Nvidia’s Huang calls China chip curbs a mistake as board investigates targeting claims
Editorial Team
••📖7 min read
Nvidia founder and CEO Jensen Huang defends chip sales to China in an interview recorded before a board intelligence inquiry. Filings document export-control costs, Chinese sales barriers and tariff burdens, while leaving the security case unresolved.
Nvidia’s founder and CEO argues that US chip restrictions strengthen Chinese competitors.
Licenses have enabled limited sales of Nvidia’s H200 data center chips, but Nvidia says Chinese restrictions and a tariff constrain the business.
A board inquiry concerns alleged intelligence targeting; China denies the allegations, and no compromise has been established.
Jensen Huang, chief executive of AI-chip supplier Nvidia, which he founded in 1993 and has led since, has called US restrictions on chip sales to China a mistake. His defense of engagement came in an interview recorded before the board opened an investigation into reports that he was targeted by a foreign intelligence operation.
The remarks bring together two questions with different standards of evidence: whether American chipmakers lose competitive ground when excluded from China, and whether Nvidia’s information or business dealings were compromised. Company disclosures establish substantial commercial costs. They do not settle either the intelligence allegations or the national-security value of export controls.
Data center computing racks illustrated in Nvidia’s 2026 proxy statement. Image: Nvidia. Source: Nvidia proxy statement.
A familiar argument from an executive with a stake
Huang defended serving Chinese customers in a Bloomberg interview with Emily Chang filmed less than a month before November 9, according to an account of the interview. He also argued that cooperation and dialogue with Chinese researchers were necessary for AI safety.
“I think we made a mistake, and this mistake is hard to unwind, but I did my best. That's all I could do,” Huang said.
His influence reflects Nvidia’s transformation from a business focused on PC graphics into a supplier of chips and systems for AI computing. Its May proxy statement records his leadership since founding and his 2026 appointment to the president’s science and technology advisory council. His company’s access to a major market gives him a direct commercial interest in the policy he is challenging.
UC Berkeley and NVIDIA are releasing tools for AI agents to assemble and improve robot programs in simulation. A warehouse operator reports a useful production update, while unsuccessful fixes and benchmark results show why testing, speed and existing infrastructure still matter.
Editorial Team
Huang contends that restrictions have accelerated China’s domestic chip and AI development, with Huawei benefiting, as described in a report on his position. Huawei is a Chinese chip designer whose Ascend AI processors offer an alternative to Nvidia’s products.
That competition predates the latest interview. In May 2025 remarks, Huang said Nvidia’s China market share had fallen from 95% to 50% since the start of the Biden administration. Those were his figures then, not an independent measurement of its November 2026 position. The account provided no detailed market definition or measurement method for the comparison.
Those remarks also addressed the Biden administration’s AI diffusion rule, which proposed dividing countries into tiers for access to advanced chips. Nvidia’s later filing says Washington announced in May 2025 that it would rescind that rule and develop a replacement. That global framework and the China-specific licensing requirements are distinct parts of the export-control debate.
Company-reported revenue from direct customers headquartered in China, including Hong Kong. Figures cover all products; customer headquarters do not establish shipping destinations, end users or AI-chip market share. Source: Nvidia. Source: Nvidia second-quarter fiscal 2027 filing.
Licenses, blocked sales and a tariff
Nvidia designed H20 chips primarily for China, adapting its products to US restrictions. Its first-quarter fiscal 2026 results show what happened when Washington required export licenses on April 9, 2025: the company recorded a $4.5 billion charge for excess H20 inventory and purchase obligations in the quarter ended April 27.
Nvidia said it sold $4.6 billion of H20 products before the new requirement and could not ship another $2.5 billion in revenue that quarter. The inventory charge and unshipped revenue measure different effects; adding them would misstate a single loss. Its subsequent full-year results also record a $180 million H20 charge reversal in the second quarter. Despite the restrictions, total fiscal 2026 revenue reached $215.9 billion, up 65%, for the year ended January 25, 2026.
