DeepSeek V4 model to run on Huawei chips, testing China's AI self-sufficiency ambitions

Waiting for DeepSeek: new model to test China's AI ambitions

Tokyo (AFP) - For weeks now, the global tech industry has been waiting for a major artificial intelligence launch from DeepSeek, seen as a benchmark for China's progress in the fast-moving field. More than a year has passed since the startup put Chinese AI on the map in early 2025 with a low-cost chatbot that performed at a similar level to US rivals. But despite reports and rumours about its imminent release, DeepSeek's next-generation "V4" model is nowhere in sight. Speculation is also swirling over the geopolitical implications of which computer chips were chosen to train and power the new system: world-leading US designs or made-in-China alternatives that the country is racing to develop. "It's important to know because at one level, it is a signal of China's AI self-sufficiency trajectory," Wei Sun, principal AI analyst at Counterpoint Research, told AFP. Tech news outlet The Information reported last week that V4 can be run on the latest chips made by China's Huawei. Such a shift would mark a milestone for China in its bid to beat US restrictions on the export of top-of-the-range AI chips from Californian titan Nvidia to the country. The report cited five people with direct knowledge of large orders for Huawei chips, made in preparation for the DeepSeek launch by tech giants including Alibaba, ByteDance and Tencent. AFP contacted DeepSeek, Huawei, Alibaba, ByteDance and Tencent but none were able to comment. 'Wake-up call' DeepSeek started life in 2023 as a side project of a hedge fund that had access to a cache of powerful Nvidia processors. It shot to attention in January 2025 with its R1 deep-reasoning chatbot, which sent US tech shares tumbling with President Donald Trump calling it a "wake-up call" for American firms. R1 was based on DeepSeek's last major AI model, V3, which was released in December 2024. The company's affordable, customisable AI tools have been widely adopted in China, and are also popular in emerging markets such as Southeast Asia and the Middle East. Stephen Wu, founder of the Carthage Capital fund, told AFP that V4 -- said to be multimodal, meaning it can generate text, pictures and video -- could again shock US tech valuations. "I expect the upcoming DeepSeek V4 release will not just be a software update; it will be a highly capable, open-source model that handles massive context windows at a fraction of the cost," he predicted. But DeepSeek's reputation as a company at the frontier of AI technology is also at stake. Its models previously relied on Nvidia chips, so a move to collaborate with domestic chipmakers would require "substantial re-engineering", Wei said. "That transition can slow development cycles and introduce performance trade-offs, especially for V4, a model expected to be state-of-the-art." Training vs inference The US cites national security concerns as the reason for its export ban on Nvidia's most powerful AI processors to China. "The ongoing wait for DeepSeek V4 points to friction in scaling advanced models without unrestricted access to top-tier Nvidia hardware," Wu said. But some reports allege that DeepSeek skirted the ban to train V4 using thousands of Nvidia's top-end Blackwell chips, dismantled in third countries and smuggled to China. Training AI models requires huge amounts of computing power -- much more than for processing generative AI queries, which is known as inference. AFP has contacted DeepSeek for comment. Nvidia did not respond to a comment request but told The Information it had not seen evidence of this and "such smuggling seems farfetched". Another Chinese AI startup, Zhipu, in January unveiled an image generator that it said had been entirely trained on Huawei chips. And Alibaba said this week it would open a new data centre for AI training and inference in southern China, powered by 10,000 of its own chips and operated by China Telecom. As for DeepSeek, "if they have successfully trained V4 entirely on Huawei silicon, it signals a material shift in the geopolitical tech landscape", Wu said.

