DeepSeek allegedly trained V4 model on banned Nvidia Blackwell chips, denied US chipmakers access

Exclusive: DeepSeek withholds latest AI model from US chipmakers including Nvidia, sources say

SAN FRANCISCO/SINGAPORE, Feb 25 (Reuters) - DeepSeek, the Chinese artificial intelligence lab whose low-cost model rattled global markets last year, has not shown U.S. chipmakers its upcoming flagship model for performance optimization, two sources familiar with the matter said, breaking from standard industry practice ahead of a major model update. Instead, the lab, which is expected to launch its next major update, V4, granted early access to domestic suppliers, including Huawei Technologies, the sources said. AI developers typically share pre-release versions of major models with leading chipmakers such as Nvidia (NVDA.O), opens new tab and Advanced Micro Devices (AMD.O), opens new tab to ensure their software performs efficiently on widely used hardware. DeepSeek has previously worked closely with Nvidia's technical staff. NVIDIA, AMD LEFT OUT For its forthcoming model, which was expected to be released around the Lunar New Year holiday, DeepSeek did not provide access to Nvidia and AMD and gave Chinese chipmakers, including Huawei, a head start of several weeks to optimize the software for their processors, the sources said. Nvidia and AMD declined to comment. DeepSeek and Huawei did not respond to requests for comment. Reuters could not immediately determine the reason for the decision. "The impact to Nvidia and AMD for general data accelerators is minimal - most enterprises are not running DeepSeek, which serves as a benchmarking model more than anything else," said Ben Bajarin, CEO of research firm Creative Strategies. He added that new AI coding tools are reducing the time it takes to make software run well on hardware, "from months to weeks." The move is likely part of a broader strategy by the Chinese government "to try to keep U.S. hardware and models disadvantaged" in China, Bajarin said. The development comes as a senior Trump administration official told Reuters DeepSeek's latest AI model was trained on Nvidia's most advanced chip, Blackwell, using a cluster in mainland China, in a move that appears to violate U.S. export controls. DeepSeek may seek to remove technical indicators revealing its use of American AI chips, and plans to publicly claim that it used Huawei's chips to train its model, according to the U.S. official. DeepSeek's models have been downloaded more than 75 million times on the open-source platform Hugging Face since the company burst onto the scene in January 2025, helping fuel a wave of Chinese open-source models competing with U.S. AI labs. Among models released in the past year, downloads for Chinese models have surpassed those from any other country on the platform. The rapid rise of Chinese open-source models has intensified debate in Washington over exporting advanced U.S. AI chips to China. U.S. authorities last year allowed Nvidia's H20 and AMD's MI308 chips -- designed for AI inference -- to resume shipments to China, even as licenses for more advanced processors remain restricted. It was unclear whether DeepSeek has secured approval to purchase those U.S. chips. The H20 and MI308 chips are aimed at inference, the process of running trained AI models. Demand for the MI308 was significant, with AMD saying it generated $390 million in sales of the chip in its most recent quarter. DeepSeek is among several Chinese AI firms expected to unveil new models this month. Reporting by Krystal Hu, Stephen Nellis in San Francisco and Fanny Potkin in Singapore; Editing by Kenneth Li, Rod Nickel Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * China Krystal Hu Thomson Reuters Krystal reports on venture capital and startups for Reuters. She covers Silicon Valley and beyond through the lens of money and characters, with a focus on growth-stage startups, tech investments and AI. She has previously covered M&A for Reuters, breaking stories on Trump's SPAC and Elon Musk's Twitter financing. Previously, she reported on Amazon for Yahoo Finance, and her investigation of the company's retail practice was cited by lawmakers in Congress. Krystal started a career in journalism by writing about tech and politics in China. She has a master's degree from New York University, and enjoys a scoop of Matcha ice cream as much as getting a scoop at work.

Exclusive: China's DeepSeek trained AI model on Nvidia's best chip despite US ban, official says