The filing for the quarter ended July 26, 2026 shows that permission from Washington did not guarantee purchases in China. Nvidia said US licenses granted from August 2025 would have allowed certain H20 sales, but Chinese government restrictions prevented it from selling its H20 inventory.
From February 2026, US licenses permitted small H200 shipments to specific Chinese customers. Nvidia said Chinese restrictions again limited sales, and it incurred a separate $0.4 billion inventory and purchase-obligation charge in the first half of fiscal 2027 as H200 demand diminished. Some shipments subsequently occurred, accounting for less than 1% of Data Center revenue in the July quarter.
Those H200 licenses required inspection in the United States before delivery. Nvidia said importing the chips for that process subjected them to a 25% tariff, which it had been unable to pass on to customers. Legal access therefore carried an additional cost for the supplier.
Nvidia said it remained effectively excluded from China’s data center compute market, while still shipping unrestricted products such as gaming and workstation graphics processors. It warned that competitors were building developer and customer ecosystems that could challenge it worldwide. That is the company’s assessment of the strategic risk, not a demonstrated measure of Chinese technological capability.
The revenue data also require care. For the six months ended July 26, 2026, Nvidia reported $12.43 billion from direct customers headquartered in China, including Hong Kong, against $13.64 billion a year earlier. Against companywide revenue of $177.84 billion and $90.81 billion respectively, that represents about 7% versus 15%, calculated from the filing’s figures. Yet China-headquartered customer revenue in the July quarter alone rose to $7.88 billion from $3.99 billion a year earlier.
These figures cover all products and classify revenue by direct customers’ headquarters. They do not identify where chips ultimately ship or are used, and they are not Nvidia’s share of China’s AI-chip market.
The inquiry concerns access as well as information
The November 9 interview account says Nvidia’s board announced an independent investigation on November 8, overseen by independent directors with external legal counsel.
It traces the intelligence allegations to The New York Times, which cited unnamed sources familiar with US intelligence reporting. Those sources said officials believed Alice Huang, a former executive at Singapore-based Nvidia chip buyer Megaspeed International, had been recruited by China’s Ministry of State Security to collect information on the CEO and potentially influence access to advanced American AI chips. The inquiry follows reports of a personal relationship between the two.
The account also says Megaspeed faced suspicions of helping China obtain restricted chips. Nvidia had previously said it checked that Megaspeed did not illegally send its AI processors to China. That company statement is not an independent finding resolving the suspicions.
Nvidia said its board had no reason at that point to believe company secrets or US national security had been compromised, according to the same account. Jensen Huang said he supported the inquiry and would cooperate.
China denied the spying allegations on November 9. Foreign Ministry spokesperson Guo Jiakun called them unsubstantiated and a smear against China, according to a report of the ministry’s briefing. The interview predates the inquiry, and Huang’s public criticism of export controls was already documented in May 2025. Neither chronology establishes whether the alleged targeting influenced any business decision.
What would resolve the security question
The board inquiry must establish whether the alleged operation affected Nvidia’s confidential information, technology or business dealings. No such finding has been reported in the accounts cited here.
Export policy has a broader stated purpose. The Commerce Department’s Bureau of Industry and Security, which administers export controls, said in its October 2022 announcement that the measures aimed to restrict China’s access to advanced chips, supercomputers and semiconductor manufacturing capabilities used for military systems and surveillance.
Assessing that policy requires evidence about the sensitive capabilities China can acquire or build, the effectiveness of enforcement and the alternatives Chinese suppliers can deliver. Sales figures alone cannot provide it. Even Nvidia’s global growth depends on more than chip demand: its July filing identifies land, electricity, data center construction and customer financing as constraints on deployment.
For Nvidia, a wider return to China requires a competitive data center product accepted by both governments, on terms customers will pay and the company can sustain. For policymakers, the unresolved test is whether restrictions meaningfully limit the capabilities they target while Chinese alternatives develop. Neither Huang’s lost-market argument nor the unproven intelligence allegations answers that question.
