy, a country that once supplied between a fifth and a quarter of Nvidia's data center revenue.
The path to zero was neither sudden nor simple. Over eighteen months, U.S. export controls were tightened, briefly loosened, and tightened again. In April 2025, Washington required a license just to ship the H20 — a chip Nvidia had designed specifically for the Chinese market — costing the company $4.5 billion in charges. By year's end, CEO Jensen Huang described Nvidia's China AI chip market share as having fallen from roughly 95% to zero. The Chinese government compounded the damage by discouraging its own tech companies from purchasing foreign chips at all.
Some movement returned in early 2026. The U.S. began issuing limited licenses, and by summer, ByteDance and Tencent had each received around 10,000 H200 units — the first meaningful deliveries in years. But Nvidia has shipped only a fraction of what those licenses allow. Every chip bound for China requires U.S. inspection and carries a 25% tariff the company cannot pass to customers. Nvidia has said plainly that under current rules, it cannot offer a competitive data center product for broad distribution in China with both governments' blessing.
On Thursday, September 24, President Trump will host President Xi at the White House, with artificial intelligence among the central topics. Jensen Huang is expected at the state dinner. The arithmetic of a potential breakthrough is clear enough: a return to pre-restriction H20 volumes would add roughly $7 billion in quarterly revenue — about 7% on top of the current forecast. But the old days are unlikely to return regardless of what is decided. Chinese competitors have had years to fill the void, domestic chip investment has accelerated, and customer habits have shifted under government pressure.
Nvidia's stock, trading near $222, does not appear to price in a reopened China. A policy shift Thursday could make the $108 billion forecast look conservative almost immediately. But the forecast stands on its own without one — and that, perhaps, is the more remarkable fact.
Nvidia's latest revenue forecast carries an unusual zero in its fine print: the company is assuming it will sell nothing—not a single dollar—of data center compute chips to China in the coming quarter. The $108 billion projection, released in late August, represents an 89% jump from the prior year, but it gets there entirely without the world's second-largest economy, a market that once supplied between a fifth and a quarter of the company's data center revenue.
This is not an accident. It is a choice, and a necessary one. The U.S. government has effectively locked Nvidia out of China's AI chip market through export controls that have been tightened, loosened, and tightened again over the past eighteen months. In April 2025, Washington told Nvidia it needed a license just to ship the H20, a chip designed specifically for the Chinese market. That decision cost the company $4.5 billion in charges. By October of that year, CEO Jensen Huang was describing Nvidia's share of China's AI chip market as having collapsed from roughly 95% to zero. The Chinese government, meanwhile, has discouraged its own tech companies from buying foreign chips, pushing them toward domestic alternatives instead.
The situation has thawed slightly, but only in the margins. Starting in February 2026, the U.S. began issuing licenses allowing Nvidia to send small quantities of H200 chips to specific customers in China. The company took another $400 million charge this year on excess H200 inventory that never found buyers. By summer, ByteDance and Tencent each received roughly 10,000 H200 units—the first significant deliveries in years. But Nvidia has shipped only a fraction of what its licenses actually permit. Each chip bound for China clears U.S. inspection and carries a 25% tariff the company cannot pass to customers. The company itself has stated that under current rules, it cannot deliver a competitive data center product for broad distribution in China with both governments' approval.
On Thursday, September 24, President Donald Trump will host Chinese President Xi Jinping at the White House. Artificial intelligence is among the main topics on the agenda. Nvidia CEO Jensen Huang is expected to attend the state dinner that evening. The question hanging over the meeting is whether anything might shift the calculus on chip exports—and if so, what it would mean for a forecast built on the assumption that nothing will.
The math is instructive. If Nvidia could somehow return to selling H20 chips at the volumes it managed before the license requirement took effect, the company would add roughly $7 billion in quarterly revenue—about 7% on top of the current $108 billion forecast. That sounds meaningful until you consider what came before. Nvidia sold $4.6 billion of H20 chips in the first quarter of fiscal 2026, before restrictions tightened, and could not ship another $2.5 billion due to the new rules. Even combined, that is a fraction of what the company once pulled from China. The previous share of data center revenue would be worth almost $20 billion a quarter on today's base—a number no serious analyst should pencil in. Chinese competitors have had years to fill the gap. Beijing has invested heavily in domestic chip development. Customers there have been discouraged by their own government from relying on foreign vendors. A reopening of the market would not mean a return to the old days.
Yet the summit matters because policy can move faster than markets expect. The stock, trading near $222, is priced at roughly 14 times fiscal 2028 earnings estimates. Whatever is baked into that valuation, it does not appear to assume a reopened China. A breakthrough on Thursday could make management's forecast look conservative in a hurry. But nothing has to happen at the summit for the investment case to hold. Nvidia's $108 billion forecast, built on the strength of markets outside China, stands on its own. Whatever the two governments decide this week is simply a bonus.
Notable Quotes
Nvidia cannot deliver a competitive data center product for broad distribution in China with both governments' approval under current rules— Nvidia company filings
Nvidia's share of China's AI chip market dropped from about 95% to zero— Jensen Huang, Nvidia CEO, October 2025
Want the full story?
Read the original at The Motley Fool ↗
Politics & Governance