Alibaba and Tencent turn to domestic chips amid U.S. export curbs
Chinese tech giants are pivoting to homegrown hardware to sustain their cloud businesses as new U.S. restrictions limit access to advanced AI processors.
The short version
- New U.S. rules require licenses for selling advanced chips to data centers in countries like Malaysia, Singapore, and Brazil.
- Tencent says its current cloud business relies on inference, which can run on a limited number of domestic chips.
- Alibaba reports that its in-house chip division, T-Head, is already capable of supporting large-scale model training.
- Tencent's president admits that the restrictions will likely slow the development of its next-generation Hunyuan model.
- Alibaba's cloud revenue growth could slow to 30% to 45% next year before potentially rebounding by 2028.
Why it matters
This shift signals a move toward technological self-reliance for China's largest tech firms. While it protects their current cloud services, it creates a bottleneck for the high-intensity computing needed to build the world's most advanced AI models.
Navigating new export rules
The U.S. Department of Commerce recently released an interim rule that expands export controls on advanced AI processors. These measures now require a license for chipmakers to sell advanced hardware to data centers in several countries, including Malaysia, Singapore, Thailand, Indonesia, and Brazil. The rule also targets large data centers that offer significant computing power to Chinese customers.
These restrictions are designed to limit the amount of high-end computing power available to Chinese companies. For giants like Alibaba and Tencent, this creates a significant hurdle for their cloud businesses and AI development. The move follows scrutiny over whether previous controls were sufficient to keep advanced chips out of the Chinese market.
The impact is already being felt across the industry. For example, the Beijing-based developer Moonshot AI reportedly scrapped a $150 million deal to access Nvidia's Blackwell chips through an Indonesian data center. These rules also make it harder for Chinese companies to manufacture their own semiconductors using overseas foundries.
Tencent's focus on inference
Tencent is drawing a clear distinction between the power needed to train new AI models and the power required to run existing ones. The company's chief strategy officer noted that their cloud business today is primarily running inference, which is the process of running an AI model. He stated that Tencent has reached a stage where this can run on a "surprisingly limited number" of domestically produced chips.
Tencent's president, Martin Lau, noted that local supply chains should be sufficient for inference, particularly for models that are smaller and more efficient. However, he acknowledged that the new restrictions are "unfortunate" for the company's long-term goals. He admitted that it will now take more time for the company to train the next version of its Hunyuan model.
Despite the delays, Lau predicted that these U.S. controls might actually speed up the development of China's domestic chip industry. He believes that all Chinese players will eventually have to rely on domestic supply chains and chips. The company reported third-quarter revenue of 207.3 billion yuan ($31 billion), with its fintech and business services segment growing 9.4% to $9.5 billion.
Alibaba's domestic chip push
Alibaba is doubling down on its own hardware to maintain its 2030 targets, which include $100 billion in external cloud revenue. The company has invested heavily in T-Head, its in-house chip division, which it plans to list in Hong Kong. CEO Eddie Wu stated that progress in developing domestic computing power is already ahead of the company's original expectations.
While Alibaba is confident in its long-term goals, the company warned of near-term friction. Wu noted that the new export control rules will affect next year's products and services, and there is still uncertainty regarding the supply chain. He indicated that cloud revenue growth might drop to a range of 30% to 45% next year before potentially rebounding to 50% or higher by 2028.
Alibaba's T-Head chips are already capable of supporting the training of large-scale models. Wu predicted that production for these chips will ramp up in 2027 or 2028. He believes that once production scales, the company will be in a better position to control the speed at which it iterates its foundation models. The firm reported third-quarter revenue of 271.5 billion yuan ($40.5 billion), with AI and cloud revenue rising 53% to $9.1 billion.
Legal and industry pressure
Alibaba is currently facing internal pressure from shareholders regarding its past ties to overseas entities. A group of shareholders sued the company, accusing it of concealing ties to a Singapore-based cloud computing company suspected of funneling chips to China. The lawsuit claims these ties contributed to a 14% drop in Alibaba's share price over four days and that the company understated the risks of export-control violations.
The broader semiconductor landscape is also shifting as a result of these policies. As Fortune reported, major manufacturers like Taiwan Semiconductor Manufacturing Company and Samsung Electronics have paused some chip deliveries to Chinese clients. These actions highlight the tightening grip on the movement of advanced technology across borders.
Data from the House Foreign Affairs Committee shows the scale of the current oversight. The Department of Commerce has approved nearly 880,000 AI chips for Chinese companies out of more than 3.4 million chips that Nvidia and AMD sought to export. This includes 75,000 chips for Alibaba and 150,000 for Tencent, illustrating the ongoing balance between trade and security.
What to watch
- Whether Alibaba's T-Head chip production meets its 2027 ramp-up targets.
- How the 30% to 45% projected cloud revenue growth for Alibaba holds up in 2025.
- Further reactions from other Chinese AI developers like Moonshot AI as they seek hardware alternatives.
Try it
Check the latest quarterly earnings reports from Alibaba and Tencent to see how they categorize their 'capital expenditure' on domestic chip production.
Sources and links
- OfficialTencent reported static.www.tencent.com
- ReportingFortune: the article we sourced
First reported by Fortune. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.