Despite years of US trade restrictions, Chinese telecom giant Huawei has quietly emerged as one of China’s strongest contenders in the field of artificial intelligence.
Based in Shenzhen, the company is not only seen as China’s answer to US AI chip leader Nvidia but was also among the first to monetize AI models through industrial applications.
“Huawei has been forced to shift and broaden its core focus over the past decade due to multiple external pressures,” said Paul Triolo, Partner and Senior VP for China at consultancy DGA-Albright Stonebridge Group.
This expansion has led the company into diverse areas—from smart cars and operating systems to the critical technologies driving the AI revolution, including advanced semiconductors, data centers, chips, and large language models.
“No other tech company has demonstrated this level of competency across so many complex and high-barrier sectors,” Triolo added.
Even Nvidia CEO Jensen Huang recently acknowledged Huawei’s progress, describing it as “one of the most powerful tech companies in the world,” and warning that Huawei could replace Nvidia in the Chinese market if Washington continues its export restrictions on advanced chips.
Nvidia’s market capitalization surpassed $4 trillion last week, making it the most valuable company globally, thanks to its cutting-edge processors and its CUDA computing platform, the industry standard for training AI models.
But that dominance is now being challenged, as Huawei proves capable of delivering high-quality performance across a broad spectrum.
From Small Distributor to National Hero
Founded in 1987 by ambitious entrepreneur Ren Zhengfei, Huawei started as a small distributor of telephone switches from a Shenzhen apartment.
Over time, it grew into a major player in the telecom space, expanding first into emerging markets such as Africa, the Middle East, Russia, and Latin America before breaking into Europe.
By 2019, Huawei was well-positioned to benefit from the global launch of 5G networks and had become one of the top smartphone manufacturers. It also began designing its own chips through its HiSilicon unit.
But its success attracted scrutiny—particularly from the United States, which repeatedly accused Huawei of threatening national security, allegations the company has consistently denied.
In 2019, the US dealt Huawei a heavy blow by placing it on a trade blacklist, barring American firms from doing business with it.
This slashed revenue from its consumer unit—then its largest—to about $34 billion in 2021, half of the previous year’s total.
Still, the company pressed on with AI chip development, even after further sanctions in 2020 severed its ties with Taiwan’s TSMC.
In 2019, Huawei launched its Ascend 910 AI processor as part of a strategy to build a comprehensive, end-to-end AI ecosystem.
While the sanctions were intended to cripple Huawei, they instead helped turn it into a national symbol, especially after the 2018 arrest of CFO and founder’s daughter Meng Wanzhou in Canada over alleged sanctions violations related to Iran.
“The US restrictions pushed Huawei into the arms of the Chinese government in a way Ren Zhengfei had long tried to avoid,” said Triolo, “but they ultimately accelerated the company’s AI progress.”
The Comeback
In 2023, Huawei’s consumer business saw a strong rebound, thanks to the launch of a smartphone featuring what analysts say is an advanced, China-made chip.
That surprise chip was indirectly linked to Chinese firm SMIC, which is also under US sanctions.
Though semiconductor analysts believe the chip’s production capacity is limited, Huawei proved it was back in the game.
Reports soon emerged about a new AI chip, the Ascend 910B, and the company is now preparing for mass production of the next-gen 910C.
“Huawei has made significant progress in emulating high-performance GPU capabilities using clusters of less advanced chips,” said Jeffrey Towson, Managing Partner at TechMoat Consulting.
In April, Huawei unveiled its CloudMatrix 384 system, which connects 384 Ascend 910C chips in a single data center cluster.
Analysts noted that the system outperforms Nvidia’s GB200 NVL72 in some benchmarks.
“Huawei isn’t just catching up—it’s redefining how AI infrastructure should work,” said a recent Forrester report.
The company also developed its own software stack to replace CUDA, called CANN, boosting its in-house software capabilities.
“The AI race isn’t just about faster chips—it’s also about development tools, and Huawei is building a full-stack solution,” the report added.
But Huawei’s ambitions go beyond just challenging Nvidia. Chips are only one building block in its broader AI strategy.
Today, Huawei’s operations span the full AI value chain—from chips and computing power to models and applications.
In 2023, its ICT infrastructure unit—which includes 5.5G networks and industrial AI systems—became its largest revenue source, generating 362 billion yuan.
Through its cloud arm, Huawei Cloud (founded in 2017 to compete with AWS and Oracle), the company runs data centers powered by Ascend chips and CloudMatrix systems.
It uses these to train its Pangu series of AI models, which are tailored for sectors like healthcare, finance, government, industry, and automotive. These models have been deployed in over 20 industries in the past year.
“Often, our engineers stay on-site for months—even in remote mines—to implement AI solutions,” said Jack Chen, VP of marketing for Huawei’s oil, gas, and mining unit.
In May, the company deployed over 100 autonomous electric trucks powered by 5G, AI, and Huawei’s cloud infrastructure to transport coal and soil.
Chen added that this technology “can be widely replicated across Central Asia, Latin America, Africa, and the Asia-Pacific region.”
Huawei has also open-sourced its Pangu models in a move to expand its global reach and bolster what it calls its “Ascend ecosystem”—a complete suite of intelligent products built on Ascend chips.
Patrick Moorhead, CEO of Moor Insights & Strategy, told CNBC: “I expect Huawei to push Ascend chips into Belt and Road countries.”
He concluded that within 5 to 10 years, the company could build a significant market share in those nations—just as it did before in telecom.
