U.S.-China Relations
NextFin News - Chinese AI developers who have ordered roughly 6 million Nvidia RTX Pro 5500 chips - a stash worth about $70 billion at list prices - expect to receive only some tens of thousands of units, or roughly 1%, over the next four months, as the US chip giant struggles to lift production fast enough to meet demand that has outstripped its ability to build the cards.
The order book, held by Alibaba Group, Tencent Holdings, ByteDance and other Chinese AI and cloud firms, is one of the largest chip orders in Nvidia's history on paper. But the fulfillment math tells a different story: a gap between headline demand and deliverable revenue that exposes how tightly US export controls have warped the China channel, and how little of the apparent opportunity can actually land in the near term. The $70 billion figure is a measure of Chinese appetite, not a forecast of what Nvidia will collect.
The Order That Cannot Be Filled
People familiar with the matter say Chinese buyers have in total ordered about 6 million RTX Pro 5500 chips, yet the companies predict receiving only roughly 1% - some tens of thousands of units - within the next four months. Nvidia has told customers and partners it cannot promise to deliver anywhere close to the requested volume because demand exceeds planned production.
The arithmetic behind the headline is straightforward. The chip has been reported at roughly $11,400 per unit in China-facing pricing, and 6 million units at that price multiply out to approximately $70 billion in sticker value. But a sticker price is not a signed contract, a wired payment, or a confirmed shipment. It is a list-value calculation based on requested quantities, and with Nvidia itself warning it cannot build enough cards, the $70 billion figure functions more as a measure of pent-up demand than as revenue likely to be recognized in fiscal 2027.
For context, the company's current-quarter revenue forecast of $108 billion assumes no China data-center chip revenue at all. Any sales that do clear the export-control hurdle would come on top of that outlook - but the 1% allocation means the incremental contribution in the next four months is measured in the hundreds of millions of dollars at most, not billions. Tens of thousands of units at roughly $11,400 each is a rounding item against a company that generates more than $50 billion a quarter.
"I think that we could only be in service of a market if the country wants us to be," Nvidia chief executive Jensen Huang said in September 2025, when Beijing told firms to stop buying the company's RTX Pro 6000D accelerator.
That line captures the two-sided gate that now defines Nvidia's China business: Washington decides what Nvidia is allowed to sell, and Beijing decides whether Chinese firms are allowed to buy it. The RTX Pro 5500 sits in the narrow corridor where both gates are currently open - and even there, the company cannot build enough to pass through.
Why the RTX Pro 5500 Exists - and Why It Can Vanish
The RTX Pro 5500 is not Nvidia's flagship chip. It is a deliberately scaled-back product engineered to clear the technical performance thresholds that US export rules impose on sales to China. With 84 GB of GDDR7 memory with error-correcting code, 21,760 CUDA cores, and a power draw of up to 600 watts, it is positioned as a professional workstation GPU for AI agents, AI inference, and simulation - not as a conventional data-center accelerator like the barred B300 or the restricted Hopper family.
That positioning is deliberate, and it is fragile. US export controls do not ban chips by name alone; they draw a line through technical parameters - total processing performance, performance density, and interconnect bandwidth - and anything above the line requires a license that is presumed denied. Nvidia's compliance strategy for years has been to design products that sit just under that line: the A800 and H800 followed the A100 and H100 bans; the H20 emerged after another tightening; the RTX 4090's China variant followed the consumer-card restrictions. The RTX Pro 5500 is the latest entry in that sequence - a chip built around a rulebook rather than around raw performance.
The strategy carries recurring policy risk, and the company has already paid for it. In April 2025, Washington required an export license for the H20 - the very chip Nvidia had designed to comply with earlier curbs - and Nvidia recorded about $5.5 billion in charges related to H20 inventory and purchase commitments. A compliant workaround today can become a prohibited good tomorrow, which is why Chinese buyers have been racing to lock in supply while the window remains open. The RTX Pro 6000D episode in September 2025, when Beijing told firms to stop buying the card and Nvidia's shares fell 2.6% that day, shows the other side of the same trap: even a chip that clears Washington's rules can be shut out by Beijing's procurement guidance.
Recent diplomatic activity has not resolved the underlying tension. Following the Trump-Xi meeting in Washington, US Trade Representative Jamieson Greer said national security export controls were taken "off the table" as a negotiating point - which is not the same as lifting them. The controls remain in force, the licensing machinery keeps running, and the RTX Pro 5500 remains a mid-tier workstation part rather than the data-center-class silicon Chinese labs actually want.
The Real Story Is Not the $70 Billion
The most consequential detail in the order news is not the size of the demand signal. It is the supply-chain admission embedded inside it: Nvidia, the world's most valuable chip company, is telling its largest customers that it physically cannot build what they have asked for. That admission reframes the headline from "China wants $70 billion of Nvidia chips" to "China wants far more Nvidia silicon than the global supply chain can deliver under current rules."
Reports have placed Nvidia's 2027 global production capacity for the RTX Pro 5500 at no more than 3 million units - half of total Chinese demand for this single chip alone, according to the reported figures. That mismatch is the structural reality underneath the cyclical allocation drama. Even if every policy gate stayed open, the foundry capacity, advanced packaging, and memory supply needed to serve China's appetite would have to come out of a global pool already stretched by data-center demand for Blackwell accelerators. The RTX Pro 5500 is built on TSMC's custom 4N process, the same family of nodes feeding Nvidia's highest-margin products; every wafer allocated to a China-compliant card is a wafer not allocated to an unrestricted data-center part.
