MobbleOpen in Mobble ⇢
Technology · Artificial intelligence · published 2026-11-16 · via Tom's Hardware

US Introduces New Rules to Block Chinese Firms from Bypassing AI Hardware Bans via Cloud Access

Image via Tom's Hardware
Image via Tom's Hardware

The U.S. Bureau of Industry and Security (BIS), in coordination with the Treasury Department and National Security Council, has issued an interim final rule to prevent non-whitelisted entities, particularly in China, from circumventing GPU export controls by accessing AI servers remotely through third-party cloud data centers. The move targets cloud rental workarounds that undermine existing restrictions on sensitive American technology.

Expanded Detail

On 16 November 2026, the U.S. Bureau of Industry and Security (BIS), working with the Treasury Department and National Security Council, introduced an interim final rule (IFR) to close a growing loophole in AI hardware export controls. The rule specifically targets non-whitelisted entities—particularly those based in China—from accessing powerful American AI servers via third-party cloud data centers located in countries like Indonesia, Malaysia, Singapore, Thailand, and Brazil. These cloud-based arrangements had allowed Chinese firms to circumvent physical export bans on high-performance GPUs, such as Nvidia’s Blackwell Ultra B300 chips, by renting remote access to AI infrastructure. The new rule mandates that foreign data centers using U.S.-origin technology must restrict access to companies headquartered in higher-risk nations and require licenses for any exports of advanced chips to firms in those regions. The regulation will take effect in 90 days. This move follows earlier reports of black market GPU sales and the discovery of restricted Nvidia chips in a China Telecom data center, highlighting the limitations of prior export controls.

The background includes years of escalating U.S.-China tech tensions, with previous export restrictions already limiting sales of advanced AI chips to Chinese companies. However, as physical exports became harder to manage, cloud rental models emerged as an effective workaround. Nvidia itself had recently urged data center operators to limit their business with Chinese clients, requiring at least 51% of revenue come from non-Chinese sources—yet enforcement remained inconsistent. The new rule reflects a broader effort by U.S. agencies to prevent circumvention through digital infrastructure. Key players include BIS, which enforces export rules; the Treasury Department, involved in financial sanctions coordination; and the National Security Council, which assesses strategic risks. While the intent is clear—to safeguard national security by limiting adversarial access to cutting-edge AI—it remains uncertain how effectively these measures can be enforced across global cloud networks or whether they will spur further innovation in evasion tactics.

Context

The new U.S. rule could significantly impact global cloud infrastructure providers, particularly those with operations in Southeast Asia and other regions serving high-risk jurisdictions. Companies may face increased compliance burdens, potentially leading to higher operational costs or reduced service availability in certain markets. For China’s AI industry, which has already faced severe constraints on hardware access, this move may further slow innovation and limit the scale of domestic large language model development. However, it could also accelerate efforts to build sovereign cloud ecosystems and indigenous AI infrastructure, reinforcing long-term strategic decoupling between U.S. and Chinese tech sectors. A comparable precedent is the 2019 export controls on Huawei, which initially restricted chip access but led to rapid domestic alternatives like Huawei’s Ascend chips. This new rule may similarly drive China to invest more heavily in closed-loop AI systems. What remains uncertain is whether these measures will be effective in practice, given the difficulty of monitoring remote server access across borders. The next phase will likely involve testing enforcement mechanisms and assessing how quickly cloud providers adapt—possibly through encryption or decentralized architectures that obscure user origins.

Expanded detail and Context are AI-generated analysis; the linked article remains the authoritative source.
Read the full article at Tom's Hardware →
Related stories
US Expands AI Chip Controls to Block China from Using Overseas Data Centers for AI Training · Artificial intelligence
Samsung and TSMC halt certain chip exports to China amid new US controls · Semiconductors
Trump Targets Philippines for AI Infrastructure Amid Domestic Opposition · Artificial intelligence
Signal Implements Default Screenshot Restrictions to Prevent AI Agents from Accessing Chats · Cybersecurity
This summary is Al-enhanced to contain extended analysis and broader social context. The original is {NAME); the linked article is the authoritative source. Original headline: “New US rules aim to restrict Chinese firms’ remote access to AI servers and infrastructure.” Browse more stories.