ByteDance is in advanced talks to spend an additional USD 500 million on Microsoft’s Azure cloud services, primarily to access OpenAI models, according to a November 18 report by The Information. The negotiations have drawn criticism from US lawmakers, adding to scrutiny of how Chinese technology companies obtain American artificial intelligence capabilities as they expand overseas.

The Information, which first reported the proposed deal, cited two people familiar with the discussions. The additional spending would expand an existing relationship between ByteDance and Microsoft, rather than establish a new one. The companies have not announced a finalized agreement, and it is not yet clear which ByteDance products or operations would use the additional capacity.

The potential arrangement also highlights a distinction in how OpenAI technology reaches Chinese customers. OpenAI restricts access to its models for most Chinese companies, but Microsoft, which holds commercialization rights under its partnership with the AI developer, does not apply the same restrictions, according to the report. The difference has created friction between the partners over Microsoft’s dealings with Chinese customers.

Microsoft told Reuters that it takes misuse of its AI services seriously and is committed to enforcing its terms of service, complying with applicable laws, and honoring partner agreements. It declined to discuss confidential customer negotiations. ByteDance has yet to publicly comment on the reported talks or the objections raised by US lawmakers.

On November 18, the House Select Committee on China requested a bipartisan briefing from Microsoft about the potential arrangement. Chairman John Moolenaar, a Republican representative from Michigan, said the committee was concerned that American companies could help China narrow the gap with the US in AI.

“Microsoft should not sell AI tools to ByteDance,” Moolenaar said in a statement. He argued that the arrangement would be particularly troubling given Microsoft’s contracts with the Pentagon, criticizing US technology companies for supporting Chinese AI development while also working with Washington.

While the committee’s intervention does not mean that the proposed deal would violate US law, it shows how commercial arrangements involving Chinese technology companies can face political opposition even when services are provided by established US companies.

The implications extend beyond ByteDance: Chinese companies developing products for international markets increasingly rely on a combination of proprietary technology, third-party AI models, and computing infrastructure distributed across borders. Access to these resources may depend not only on commercial agreements, but also on decisions by US regulators and technology suppliers.

The debate comes amid the controversy surrounding Nvidia, which has made these dependencies more salient. Nvidia co-founder Jensen Huang resigned as chairman, CEO, and board member earlier this month amid allegations concerning his personal relationships. Scrutiny has also intensified around Singapore-based cloud computing company Megaspeed, an Nvidia partner under investigation over concerns that it may have helped Chinese customers obtain advanced chips in violation of US export restrictions.

On November 16, the US Department of Commerce announced an interim final rule addressing remote access to advanced American semiconductors. The rule, which takes effect on November 19, targets arrangements in which companies in restricted jurisdictions access computing power through data centers elsewhere, rather than importing the chips directly.

Under the rule, exports and transfers of covered advanced chips and servers to data centers worldwide can require a license when the facilities provide computing power above specified thresholds to companies headquartered in China or other restricted markets. Certain existing arrangements have 90 days to obtain authorization or cease the covered activity.

The rule also extends destination-specific licensing requirements for advanced computing hardware to Singapore, Malaysia, Indonesia, Thailand, and Brazil, with exceptions for qualifying US-headquartered customers. The new controls primarily target hardware exports and remote access to advanced computing capabilities rather than impose a blanket ban on sales to those markets.

In a November 18 analysis, Kate Koren and Kevin Kurland of the Center for Strategic and International Studies said the restrictions are nevertheless likely to slow Chinese access to advanced computing capacity in the near term. The two analysts noted that suppliers face greater due diligence obligations and licensing burdens, while data center projects could face delays. They also cautioned that procurement networks could shift to new locations, requiring regulators to continue adapting their approach.

The disputes are unfolding as US President Donald Trump and Chinese President Xi Jinping are scheduled to meet in Shenzhen on November 19 during the Asia-Pacific Economic Cooperation summit, which is also being attended by other world leaders, including Russian President Vladimir Putin. AP previously reported, citing administration officials, that Trump did not plan to confront Xi over the Nvidia controversy. There is no indication of whether the ByteDance-Microsoft talks will be discussed.