The second-order question most investors are not asking: what happens to the 99% of demand that does not get filled? Every unfilled RTX Pro 5500 order is a live lead for Huawei, Cambricon, and Enflame - the domestic suppliers that already hold nearly 80% of China's AI server market. Nvidia's production shortfall is not a neutral event; it is a forced migration of Chinese buyers toward local silicon, and each month of under-delivery deepens the switching costs, software integration, and procurement relationships that make those buyers harder to win back later. A customer that rewrites its inference stack for Huawei's Ascend architecture to survive a shortage does not revert to Nvidia the moment a few tens of thousands of cards arrive.
There is also a pricing dynamic worth noting. The RTX Pro 5500 has been reported at roughly RMB 85,000 to 90,000 in China, comparable to Huawei's Ascend 950PR - meaning Nvidia is competing for this business on price as well as availability, not on uncontested superiority. Scarcity without pricing power is the worst combination for a company accustomed to both.
What the Market Is Pricing - and What It Is Not
Nvidia shares closed at $206.69, up $4.63, or 2.29%, on the session, with a market capitalization of about $4.98 trillion. The stock trades below its 50-day average of roughly $221.50 and well under its 52-week high near $243, reflecting a market that has spent recent weeks weighing China policy headlines against the durability of the broader AI capital-expenditure cycle.
The muted-to-positive reaction to the order news is consistent with how this stock has traded the China channel: the upside is capped because Nvidia itself is the one warning it cannot fulfill the orders, while the downside is cushioned because the company's base forecast already assumes zero China data-center revenue. In other words, the bad news is priced in, and the good news is constrained by physics and policy rather than by demand. A 1% fill rate moves neither the revenue model nor the earnings multiple.
Consensus data shows all 31 tracked analysts rate the stock a buy, with an average price target around $324 - a target built on the global data-center buildout, not on China reopening. The next hard test arrives with the company's third-quarter fiscal 2027 earnings on November 17, when management will face questions on how much of the RTX Pro 5500 pipeline can convert into actual shipments and revenue, and whether the line earns a place in the growth narrative or remains a footnote.
The Counter-Thesis: This Is Not a Breakthrough - It Is a Pressure Valve
The bull case reads the order book as evidence that China still wants Nvidia silicon, that the recent summit helped thaw the channel, and that the RTX Pro 5500 is the thin edge of a reopening that could restore billions in revenue. Under that reading, the 1% allocation is a starting point that expands as production ramps and licensing clarity improves.
The stronger counter-thesis is simpler and harder to dismiss. A 1% fill rate over four months is not a reopening; it is a pressure valve. The order volume reflects panic-buying ahead of the next policy shift, not a normalized commercial relationship. Beijing has spent years pursuing semiconductor self-sufficiency as a strategic objective, and a government that signals approval for one mid-tier workstation card while its domestic champions capture 80% of the AI server market is managing dependency, not ending it. The structural direction of travel - toward domestic silicon - is unchanged; the RTX Pro 5500 is a tactical exception, not a regime change.
This is where the cyclical-versus-structural call matters, because getting it wrong flips the conclusion. The production bottleneck itself is cyclical: capacity can be added, yields can improve, and a 1% allocation can become 10% if Nvidia dedicates more 4N wafer starts to the part. But the constraint that shapes the opportunity is structural and will not self-correct: US export controls are a durable feature of the relationship, not a temporary disruption, and China's domestic-substitution drive is a state-backed multi-year program, not a stopgap. The cyclical leg can improve; the structural leg keeps pulling the other way.
The falsifying signal is specific. If China grants broad, sustained approvals for higher-performance data-center chips - the H200, the B300-class parts, or any accelerator above the current performance threshold - and Nvidia begins recognizing meaningful China revenue above its zero-assumption forecast, then the "pressure valve" read is wrong and a genuine reopening is underway. Absent that, the 1% allocation is the regime.
What to Watch
Short term (next four months): actual RTX Pro 5500 shipment volumes and the realized fill rate against the 6 million-unit order book. Tens of thousands of units delivered confirms the constraint; anything approaching even 10% would signal either a production ramp or a policy shift. Watch Nvidia's own language on whether it "cannot guarantee" fulfillment hardens into formal allocation notices.
Medium term (through fiscal 2027): the November 17 earnings call and any update to the $108 billion quarterly forecast. Management commentary on China contribution, and whether the RTX Pro 5500 line is treated as a growth vector or a rounding item, will set the tone. Also watch for any charge or inventory disclosure that echoes the H20 episode.
Long term (structural): US Commerce Department licensing posture and Beijing's procurement guidance. If Washington tightens below the RTX Pro 5500's performance threshold - as it did with the H20 - the entire channel closes again and the order book converts from opportunity into write-off risk. If Beijing widens approvals to data-center-class parts, the domestic-substitution thesis weakens and the stock re-rates on reopening optionality.
The base case is continued tight allocation: enough shipments to keep Chinese customers engaged, not enough to move Nvidia's revenue needle. The upside case is a negotiated reopening that lifts the performance ceiling. The downside case is a repeat of the H20 episode - a compliant chip grandfathered into prohibition.
The $70 billion number is real as a measure of Chinese appetite. It is misleading as a measure of what Nvidia will collect. In the chip war, the most important figure is not the order book - it is the 1%.
Explore more exclusive insights at nextfin.ai.
