Huawei Technologies is preparing to mass-ship a pair of advanced artificial intelligence chips – the Ascend 910C and upcoming Ascend 920 – marking a big moment in the global AI hardware arena. These new chips are poised to fill a void left by U.S. export restrictions that have curbed China’s access to top-tier AI accelerators from U.S. firms like Nvidia.
Huawei’s move not only underscores China’s determination to forge ahead in semiconductor self-reliance, but also foreshadows a potential reordering of the global AI supply chain. In a climate of U.S.-China tech tensions, the company’s chip ambitions are set to reverberate far beyond its domestic market, hinting at an emerging bifurcation in the world’s AI development ecosystems.
Huawei’s Ascend Chips Aim to Fill the Nvidia Void
According to sources cited by Reuters, Huawei will begin mass shipments of its Ascend 910C AI chip to Chinese customers as early as May. Initial deliveries have reportedly already occurred, signaling Huawei’s readiness to step into the breach created by U.S. bans on Nvidia’s high-end GPUs.
The 910C is a cutting-edge AI processor designed to match the performance of Nvidia’s flagship H100 accelerator by an inventive means: it packages two of Huawei’s previous-generation 910B chips into one module. This chiplet integration effectively doubles the computing power and memory, yielding performance comparable to Nvidia’s H100, which has been barred from China since 2022
The Ascend 910C is not an entirely new architecture but rather an evolutionary upgrade, leveraging Huawei’s proprietary Da Vinci architecture. With roughly 780–800 TFLOPS of AI performance (in BF16/FP16 precision), it achieves about 60% of the Nvidia H100’s performance on certain tasks – a significant feat given China’s current manufacturing constraints.
The chip supports mainstream AI frameworks (like TensorFlow and PyTorch) in addition to Huawei’s own MindSpore, making it relatively practical for Chinese companies to adopt. By offering a domestic alternative with high performance, Huawei is effectively filling the gap left by Nvidia’s absence. The timing is opportune: just weeks ago the U.S. government tightened export rules, blocking Nvidia’s China-only “H20” AI chips without a license. With Nvidia’s advanced silicon suddenly off-limits, Huawei’s 910C arrives as a lifeline for China’s tech industry – one developed on its own terms. This move ensures that China’s AI labs and data centers can continue training large AI models and deploying advanced analytics, albeit on homegrown hardware. In short, Huawei’s new chip shipments signal that Chinese firms won’t be left stranded by geopolitics; instead, they are pivoting to domestic solutions to keep their AI ambitions on track.
Ascend 910 (Huawei)
U.S. Sanctions Spur China’s Tech Self-Reliance
Huawei’s push into high-end AI chips is occurring against the backdrop of an intensifying tech “cold war” between the United States and China. Washington has imposed successive rounds of export controls to limit China’s access to cutting-edge semiconductors, citing national security concerns.
These include the late-2022 ban on Nvidia’s A100/H100 GPUs for China, extended in 2023/24 to cover even pared-down versions (like Nvidia’s A800, H800, and the H20), as well as similar curbs on advanced processors from AMD (MI300 series). The intent is to hamper China’s progress in AI and supercomputing, but an unintended consequence is becoming apparent: the restrictions are galvanizing China’s drive for semiconductor self-sufficiency. Beijing has poured massive investments into its chip sector (including a state-backed $47.5 billion “Big Fund” for semiconductors), and companies like Huawei are the tip of the spear for these efforts.
Building world-class AI chips under sanctions is no easy task. Huawei must navigate around a U.S. technology blockade that cut off its access to top silicon fabrication and IP. The Ascend 910C provides a case study in resourcefulness. Part of the chip is reportedly fabricated by China’s leading foundry, SMIC, on a 7-nanometer process. In addition, Huawei has had to get creative in sourcing components: some 910C units may incorporate chips originally made by TSMC for a third-party (Sophgo) that were acquired through intermediaries. U.S. regulators are reportedly investigating such workarounds, underscoring how closely Washington is watching for any sanction evasion.