DeepSeek V4 model to use Huawei chips? But, can Huawei truly match Nvidia in power and scale

DeepSeek is moving V4 AI model to Huawei chips, reducing reliance on Nvidia hardware. This shift reflects China's AI independence strategy and export restriction pressures. Huawei's Ascend chips promise lower costs and scalable inference performance. However, Nvidia still leads in high-end AI training power globally. Early reports suggest hybrid usage may continue for stability. Big tech firms are already stockpiling Huawei chips rapidly. The real question now, can Huawei close the performance gap faster than expected? The DeepSeek V4 model to use Huawei chips is emerging as one of the most significant shifts in the global AI race, with reports suggesting that hundreds of thousands of Huawei AI chips have already been ordered ahead of launch. According to multiple sources, the Hangzhou-based AI startup DeepSeek is preparing to roll out its next-generation model within weeks, marking a decisive move toward domestic semiconductor independence. The decision answers a key question dominating search trends: why is DeepSeek moving away from US chipmakers? The answer lies in geopolitics, supply constraints, and strategic control over AI infrastructure. This development places Huawei Technologies at the center of China's AI ambitions, while sidelining US giants like Nvidia and AMD. With AI models becoming increasingly dependent on hardware optimization, this move could reshape the competitive balance between the US and China in artificial intelligence. The decision behind the DeepSeek V4 model to use Huawei chips is not just technical -- it's deeply strategic. Traditionally, AI companies collaborate with multiple chipmakers to optimize performance before launching new models. However, DeepSeek has reportedly denied early access to Nvidia and AMD, breaking industry norms. Instead, it has worked closely with Huawei and Cambricon Technologies to rewrite core parts of its model. This ensures that V4 runs efficiently on domestically produced chips, reducing reliance on foreign hardware. This shift comes at a time when US export controls -- strongly backed by Donald Trump -- have restricted China's access to cutting-edge chips like Nvidia's Blackwell architecture. By building a model optimized for Huawei chips, DeepSeek is effectively future-proofing itself against further restrictions. Reports indicate that Chinese tech giants, including Alibaba Group, ByteDance, and Tencent Holdings, have placed bulk orders totaling hundreds of thousands of Huawei chips. This surge in demand signals strong industry confidence in Huawei's AI hardware capabilities. It also reflects a broader shift where Chinese firms are aligning their infrastructure with domestically produced technology. The DeepSeek V4 model to use Huawei chips is expected to benefit directly from this scale. Larger chip availability means faster deployment, improved training cycles, and reduced dependency on global supply chains that are increasingly volatile due to geopolitical tensions. The implications of the DeepSeek V4 model to use Huawei chips extend far beyond China. Nvidia, long considered the dominant force in AI hardware, could face a gradual erosion of its market share in China. Despite claims by Nvidia CEO Jensen Huang that China remains only "nanoseconds behind" the US in AI development, this move suggests that the gap may close faster than expected. If Chinese firms successfully optimize their models for local chips, they may no longer need US hardware at scale. Meanwhile, AMD also faces a similar challenge, as both companies lose early access opportunities that are critical for fine-tuning chip performance with next-gen AI models. The situation is further complicated by reports that DeepSeek may have previously trained models using advanced Nvidia chips despite export restrictions. This adds another layer of tension to an already complex US-China tech rivalry. The DeepSeek V4 model to use Huawei chips could mark the beginning of a new phase in China's AI ecosystem -- one that is more self-reliant and less vulnerable to external pressure. DeepSeek is not stopping at a single model. Reports suggest that it is developing two additional V4 variants, each optimized for different capabilities and designed specifically for Chinese hardware. This multi-model strategy could give China a diversified AI portfolio capable of competing globally. Moreover, the success of earlier models like V3 and R1 -- known for their lower cost and competitive performance -- has already shaken investor confidence in high-spending US AI firms. V4 is expected to build on that momentum. 1. What is the DeepSeek V4 model to use Huawei chips and why is it important? The DeepSeek V4 model to use Huawei chips represents a major shift toward China's self-reliant AI ecosystem, reducing dependence on US chipmakers like Nvidia and AMD. This move is driven by export restrictions and the need for stable, domestic hardware supply. It is important because it could reshape the global AI competition and accelerate China's semiconductor innovation. 2. How will the DeepSeek V4 model to use Huawei chips impact the global AI chip market? The DeepSeek V4 model to use Huawei chips could significantly impact the global AI chip market by weakening Nvidia and AMD's dominance in China. As Chinese firms adopt Huawei-based infrastructure, demand for US chips may decline in the region. This shift may also push other countries to explore localized AI hardware strategies, intensifying global competition. (You can now subscribe to our Economic Times WhatsApp channel)

DeepSeek To Run New Model on Huawei Chips; Concern for US Chipmakers?