WASHINGTON, Feb 23 (Reuters) - Chinese AI startup DeepSeek's latest AI model, set to be released as soon as next week, was trained on Nvidia's (NVDA.O), opens new tab most advanced AI chip, the Blackwell, a senior Trump administration official said on Monday, in what could represent a violation of U.S. export controls. The official said the U.S. believed DeepSeek would remove the technical indicators that might reveal its use of American AI chips. The official declined to say how the U.S. government obtained the information. Nvidia declined to comment. The Chinese embassy in Washington said in a statement that Beijing opposes "drawing ideological lines, overstretching the concept of national security, expansive use of export controls and politicizing economic, trade, and technological issues." The Commerce Department and DeepSeek did not immediately respond to requests for comment. The official did not provide information on how DeepSeek obtained the Blackwells, but noted that U.S. policy is "we're not shipping Blackwells to China," emphasizing that DeepSeek's possession of the chips could represent an export control violation. The news, not previously reported, could further divide Washington policymakers as they struggle to determine where to draw the line on Chinese access to the crown jewels of American AI semiconductor chips. China hawks fear chips could easily be diverted from commercial uses to help supercharge China's military and threaten U.S. dominance in AI. But White House AI Czar David Sacks and Nvidia CEO Jensen Huang argue that shipping advanced AI chips to China discourages Chinese competitors like Huawei from redoubling efforts to catch up with Nvidia's and AMD's (AMD.O), opens new tab technology. U.S. export controls, overseen by the Commerce Department, currently bar Blackwell shipments to China. In August, U.S. President Donald Trump opened the door to Nvidia selling a scaled-down version of the Blackwell in China. But he later reversed course, suggesting the firm's most advanced chips should be reserved for U.S. companies and kept out of China. Trump's decision in December to allow Chinese firms to buy Nvidia's second most advanced chips, known as the H200, drew sharp criticism from China hawks, but shipments of the chips remain stalled over guardrails built into the approvals. The official declined to comment on how the latest news would impact the Trump administration's decision on whether to allow DeepSeek to buy H200s. The U.S. official also said DeepSeek's Blackwells are likely part of a cluster at its data center in Inner Mongolia, an autonomous region of China. The model they helped train likely relied on the "distillation" of models made by leading-edge U.S. AI companies, including Anthropic, Google, OpenAI, and xAI, echoing allegations made by OpenAI and Anthropic, the official added. The technique known as distillation involves having an older, more established and powerful AI model evaluate the quality of the answers coming out of a newer model, effectively transferring the older model's learnings. Hangzhou-based DeepSeek shook markets early last year with a set of AI models that rivaled some of the best offerings from the U.S., fueling concerns in Washington that China could catch up in the AI race despite restrictions. Reporting by Steve Holland and Alexandra Alper; Editing by Chris Sanders and Sonali Paul Our Standards: The Thomson Reuters Trust Principles., opens new tab

DeepSeek has reportedly denied Nvidia and AMD early access to its new V4 AI model, giving Huawei and other Chinese chipmakers a head start

Sources speaking to Reuters have claimed that DeepSeek, the Chinese AI company that gave the US AI market a scare last year with its surprisingly-competitive AI models, has withheld both Nvidia and AMD from early access to its upcoming V4 update. However, early access is said to have been granted to domestic suppliers, including Huawei. If true, this would be a break from standard industry practices, allowing Chinese chipmakers several weeks to optimise their processors for the new model while their US competitors are left twiddling their thumbs. All this comes off the back of fresh concerns over US/China chip export controls. An unnamed "senior Trump administration official" has claimed that the V4 model was trained on a cluster of Nvidia Blackwell chips located in Inner Mongolia, an autonomous region of China. This would be in breach of current US policy on the matter, which the source confirmed was: "We're not shipping Blackwells to China." While the US government has approved sale of Nvidia's older H200 chips to the country, its newer and much more powerful Blackwell chips appear to be an absolute sticking point. President Trump has repeatedly confirmed that Nvidia's Blackwell hardware would not sold to Chinese customers at any time in the foreseeable future -- although reports of chip smuggling efforts suggest that many have made it to the country already. Last year, Nvidia CEO Jensen Huang claimed that China was "nanoseconds behind" the US in terms of AI technology and chipmaking, and has long argued that the company should be allowed to sell its chips to Chinese customers. And while H200 GPUs now appear to have been approved on both sides of the border, and Chinese buyers are lining up to get their hands on Nvidia's tech, it seems the adversarial relationship between the two nations in regards to the AI race continues. The US has played its cards in regards to holding back certain AI hardware advances via export controls, but some analysts have suggested that DeepSeek's recent move may be part of a broader strategy by the Chinese government to keep the US on the backfoot. Given the success of DeepSeek's earlier efforts -- and assuming Reuters' sources are correct -- providing Chinese firms with an extra, early advantage towards its new AI models strikes as the most recent attempt to even the odds.

US Claims China's DeepSeek Used Banned Nvidia Blackwell Chips - NVIDIA (NASDAQ:NVDA)

