← Latest articles

Nvidia Leadership Scandal Reaches Alibaba as US Scrutiny Focuses on Foreign Chip Rentals

This report is part of our Nvidia leadership timeline. Follow the dated updates and source status there.

Jensen Huang — file photograph.
Photo credits
Photo: ETC-USC · CC BY 2.0 · Cropped to 16:9 around reviewed heads and significant details; resized and converted to WebP.
Jensen Huang (file photo).

The leadership scandal surrounding Nvidia has shifted attention toward Alibaba as US officials examine how Chinese artificial intelligence developers rent computing power overseas. The situation highlights Alibaba's central role in China's artificial intelligence sector and the growing friction over how foreign technology is accessed.

Companies worldwide are increasingly adopting open-weight artificial intelligence models—systems that anyone can download and run for free—because they are much cheaper to operate than proprietary software. Duolingo co-founder and CEO Luis von Ahn stated at Yahoo Finance Invest that each AI practice call in his app used to cost about 30 cents, but dropped to "less than a cent" after "moving most of this to open weight models". Decagon co-founder and CEO Jesse Zhang added that new customers at his AI startup run "almost entirely" on those same freely accessible models.

Among these, the most widely used family of free models comes from China, with Alibaba serving as the largest producer by a wide margin. Hugging Face co-founder Thomas Wolf noted on a broadcast that "Right now there is massive usage of ... a Chinese family called Qwen, which is the most used one". These Chinese models have closed the performance gap with American alternatives significantly. NYU Stern professor emeritus Nouriel Roubini explained at the Invest conference that Chinese open models achieve about 95% of the capability of US models because "we allowed distillation — that's over — and we allowed them through Singapore to get all these advanced chips". Distillation refers to the process where a newer AI model learns by copying the answers of a stronger model.

The scrutiny intensified after Nvidia CEO Jensen Huang resigned following his admission to an affair with a woman that US intelligence officials believe passed information to Beijing. Reports identified her as Alice Huang, a former executive at Megaspeed, a Singapore-based company that rents out computing power built on Nvidia chips. Financial Times reports indicate that Alibaba was Megaspeed's founding customer and remains its largest client. While the US investigation focuses on Megaspeed rather than Alibaba, and Megaspeed denies diverting chips, the connection has drawn an immediate market reaction. Alibaba shares fell 7% to their lowest close since January 2025 as reports circulated that Washington intends to close the Singapore route.

Because China is legally barred from buying Nvidia's most advanced chips directly, renting them through foreign intermediaries has served as a workaround. The US Commerce Department is now finalizing rules that would prohibit Chinese artificial intelligence companies from utilizing advanced chips in data centers located outside China, including in Singapore and Malaysia. Furthermore, financial analysts point out that replacing the computing capacity Alibaba rents through Megaspeed could prove both costly and slow as the company faces tightening export controls and heavy spending.

Summary

  • The investigation into Megaspeed and potential US restrictions on overseas chip rentals threaten the computing capacity that powers Alibaba's expanding AI cloud services.
  • The departure of Nvidia's CEO removes a vocal supporter of open-weight models, potentially empowering US lawmakers pushing for stricter technology export curbs.
  • Without access to advanced chips housed abroad and the ability to train on stronger models, Chinese AI developers may face renewed hurdles in keeping pace with US advancements.