What's more, DeepSeek did not provide early versions of their upcoming flagship model to Nvidia and AMD for performance optimisation In a move that is likely to cause some concern amongst top global chipmakers, Chinese AI company DeepSeek is planning to run its new foundational model on the latest semiconductors designed and manufactured by Huawei Technologies. A report published by The Information said the Chinese tech giants led by the likes of Alibaba, ByteDance and Tencent Holdings, were all set to launch the new model had also placed larger orders for Huawei's new chip. How much of a bearing this could have on the likes of Nvidia and others is something we will have to wait and see. That DeepSeek would ignore US chipmakers was highlighted by a news report carried by wire agency Reuters last month itself. They had said that the Chinese AI company had not shared their flagship model for performance optimisation. Instead, they gave access to Huawei Technologies. Readers would recall that developers usually share pre-release versions of their major models with large chipmakers such as Nvidia and AMD in order to ensure that their software functions efficiently and across widely used hardware. In the past, DeepSeek had worked closely with Nvidia, but post President Trump's tariff scare on China and a ban on selling chips there, the relationship seems to have soured. From what we see in the Chinese media, the next foundational model from DeepSeek could be out over the next few weeks. Over the past few months, the company has been in collaboration with Huawei and its rival Cambricon Technologies to rewrite pieces of the model's underlying code as well as for testing the same. The report also noted that DeepSeek had two additional variants of its V4 model and both were optimised and built to run on Chinese-made chips. What remains to be seen is whether this release would mimic a similar instance (V3 and R1 release) last year when it caused a tech stock selloff on one crucial question: Do AI firms in the US need to spend billions of dollars for computing power? The last time round, Deepseek's release caused a major upheaval in the AI-heavy Nasdaq with the composite index falling 3%. Shares of Nvidia plummeted 17% during that period and wiped out a whopping $600 billion in a flash. Though both have since recovered, the release of the V4 flagship model could potentially trigger a sequel to that meltdown. What's more, if such an event does take place and DeepSeek is perceived as competent enough to compete with the current models of OpenAI and Anthropic, things could get really tense. Because these two AI startups are targeting an IPO in 2026 and other Big Tech giants like Google, Meta and Microsoft have invested hundreds of billions of dollars around AI infrastructure in 2025 and plan to shell out over $650 billion in 2026.

Waiting for DeepSeek: new model to test China's AI ambitions - VnExpress International

