Export controls impact Chinese AI compute

What AI 2027 Predicted

The scenario describes a dynamic where US export controls genuinely constrain Chinese access to frontier compute (especially NVIDIA GPUs), but China adapts faster than expected through domestic chip alternatives and creative workarounds.

How We Track This

We monitor:

• US export control policy updates and enforcement actions
• Chinese domestic chip development (Huawei Ascend, etc.)
• Training runs confirmed to use domestic Chinese hardware
• NVIDIA revenue reports and China-specific GPU production decisions

Current Evidence

China is adapting faster than expected. GLM-5 (744B parameters) was trained entirely on Huawei Ascend chips. DeepSeek V4 (1T parameters) is optimized for Ascend. NVIDIA halted China-bound H200 production in March 2026. But China’s 15th Five-Year Plan has an aggressive AI infrastructure strategy with military-civil fusion. Export controls are having impact but China is finding workarounds.

Reuters reported on April 29, 2026 that large Chinese tech firms were scrambling to secure Huawei AI chips after the DeepSeek V4 launch, and that Huawei planned around 750,000 Ascend 950PR units in 2026 with mass production beginning in April and full-scale shipments expected in the second half of the year. This reinforces both halves of the prediction: export controls continue to shape procurement, while China is scaling domestic alternatives.

CNA/Reuters reported on May 15, 2026 that US Trade Representative Jamieson Greer said semiconductor export controls were not a major topic in US-China talks, while also noting that no H200 deliveries had occurred despite earlier US clearances for some Chinese buyers. This is a mixed update: controls remain a live constraint, but the policy environment is fluid rather than a simple tightening path.

AP reported on May 21, 2026 that Taiwanese authorities were investigating three people suspected of using forged documents to smuggle servers containing advanced Nvidia chips to China. This adds another enforcement and circumvention datapoint: controls remain binding enough to create smuggling incentives, while access attempts continue through illicit or gray-market routes.

Late-May reporting said China certified nine domestic AI processors for government procurement under a new AI training/inference chip category in the Anke security-certification framework. This is incremental evidence for the adaptation side of the export-control dynamic: Chinese procurement is being steered toward domestic alternatives, though certification does not by itself show frontier parity with Nvidia.

On May 31, 2026, BIS clarified that exporters still need licenses for advanced computing items when the buyer is headquartered in, or ultimately parented from, Country Group D:5 or Macau, even if the purchasing entity is located elsewhere. This strengthens the enforcement side of the export-control picture by targeting offshore routes used to obtain AI accelerators.

Sources:

• How China’s GLM-5 Works: 744B Model on Huawei Chips
• DeepSeek V4: 1T Parameters, Huawei Ascend-Optimised
• China’s 5-Year Plan Has Moved Beyond the Chip War — The Diplomat
• Big Chinese tech firms scramble to secure Huawei AI chips — Reuters
• Chip export controls not major topic in China talks — CNA/Reuters
• Taiwan prosecutors investigate Nvidia-chip server smuggling — AP
• Tom’s Hardware: China certifies nine domestic AI chips for government procurement
• BIS guidance on advanced computing items for D:5 and Macau-headquartered entities

Counterevidence & Limitations

• China’s domestic alternatives remain less efficient than frontier NVIDIA hardware — Ascend chips reportedly require significantly more engineering effort and deliver lower throughput per watt
• The Ascend ecosystem is still maturing; software tooling (compilers, frameworks, debugging) lags CUDA significantly, adding friction to large-scale training
• It remains unclear whether domestic workarounds can scale to the largest training runs (10^26+ FLOPs), where hardware reliability and interconnect performance become critical
• Export controls are tightening, not loosening — the long-term cumulative impact may be larger than current snapshot assessments suggest
• The “adaptation” narrative may overweight high-profile examples (GLM-5, DeepSeek V4) while underrepresenting the broader ecosystem of Chinese AI companies that lack DeepSeek-level engineering talent

What Would Change Our Assessment

• Downgrade: If China demonstrated full compute parity with US labs despite controls
• Upgrade: Already confirmed; the dual dynamic (impact + adaptation) is clearly happening

