The Tamesis
People are awaiting the launch of China DeepSeek’s latest artificial intelligence model, which will be seen as a benchmark for Beijing’s progress in technology and semiconductor development.
It’s been more than a year since the startup put Chinese artificial intelligence on the map in early 2025 with a low-cost chatbot that shocked Silicon Valley.
Despite reports and rumors of an imminent release, DeepSeek’s next-generation “V4” model is nowhere in sight.
AF also reported: Samsung Electronics expects AI boom to drive profit jump 755%
Speculation is also rife over the geopolitical implications of which computer chip will be chosen to train and power the new system: a world-leading American design or a Chinese-made alternative that China is racing to develop.
“It’s important to understand this because, to some extent, it’s a signal of China’s self-sufficiency trajectory in AI,” Sun Wei, chief AI analyst at Counterpoint Research, told AFP.
Tech news outlet The Information reported last week that V4 could run on the latest chips made by China’s Huawei.
The shift would be a milestone in China’s bid to break U.S. restrictions on California giant Nvidia’s export of top artificial intelligence chips to China.
The report cited five people with direct knowledge of Huawei’s large chip orders, which are in preparation for the launch of DeepSeek by technology giants such as Alibaba, ByteDance and Tencent.
Model breakthrough or development slowdown?
DeepSeek was launched in 2023 as a side project for a hedge fund that had access to powerful Nvidia processor caches.
In January 2025, it attracted attention with its R1 deep inference chatbot, causing U.S. technology stocks to plummet and President Donald Trump calling it a “wake-up call” for American companies.
R1 is based on DeepSeek’s last major AI model, V3, which was released in December 2024.
The company’s affordable, customizable AI tools have been widely adopted in China and are also popular in emerging markets such as Southeast Asia and the Middle East.
V4 – which is said to be multi-modal, meaning it can generate text, images and videos – could once again shock the valuations of US technology companies, Stephen Wu, founder of Carthage Capital Fund, told AFP.
“I anticipate that the upcoming DeepSeek V4 release will be more than just a software update; it will be a powerful open source model capable of handling large numbers of context windows at very low cost,” he predicted.
But DeepSeek’s reputation as a company at the forefront of artificial intelligence technology is also at stake.
Wei said its models previously relied on Nvidia chips, so the move to partner with domestic chipmakers will require “a lot of redesign.”
“This shift could slow down development cycles and introduce performance tradeoffs, especially for V4, the model expected to be the most advanced.”
Potential uses of Huawei chips
The United States has banned the export of Nvidia’s most powerful artificial intelligence processor to China, citing national security concerns.
“The continued wait for DeepSeek V4 demonstrates the friction in scaling advanced models without unrestricted access to top-tier Nvidia hardware,” Wu said.
But some reports say DeepSeek circumvented the ban by using thousands of Nvidia’s high-end Blackwell chips to train V4, which were dismantled in third countries and smuggled into China.
Training AI models requires massive amounts of computing power—much more than processing generative AI queries (known as inference). Nvidia told The Information that it has not seen evidence of this and that “such smuggling seems far-fetched.”
Meanwhile, another Chinese AI startup, Zhipu, launched an image generator in January that it said was trained entirely on Huawei chips.
Alibaba said this week it would open a new artificial intelligence training and inference data center in southern China, powered by 10,000 of its own chips and operated by China Telecom.
As for DeepSeek, “If they successfully train V4 entirely on Huawei chips, it would mark a major shift in the geopolitical technology landscape,” Wu said.
- Vishakha Saxena Additional Editor AFP
Also read:
Nexperia’s Chinese subsidiary’s chips “nearly fully localized”
China beats U.S. again in global patent battle, Huawei maintains lead
China now requires chipmakers to use at least 50% domestic equipment
Bytedance’s viral video model shows China’s growing control over AI
Study finds artificial intelligence majors fail security tests, with Chinese companies worst performing
DeepSeek researchers are pessimistic about AI’s impact on humanity
Nvidia CEO says U.S. export restrictions on AI chips are ‘flawed’ policy
China actively lobbies for Nvidia H200, but local chips remain key
China’s big technological secret: Dutch chip-making machine replicated in laboratory
Chinese AI firms form alliance with chipmakers to ditch foreign technology
Satellite images show Huawei’s expanding chip facilities – FT
China’s “tech economy” trumps its bleak growth prospects
4,000 CEOs say AI has no impact: Study spurs massive sell-off
Researchers say artificial intelligence doesn’t make work easier, it just makes it harder
Beijing and its billions in artificial intelligence