Cloud Pathways lets customers coordinate large TPU workloads from one Python client, extending the system Google used for Gemini training. General availability on GKE follows earlier customer access; substantial reported usage and internal recovery results still leave customer cost savings unmeasured.
NVIDIA’s GR00T 2 predicts video and robot actions together to transfer skills across bodies. Company tests show gains from human video, but selected tasks, corrective data and recovery agents qualify its strongest success rates.
Anthropic says its new AI model produced circuit boards, enclosures and firmware for three prototypes that needed no electrical fixes. The demonstration avoided factory revisions, according to the company, while leaving certification, manufacturing scale and total costs unproven.
Meta is reportedly building a consulting channel for corporate AI sales after announcing an enterprise platform led by former MongoDB CEO Chirantan Desai. Its hiring remit also covers advertising and messaging, while prospective partners already work with rival AI suppliers.
OpenAI’s $200 Chromatic: Codex Edition adds AI-assisted development to an existing cartridge-game platform. The public bundle includes a programmable cartridge, but creation requires a computer, established tools and playtesting; the store now lists a November 19 restock.
PwC US plans a December launch for a critical-thinking and judgment certificate, with completion targeted by June 2027. It will assess staff through performance reviews as the firm seeks new ways to develop skills once practiced through routine work.
Linnea Quigley and Beverly Randolph say Aftermath departed from the musical they agreed to make, exposing a gap between a marketed cast reunion and performers’ control over their screen identities.
Anthropic’s Claude Fable 5.2 and Mythos 5.2 bring lower token prices and stronger technical performance. The release also limits capabilities that could accelerate frontier AI development, as new evidence puts the pace of research close to a sensitive policy boundary.
Meta’s smallest Muse model matches Claude Haiku 5.5’s rounded intelligence score at about one-third of the benchmark task cost. The case for delegating agent work depends on caching, task difficulty and higher charges for long requests.
SpaceX proposes a gas-fired power plant and super-compute facility on about 444 acres in Starbase, Texas. A December hearing will consider local incentives, while the public notice leaves the tax concession and operating demands unspecified.
The California state senator elected to Nancy Pelosi’s open seat wants a federal AI regulator, independent safety checks and penalties for developers. His California record shows both the reach and limits of that agenda; political support from an Anthropic-funded group adds scrutiny.
Apple Creator Studio subscribers can buy extra Logic Pro assistant usage for $49.99 in the U.S. The optional packs last up to a year, but Apple’s public support document does not quantify their capacity.
Boston Dynamics has fitted its electric Atlas humanoid with scooter wheels to test learned controls against real motor limits. The research feeds into the walking robot as Hyundai plans factory deployment, but the demonstration offers no quantified reliability gains.
Nominal raised $240 million at a $3 billion valuation, three times its March valuation, to expand its hardware engineering software into AI agents and lifecycle applications. Its defense foothold and company-reported workflow gains provide a foundation, but do not establish the newer agents’ performance or economics.
The former Megaspeed executive sought backing for a data centre project worth several hundred million dollars, citing GPU access and Alibaba business. The reported pitch adds commercial detail to scrutiny of her ties to Nvidia’s chief, while espionage and chip-smuggling allegations remain unproven.
President Donald Trump offered an unnamed executive’s alleged reversal as evidence that AI risks can be managed. His account identifies neither the conduct at issue nor the legal basis for a jail threat, leaving its value as a safety precedent unproven.
The Texas startup is building factory-made data centers for AI customers, with testing reportedly planned for early 2027. Its financing backs a manufacturing strategy whose cost and power-access advantages remain to be demonstrated.
Baidu says its automated research system has improved inference for the forthcoming ERNIE-5.5-Flash, alongside public benchmark gains. The results show bounded engineering progress, while the cost of finding those improvements remains undisclosed.