Update History

Date	Update
2026-06-08	BIS issued guidance clarifying that advanced-computing license requirements apply to entities headquartered in, or ultimately parented from, Country Group D:5 or Macau even when the entity is located abroad. This reinforces the active constraint side of the export-control dynamic while leaving the China adaptation evidence unchanged.
2026-06-01	Added reporting that China certified nine domestic AI processors for government procurement. This reinforces the confirmed pattern of export controls pushing Chinese procurement toward domestic alternatives, while not changing the assessment of the remaining performance gap.
2026-05-25	AP reported that Taiwanese authorities are investigating three people suspected of using forged documents to smuggle servers containing advanced Nvidia chips to China. This adds another incremental enforcement/circumvention data point supporting the confirmed export-control dynamic.
2026-05-23	Added CNA/Reuters reporting that semiconductor export controls were not a major topic in May US-China talks and that no H200 deliveries had occurred despite earlier US clearances for some Chinese buyers. This modestly reinforces the active-constraint picture while preserving the caveat that policy is fluid.
2026-05-04	Reuters reported Chinese cloud and tech firms scrambling to secure Huawei AI chips after DeepSeek V4, with Huawei planning roughly 750,000 Ascend 950PR units in 2026 and mass production beginning in April. This strengthens evidence for both constraint and adaptation under export controls. Confidence adjusted 0.85 → 0.90.
2026-04-13	Mixed signals on export control effectiveness. CFR analysis (Dec 2025) argues controls ARE working — Huawei “falling further behind,” recommends maintaining restrictions (CFR). However, Trump admin approved H200 exports to China (Dec 2025), partially loosening controls. DeepSeek V4 on Huawei chips shows constrained self-sufficiency. Reports of Nvidia Blackwell chip smuggling for V4 training suggest controls partially circumvented. Alibaba opening datacenter with 10K own chips + China Telecom — domestic alternatives growing. Overall: controls still constraining but effectiveness gradually eroding. No status or confidence change.
2026-04-06	DeepSeek V4 confirmed to run exclusively on Huawei chips (Reuters, The Information). Alibaba, ByteDance, and Tencent ordered hundreds of thousands of Huawei chips in preparation. This represents the most significant evidence of China achieving domestic AI chip self-sufficiency — a direct response to US export controls as the scenario predicted. Meanwhile, the US trade war is creating pressure on America’s AI dominance, with Trump administration policies complicating the export control regime (WebAndITNews). The dual dynamic (export controls constrain + China adapts) continues to play out exactly as predicted. No status or confidence change.
2026-03-23	Supermicro co-founder and two others charged with smuggling $2.5B in Nvidia AI chip servers to China via Southeast Asia shell companies (CNN, NBC News). FDD analysis finds Chinese firms routinely circumvent controls by purchasing through US-based entities (FDD). Meanwhile ByteDance secured Nvidia chip access for global ops outside China (Breitbart). Confidence adjusted 0.80 → 0.85 — the dual dynamic of enforcement and circumvention is playing out more dramatically than expected.
2026-03	Clear evidence of both impact (Chinese labs face compute constraints) and adaptation (Huawei Ascend, smuggling networks, cloud workarounds). Dynamic playing out as predicted.
2025-12	Trump reversed H200 chip export controls to China, allowing Nvidia to resume H200 sales in exchange for 25% of revenue going to the US government. The H200 is approximately 6× more powerful than the H20. This is a significant policy reversal from the “small yard, high fence” regime the AI 2027 essay assumed would persist. Status downgraded to mixed.
2025-10	BIS 50% Rule takes effect October 1. GAIN AI Act passes in Senate NDAA amendment (October 9), requiring chipmakers to prioritize U.S. customers before exporting. Export control regime tightening procedurally, though Chinese labs continue releasing frontier-competitive models.
2025-09	BIS 50 Percent Rule announced (effective September 29), extending export control restrictions to all companies 50%+ owned by blacklisted entities — closing a key enforcement gap. DeepSeek reveals R1 training cost of $294,000, demonstrating China can build competitive reasoning models with dramatically lower compute — counterevidence to the “compute scarcity constrains China” mechanism.
2025-07	Updated US export controls further restrict AI chip sales to China. NVIDIA develops China-specific variants within compliance limits.
2025-04	US imposes indefinite export license requirement on Nvidia H20 chips to China (April 9). Nvidia expects $5.5B in charges. The H20 was the highest-performance chip permitted under prior rules; this closes that route. Strengthens the prediction’s core mechanism.