U.S.-China AI tensions intensified after a senior Trump administration official said Washington believes Chinese startup DeepSeek used Nvidia Corp.'s (NASDAQ:NVDA) most advanced chips to train its upcoming model, raising fresh export control concerns. U.S. Flags Potential Blackwell Export Violation A senior Trump administration official said U.S. officials believe Chinese AI startup DeepSeek trained its next AI model, which could launch by next week, on Nvidia's most advanced AI chip, the Blackwell, potentially in violation of U.S. export controls. The official said the U.S. expects DeepSeek to strip out technical indicators that could reveal the use of American AI chips and believes the Blackwell chips are likely clustered at DeepSeek's data center in Inner Mongolia, Reuters reported on Monday. Benzinga has contacted Nvidia's investor relations team for comment and is awaiting a response. The report adds pressure to an already divisive debate in Washington over where to draw the line on China's access to top-tier U.S. AI chips. The Commerce Department's export controls currently bar Blackwell shipments to China. Trump previously signaled openness to Nvidia selling a scaled-down Blackwell in China in August, but later reversed course, arguing the most advanced chips should stay with U.S. companies. China's AI Push And Nvidia's Supply ResponseStrong Global Demand And China's Revenue Potential Despite export uncertainty, Nvidia continues to benefit from robust global AI demand. The company has disclosed a backlog exceeding $500 billion as hyperscalers invest heavily in AI infrastructure. JPMorgan analyst Harlan Sur said U.S. export approvals could unlock billions in additional revenue for Nvidia, making China a significant potential upside catalyst for the stock. Chinese customers continue to show a strong appetite for AI compute, and any easing in export approvals could immediately convert that demand into revenue. NVDA Price Action: Nvidia shares were up 0.37% at $192.25 during premarket trading on Tuesday, according to Benzinga Pro data. Photo by Below the Sky via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

DeepSeek trained upcoming AI model on Nvidia's new AI chip

Chinese AI firm DeepSeek's (DEEPSEEK) latest AI model, which is expected to be released as soon as next week, was trained on Nvidia's (NVDA) most advanced AI chip series called Blackwell, Reuters reported, citing a senior Trump administration official. If Potential US policy violations increase regulatory scrutiny and could create risks for Nvidia and related firms' investor sentiment. DeepSeek's reliance on model distillation from US AI firms could raise intellectual property and competitive concerns for leading US AI companies such as Anthropic, Google, OpenAI, and xAI. Changing US-China chip export regulations, including tariffs and selective approvals, affect Nvidia's and AMD's ability to access the Chinese market and influence their revenue and growth prospects.

DeepSeek withholds latest AI model from US chipmakers including Nvidia, sources say

SAN FRANCISCO/SINGAPORE, Feb 25 (Reuters) - DeepSeek, the Chinese artificial intelligence lab whose low-cost model rattled global markets last year, has not shown U.S. chipmakers its upcoming flagship model for performance optimization, two sources familiar with the matter said, breaking from standard industry practice ahead of a major model update. Instead, the lab, which is expected to launch its next major update, V4, granted early access to domestic suppliers, including Huawei Technologies, the sources said. AI developers typically share pre-release versions of major models with leading chipmakers such as Nvidia and Advanced Micro Devices to ensure their software performs efficiently on widely used hardware. DeepSeek has previously worked closely with Nvidia's technical staff. NVIDIA, AMD LEFT OUT For its forthcoming model, which was expected to be released around the Lunar New Year holiday, DeepSeek did not provide access to Nvidia and AMD and gave Chinese chipmakers, including Huawei, a head start of several weeks to optimize the software for their processors, the sources said. Nvidia and AMD declined to comment. DeepSeek and Huawei did not respond to requests for comment. Reuters could not immediately determine the reason for the decision. "The impact to Nvidia and AMD for general data accelerators is minimal - most enterprises are not running DeepSeek, which serves as a benchmarking model more than anything else," said Ben Bajarin, CEO of research firm Creative Strategies. He added that new AI coding tools are reducing the time it takes to make software run well on hardware, "from months to weeks." The move is likely part of a broader strategy by the Chinese government "to try to keep U.S. hardware and models disadvantaged" in China, Bajarin said. The development comes as a senior Trump administration official told Reuters DeepSeek's latest AI model was trained on Nvidia's most advanced chip, Blackwell, using a cluster in mainland China, in a move that appears to violate U.S. export controls. DeepSeek may seek to remove technical indicators revealing its use of American AI chips, and plans to publicly claim that it used Huawei's chips to train its model, according to the U.S. official. DeepSeek's models have been downloaded more than 75 million times on the open-source platform Hugging Face since the company burst onto the scene in January 2025, helping fuel a wave of Chinese open-source models competing with U.S. AI labs. Among models released in the past year, downloads for Chinese models have surpassed those from any other country on the platform. The rapid rise of Chinese open-source models has intensified debate in Washington over exporting advanced U.S. AI chips to China. U.S. authorities last year allowed Nvidia's H20 and AMD's MI308 chips -- designed for AI inference -- to resume shipments to China, even as licenses for more advanced processors remain restricted. It was unclear whether DeepSeek has secured approval to purchase those U.S. chips. The H20 and MI308 chips are aimed at inference, the process of running trained AI models. Demand for the MI308 was significant, with AMD saying it generated $390 million in sales of the chip in its most recent quarter. DeepSeek is among several Chinese AI firms expected to unveil new models this month. (Reporting by Krystal Hu, Stephen Nellis in San Francisco and Fanny Potkin in Singapore; Editing by Kenneth Li, Rod Nickel)