For weeks now, the global tech industry has been waiting for a major artificial intelligence launch from DeepSeek, seen as a benchmark for China's progress in the fast-moving field. More than a year has passed since the startup put Chinese AI on the map in early 2025 with a low-cost chatbot that performed at a similar level to U.S. rivals. But despite reports and rumors about its imminent release, DeepSeek's next-generation "V4" model is nowhere in sight. Speculation is also swirling over the geopolitical implications of which computer chips were chosen to train and power the new system: world-leading U.S. designs or made-in-China alternatives that the country is racing to develop. "It's important to know because at one level, it is a signal of China's AI self-sufficiency trajectory," Wei Sun, principal AI analyst at Counterpoint Research, told AFP. Tech news outlet The Information reported last week that V4 can be run on the latest chips made by China's Huawei. Such a shift would mark a milestone for China in its bid to beat U.S. restrictions on the export of top-of-the-range AI chips from Californian titan Nvidia to the country. The report cited five people with direct knowledge of large orders for Huawei chips, made in preparation for the DeepSeek launch by tech giants including Alibaba, ByteDance and Tencent. AFP contacted DeepSeek, Huawei, Alibaba, ByteDance and Tencent but none were able to comment. 'Wake-up call' DeepSeek started life in 2023 as a side project of a hedge fund that had access to a cache of powerful Nvidia processors. It shot to attention in January 2025 with its R1 deep-reasoning chatbot, which sent U.S. tech shares tumbling with President Donald Trump calling it a "wake-up call" for American firms. R1 was 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. Stephen Wu, founder of the Carthage Capital fund, told AFP that V4 -- said to be multimodal, meaning it can generate text, pictures and video -- could again shock U.S. tech valuations. "I expect the upcoming DeepSeek V4 release will not just be a software update; it will be a highly capable, open-source model that handles massive context windows at a fraction of the cost," he predicted. But DeepSeek's reputation as a company at the frontier of AI technology is also at stake. Its models previously relied on Nvidia chips, so a move to collaborate with domestic chipmakers would require "substantial re-engineering", Wei said. "That transition can slow development cycles and introduce performance trade-offs, especially for V4, a model expected to be state-of-the-art." Training vs inference The U.S. cites national security concerns as the reason for its export ban on Nvidia's most powerful AI processors to China. "The ongoing wait for DeepSeek V4 points to friction in scaling advanced models without unrestricted access to top-tier Nvidia hardware," Wu said. But some reports allege that DeepSeek skirted the ban to train V4 using thousands of Nvidia's top-end Blackwell chips, dismantled in third countries and smuggled to China. Training AI models requires huge amounts of computing power -- much more than for processing generative AI queries, which is known as inference. AFP has contacted DeepSeek for comment. Nvidia did not respond to a comment request but told The Information it had not seen evidence of this and "such smuggling seems farfetched". Another Chinese AI startup, Zhipu, in January unveiled an image generator that it said had been entirely trained on Huawei chips. And Alibaba said this week it would open a new data centre for AI training and inference in southern China, powered by 10,000 of its own chips and operated by China Telecom. As for DeepSeek, "if they have successfully trained V4 entirely on Huawei silicon, it signals a material shift in the geopolitical tech landscape", Wu said.

DeepSeek's V4 model will run on Huawei chips, The Information reports

April 3 (Reuters) - China's DeepSeek's new model called V4 will run on the latest chips designed by Huawei Technologies, U.S. digital news outlet The Information reported on Friday. In preparation for V4's launch, Chinese tech giants, including Alibaba Group, ByteDance and Tencent Holdings, have placed bulk orders for Huawei's upcoming chip totaling hundreds of thousands of units, the report said, citing five people with direct knowledge of the purchase. The next-generation model will likely be launched in the next few weeks, the report said. Huawei Technologies and DeepSeek did not immediately respond to Reuters requests for comment sent outside normal office hours. DeepSeek has spent the past few months working directly with Huawei and another Chinese chip designer, Cambricon Technologies, to help rewrite pieces of the model's underlying code, and in testing, the report said, citing two people close to the company. DeepSeek is also working on two additional V4 variants, each optimized for different capabilities and built to run on Chinese chips, the report said. Reuters had earlier this year reported that DeepSeek has not shown U.S. chipmakers its upcoming flagship model for performance optimization, breaking from standard industry practice ahead of a major model update. The lab instead granted early access to domestic suppliers, including Huawei Technologies. The release of DeepSeek's low-cost models V3 and R1 triggered a global tech stock selloff last year, causing investors to question whether U.S. AI firms needed to spend billions of dollars on AI computing power. Since then, there has been a great deal of interest in DeepSeek-V4, a next-generation model that has yet to be released. (Reporting by Sneha S K in Bengaluru; editing by Barbara Lewis and Arun Koyyur)