Monday, November 23, 2026
AI Infrastructure · News & Analysis
Home › Chips & Hardware › Report
Chips & Hardware · Report

Export Control Progress: 10 Chinese Companies Allocated 75,000 H200 Units Each as 50% Total Volume Cap Takes Effect

Trade pressSlicast · October 4, 2026 at 21:08 UTC · Global · Source: Spheron
importance 79

Nvidia's China AI chip market share collapsed from roughly 95% to zero in about two years, according to Jensen Huang himself. That's not a supply problem—it's a policy outcome, crystallized while the US opened a narrow legal path for H200 and MI325X sales to China, only to have Congress attempt to slam it shut a week later. For anyone renting GPUs from a provider with cross-border chip exposure, this is the story that matters in 2026—separate from the HBM and CoWoS supply shortages we've covered before. That covers why chips are scarce. This explains who's legally allowed to sell them where, and what it's doing to what you pay.

**Three Policy Shifts in Six Days**

The January 15 BIS Rule shifted export license review for H200 and MI325X chips from presumption of denial to case-by-case review. Presumption of denial means BIS assumes no unless proven otherwise. Case-by-case means BIS will actually examine the application. Chips must clear two thresholds: Total Processing Performance under 21,000 and DRAM bandwidth under 6,500 GB/s. Above that, older, tighter restrictions apply.

Qualifying chips require exporters to prove four conditions: the sale won't reduce US customer supply availability; the buyer has adopted KYC and export-compliance screening; an independent US-based third party has tested the chips; and production doesn't divert foundry capacity from US orders. Reexport or transfer licenses—chips made in Taiwan headed to China—remain under presumption of denial with mandatory US screening.

On January 14, President Trump signed a proclamation imposing 25% duty on advanced computing chips at the same performance thresholds, with the US government collecting 25% of resulting China-bound revenue. This stacks on top of an August 2025 arrangement where Nvidia and AMD agreed to hand the US government 15% of China AI chip sales revenue in exchange for export licenses. Two separate government cuts sit inside the price before shipment.

The volume cap adds a hard ceiling: China-bound shipments of qualifying chips can't exceed 50% of comparable US volumes. Chinese buyers reportedly had over 2 million H200s on order; the cap meant roughly 1 million units were actually eligible to ship. Nvidia readied approximately 82,000 AI GPUs for China with the 25% tax applied as shipments began.

The US Commerce Department cleared roughly 10 Chinese firms—including Alibaba, ByteDance, Tencent, and JD.com—to buy H200s, each capped at 75,000 units. As of mid-2026, not one had actually shipped. Beijing directed domestic tech companies to pause H200 orders, and Chinese customs separately told import agents the chips weren't permitted into the country at all. Washington opened the door; Beijing kept it shut for eight months.

**Congressional Response**

Six days after the BIS rule, on January 21, the House Foreign Affairs Committee voted 42-2 to advance the AI OVERWATCH Act (H.R. 6875), which would statutorily ban Blackwell-class chip sales to China for at least two years and give Congress veto power over future AI chip export licenses. That directly challenges the executive branch's loosened posture, specifically targeting the generation after H200.

The Remote Access Security Act (H.R. 2683) passed the full House 369-22 on January 12 and extends export-control logic past physical shipments to cloud-based remote GPU access. It closes a specific loophole: a restricted buyer couldn't legally import a controlled chip, but nothing stopped them from renting compute time on that same chip through a cloud provider outside their borders. If the Senate passes its companion bill, cloud providers—not just hardware exporters—will need customer vetting and licensing procedures.

**China's Pivot to Self-Sufficiency**

China's response to loosened US posture wasn't to buy. It was to build faster. Huawei's Ascend 950PR entered mass production in March 2026 and has captured the majority of China's AI chip orders for the year, with an upgraded Ascend 950DT planned for Q4 2026. Huawei brought memory in-house: its HiBL 1.0 stack removes dependency on SK Hynix or Samsung HBM, directly targeting the exact chokepoint export controls aim at. Huawei's upgraded CANN Next software stack implements a CUDA-like programming model—thread blocks, warps, kernel launches—that meaningfully cut porting costs for developers in Nvidia's ecosystem. ByteDance and Alibaba both placed orders after customer testing.

Jensen Huang stated plainly: "The goals of the export controls are not being achieved" and US export policy "has already largely backfired" by accelerating China's push toward chip self-sufficiency instead of slowing it. Nvidia took a $4.5 billion charge in fiscal Q1 2026 tied to H20 excess inventory and purchase obligations it could no longer fulfill under existing export restrictions. Nvidia's outlook still excludes Chinese data center compute revenue entirely, with CFO Colette Kress citing continued uncertainty over whether H200 imports will clear Chinese customs even after export licenses were approved.

**Supply Reallocation**

Every chip not shipping to a Chinese buyer because of stalled licenses, paused orders, or volume caps gets sold elsewhere. Combine the ~1 million-unit H200 cap, paused Chinese orders, and the proposed Blackwell ban, and meaningful supply reallocates toward US, EU, and Gulf state buyers already competing for limited Hopper and Blackwell inventory. That reallocation lands on an already tight market where circular financing arrangements between Nvidia and vendor-backed neoclouds commit much new capacity before it reaches the open spot market.

Read the original
Export Control Progress: 10 Chinese Companies… · Slicast