Did China build a top-tier AI model by itself? A new report suggests Nvidia chips played a role.
By Britney Nguyen
Alibaba has an agreement for Moonshot to use its Nvidia H200s, according to Bloomberg - but the Chinese tech giant pushed back against the report
Alibaba's stock got a boost on Friday.
The release of Chinese artificial-intelligence model Kimi K3 earlier this month raised questions about China's increasing self-sufficiency in advanced technology, even as the U.S. attempts to curb its access to top tools.
But a new report suggests U.S. technology may have had a role in the origin of Kimi K3, which performed comparably to leading U.S. AI models on benchmarks - despite being purportedly far cheaper to train.
Bloomberg reported Friday that Kimi developer Moonshot AI has an agreement with cloud provider Alibaba Group Holding (BABA) for compute power to the tune of 20,000 chips from Nvidia (NVDA).
Alibaba's American depositary receipts rose 5.1% on Friday following the report, while Nvidia's stock was up 2.9%.
The chip maker's Hopper generation makes up a large share of Moonshot's overall capacity from Alibaba, the report said, citing unnamed people familiar with the matter. The compute agreement between Moonshot and Alibaba reportedly covers Nvidia's H200 chips.
An Alibaba spokesperson pushed back on the report, however, saying in a statement shared with MarketWatch that "the alleged supply of H200 chips to Moonshot from us is completely groundless."
Alibaba has its own Qwen series of AI models that compete with Kimi, and the company is a major investor in Moonshot.
The H200 and advanced chips from the Hopper family have been under strict U.S. export controls against China, which led Nvidia to develop downgraded versions in an attempt to keep its market share in the country. The H800 came first but was eventually put under trade restrictions, leading to the downgraded H20 alternative, which came under fire from the Trump administration last spring.
Nvidia has since pushed for the ability to sell its most advanced chips, including its Blackwell line, to Chinese customers - arguing that it would deter China from building competitive products that could pose a further threat to U.S. technological dominance.
The Information also reported this week that Moonshot is looking for Blackwell chips to power its next AI offering.
The use of Nvidia's advanced chips by China's leading AI labs to train models has been clear, Daniel Newman, CEO of the Futurum Group, told MarketWatch. Despite efforts to cut down on access, Newman said the practice "has been difficult to monitor and will likely be for the foreseeable future."
China has been able to use less powerful chips to train and run some AI models, but its more advanced versions "are clearly benefiting from Nvidia," Newman said in emailed comments. He also noted claims that Chinese companies have used distillation techniques to build off of models based in the U.S.
Though the U.S. government has approved some licenses for H200 shipments to China-based customers, Nvidia said on its earnings call in May that it has "yet to generate any revenue" there.
Neither Nvidia nor Moonshot responded to requests for comment from MarketWatch.
-Britney Nguyen
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
07-31-26 1623ET
Copyright (c) 2026 Dow Jones & Company, Inc.The articles, information, and content displayed on this webpage may include materials prepared and provided by third parties. Such third-party content is offered for informational purposes only and is not endorsed, reviewed, or verified by Morningstar.
Morningstar makes no representations or warranties regarding the accuracy, completeness, timeliness, or reliability of any third-party content displayed on this site. The views and opinions expressed in third-party content are those of the respective authors and do not necessarily reflect the views of Morningstar, its affiliates, or employees.
Morningstar is not responsible for any errors, omissions, or delays in this content, nor for any actions taken in reliance thereon. Users are advised to exercise their own judgment and seek independent financial advice before making any decisions based on such content. The third-party providers of this content are not affiliated with Morningstar, and their inclusion on this site does not imply any form of partnership, agency, or endorsement.
Popular
4 Stocks to Buy Before They Rise Further
2 Undervalued Stocks to Buy Before They Rebound
The 10 Best Companies to Invest in Now
The 10 Best Dividend Stocks
