Outside of China, Alibaba is mostly known as an e-commerce titan.
But inside the country, the company is obsessed over catching up to DeepSeek on its development of AI models, and catching up to Huawei on the chips that power them.
When Alibaba’s chip design unit T-Head unveiled its latest AI chip, the Zhenwu M890, last week, it also outlined a multi-year chip roadmap showing how the M890’s future successors would deliver massive performance gains in the next few years. Less than a year ago, Huawei had laid out a similar timeline that ran until 2028.
Unlike Huawei, Alibaba has not been targeted by U.S. sanctions, nor has it made its AI chip push as explicitly about beating Nvidia as the Shenzhen-based conglomerate has. But its decision to chase Huawei highlights the huge business opportunity presented by Beijing’s push to replace Nvidia.
Whatever the real size of the void left by Nvidia, Alibaba and other domestic companies are not ready to hand it over to Huawei. Of the 1.65 million chips shipped by Chinese vendors in 2025, 812,000 came from Huawei, while Alibaba claimed second place, shipping around 265,000. Several other tech giants and startups are eyeing Huawei’s half of the pie.
As China nudges its AI industry to a post-Nvidia era, Huawei will face more competition. Besides Alibaba, ByteDance and Tencent have the resources to mount a serious challenge. DeepSeek’s unquestionable lead among Chinese AI models lasted less than a year. Huawei might only have a few more.
In this week’s issue, we look at how Chinese chipmakers are trying to displace Nvidia from their country while searching for alternative chipmaking processes and theories to neutralise U.S. sanctions.
OUR LATEST REPORTING IN TECH AND AI
Huawei's 'chip queen' etches her name in China's tech folklore
Anthropic's Olah says AI must be guided from outside Big Tech
Pope, urging AI regulation, warns some weapons now beyond human control
Trump postpones AI executive order, cites need to compete with China
Exclusive - Grok falls flat in Washington, undercutting SpaceX's AI growth story
The AI bots are coming and the young are booing, not applauding
BREAKTHROUGH OR HYPE?
Huawei wants the world to remember the name He Tingbo, the semiconductor and telecom giant’s 30-year company veteran and “chip queen” it aims to write into the annals of tech history. It has (sort of) named a new tech guiding principle after her. This week, Huawei claimed to have discovered and applied a new principle of chipmaking, the Tau Scaling Law, also clunkily named “Her’s Law” in He’s honour.
The idea is framed as a replacement of Moore’s Law, the universally recognized pattern that describes how advances in computing are driven by chipmakers finding ways to shrink transistors and pack ever more of them onto a single chip, making computers faster, cheaper, and more energy efficient.
But as transistors approach atomic-scale limits, Moore’s Law is under strain. Instead of playing this game, Her’s Law looks for system-level performance gains, where the way a group of chips is integrated and works together matters most.
Huawei claims Her’s Law will allow it to produce a chip with transistor density equivalent to 1.4 nanometers by 2031. TSMC is expected to roll out 1.4nm chips in 2028. The eye-catching projection is likely to raise alarm in Washington, which has sought to halt China’s AI development by, among other measures, making it illegal for Huawei and other Chinese companies to acquire the tools — namely ASML lithography machines — that are needed to make sub-7nm chips.
Had Huawei actually developed a sanctions-resistant chipmaking technology as the projection seems to suggest, the implications for Chinese AI would be even greater than DeepSeek’s low-cost models tanking global tech stocks. But for now, competitors outside of China have shrugged it off. When U.S. markets opened on Tuesday, Nvidia, AMD, and Intel shares rose.
That’s because transistor density is not yet equal to performance. Moreover, techniques developed under Her’s Law will feature in phone chips later this year. For the company’s Ascend AI chips, the integration of this new principle will only be complete in 2030.
Huawei is hoping this strategy can improve the efficiency of AI clusters and data centers by drastically cutting the time and energy these systems spend on moving data between and within chips. But it’s not clear how issues like overheating will be solved by Her’s Law.
Moreover, post-Moore’s Law solutions are not new. Industry stateside has invested in these alternative pathways, as have Chinese researchers and firms, from advanced packaging to chiplets.
Huawei has in recent years repeatedly sought to signal that U.S. sanctions have failed to halt the development of its chip business. Its latest unveiling, however, suggests its investment, some of it secret, in creating China’s answer to ASML has failed to achieve breakthroughs, pushing it to double down on a good-enough alternative path.
