Both the United States and China independently blocked NVIDIA chip trade through opposite mechanisms — export permissions that produce zero shipments, procurement bans that mandate domestic alternatives. NVIDIA posted record earnings without a dollar of Chinese data center revenue, making the bifurcation official.
NVIDIA posted eighty-one point six billion dollars in quarterly revenue — the largest in semiconductor history — without a dollar of data center compute revenue from China. Not because Washington blocked the sale. Because both governments did.
The United States cleared H200 exports to roughly ten Chinese firms in May 2026, including Alibaba, Tencent, and ByteDance. Each buyer was approved for up to seventy-five thousand chips. On paper, it reads like a thaw. In practice, zero deliveries have been completed. Beijing customs has blocked chip imports since January. The Commerce Department issues permissions it knows will not clear the other side.
China's veto operates from the opposite direction. Beijing banned foreign AI chips from all state-funded data centers — a retroactive mandate affecting over one hundred billion dollars in public investment since 2021. Projects under thirty percent construction must remove any installed NVIDIA, AMD, or Intel accelerators. Advanced builds face case-by-case review. The guidance was never formally published but is enforced through procurement rules that route every state-funded chip purchase to Huawei's Ascend line.
The Corporate Acknowledgment
NVIDIA's Q1 FY2027 guidance contained a sentence that corporate earnings reports almost never include: the company is not assuming any data center compute revenue from China in its outlook. This is the bilateral blockade written into a financial forecast — not a geopolitical white paper, but a binding forward estimate that Wall Street will price positions against.
The earnings beat was massive. Revenue of eighty-one point six billion exceeded the seventy-nine point two billion consensus by three percent. Earnings per share of one dollar eighty-seven cents beat one dollar seventy-seven. Q2 guidance of approximately ninety-one billion — midpoint of an eighty-nine to ninety-three billion range — exceeded consensus of eighty-six point eight billion by over four billion. The company announced an eighty-billion-dollar buyback and raised its quarterly dividend twenty-five-fold, from one cent to twenty-five cents. Free cash flow reached forty-eight point six billion.
The stock slipped in after-hours trading. A company that beat every estimate and posted the most profitable quarter in chip history still declined, because the beat was expected and the blockade was already priced.
The Structure of the Blockade
What makes this blockade bilateral — and therefore durable — is that each side has independent reasons to maintain it.
Washington's motivation is strategic denial. Export controls on advanced AI chips have tightened every year since 2022. The H200 clearance looks permissive until you examine the implementation: seventy-five thousand chips per firm sounds generous, but not a single unit has shipped. The approvals function as diplomatic gestures that let the administration claim trade engagement while the customs apparatus prevents material flow.
Beijing's motivation is industrial policy. The state procurement ban is not primarily about retaliation. It is about creating a captive market for domestic semiconductors. Huawei's Ascend chips become the mandatory architecture for every state-funded AI project. Over one hundred billion in government spending now flows exclusively to Chinese silicon. The ban makes Huawei's competitive position a policy outcome rather than a market outcome.
The result is a blockade that neither side needs to enforce through confrontation. Each veto is self-reinforcing. Washington grants permissions that Beijing blocks. Beijing mandates alternatives that make the permissions irrelevant. The diplomatic posture is engagement. The practical outcome is bifurcation.
What Gets Built During the Blockade
NVIDIA's quarter answers the financial question: the company does not need China. Data center revenue of seventy-five point two billion came entirely from non-Chinese customers. The AI infrastructure buildout across the United States, Europe, Japan, and the Middle East generates enough demand to post record earnings with the world's second-largest economy completely excluded.
The harder question is what happens to AI development trajectories. Two compute ecosystems are now being constructed in parallel — one on NVIDIA's architecture, one on Huawei's Ascend. The performance gap between them determines whether the bifurcation is a strategic nuisance or a structural disadvantage. If Ascend reaches competitive performance within two years, the blockade accelerated Chinese semiconductor independence at a modest capability cost. If the gap persists or widens, China's state-funded AI projects operate under a permanent compute ceiling that no amount of software optimization can fully close.
Neither side is investing in reopening trade. Both are investing in making it unnecessary.
What Would Break It
Two conditions would falsify the bilateral blockade thesis. H200 deliveries to China actually begin within ninety days — meaning both Washington lifts its customs hold and Beijing allows imports. Or China reverses its state procurement ban, signaling that domestic alternatives cannot meet performance requirements.
Neither requires just one government to act. Both require mutual coordination that serves neither side's current domestic political incentives. Washington needs export controls to demonstrate strategic seriousness. Beijing needs the procurement ban to justify Huawei's subsidized expansion.
The bilateral blockade is not a crisis. It is an equilibrium — stable precisely because both sides prefer it to the alternatives.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)