Huawei denies using illicit parts, and TSMC asserts it no longer directly supplies Huawei. Meanwhile, crucial memory like HBM (high-bandwidth memory) for these AI boards may also be procured via middlemen, given that leading memory makers are also subject to U.S. pressure. All of this illustrates the complex cat-and-mouse dynamic at play: China’s tech giants are forced to innovate and improvise to overcome barriers, and in doing so they are gradually chipping away at the country’s reliance on Western technology.
Far from halting China’s AI development, the pressure from sanctions appears to be accelerating it. In the absence of U.S. chips, a cadre of Chinese companies is rushing to fill the void. Huawei’s Ascend series is joined by a growing lineup of domestic AI chips from players like Baidu (Kunlun chips), Alibaba (T-Head division), startup Biren Technology, and others. Even relatively young firms are now entering a market long dominated by Nvidia.
This surge of innovation suggests that China is determined to control its own destiny in the AI age. Chinese authorities have even informally advised local tech companies to prioritize domestic chips over foreign alternatives, ensuring a built-in customer base for made-in-China silicon. The immediate payoff of this strategy is continuity – Chinese companies can keep training AI models without interruption. The longer-term payoff could be a robust, homegrown semiconductor ecosystem that is far less vulnerable to external shocks. In essence, the U.S.-China tech rivalry has entered a new phase: one where export controls and technology bans are met with an equal and opposite force of domestic innovation. Huawei’s new chips are a tangible result of that dynamic.
Nvidia’s Market Dominance Faces a New Challenge
For years, Nvidia has enjoyed an almost unassailable lead in the AI chip market worldwide, with its GPUs serving as the workhorses for machine learning in both industry and research. That dominance has translated into booming business – until now. With the Chinese market effectively fenced off by U.S. policy, Nvidia is bracing for the financial fallout.
In the wake of the latest restrictions, Nvidia’s stock took a noticeable hit (dropping nearly 7% on the news) amid investor fears of lost sales. The company even warned it may have to write off up to $5.5 billion in inventory built for China that can no longer be sold freely. Analysts have estimated that if the U.S. continues to tighten chip exports, Nvidia could eventually forfeit tens of billions of dollars in potential revenue from the China market. For a company that in 2024 briefly reached a $1 trillion market capitalization on the back of AI enthusiasm, losing access to one of the world’s biggest tech markets is a serious setback.
Huawei’s emergence as a viable GPU competitor thus poses a twofold challenge to Nvidia. First, it threatens to erode Nvidia’s share in China, the second-largest economy, which had been a key source of growth. Chinese tech giants and cloud providers that once bought Nvidia chips by the thousands are now strongly incentivized – by necessity and policy – to switch to domestic alternatives. This cost advantage, combined with geopolitical tailwinds, means Nvidia could see a significant portion of its Chinese customer base migrate to homegrown chips.
Second, a successful rollout of Huawei’s AI chips could eventually inspire confidence (and capital) in other markets for non-Nvidia solutions. While Western companies are unlikely to replace Nvidia hardware with Chinese chips anytime soon due to trade restrictions and security concerns, the mere existence of a credible alternative underscores that Nvidia’s technological lead is not insurmountable.
That said, Nvidia’s global dominance is not toppling overnight. The company’s GPUs still set the gold standard for AI performance and have a deeply entrenched software ecosystem that Huawei and others must compete with. Outside of China, Nvidia remains the default choice for AI infrastructure, and even within China, Nvidia’s prior generations (like GPUs equivalent to the A100) are still in use where available. Huawei’s 910C, impressive as it is, operates at perhaps ~60–70% of the performance of Nvidia’s latest flagship in many scenarios. Moreover, Huawei will need to demonstrate that it can manufacture these chips in volume and support them with software and developer communities.
Nvidia’s market position in the West is secure for now, bolstered by exploding AI demand globally (from Silicon Valley to Europe to India) that far exceeds supply. The true test will be whether Huawei’s next chip generation can narrow the gap further. If Huawei can deliver on that promise, it will cement the company’s role as a serious long-term rival in AI silicon, at least within its sphere of influence.
(Unite AI/Alex McFarland)
Toward a Bifurcated AI Ecosystem?
Huawei’s latest moves highlight a broader trend: the potential bifurcation of the global AI ecosystem into two parallel tracks. On one side, the U.S. and its allies continue to advance with chips from companies like Nvidia, AMD, and Intel, along with specialized AI accelerators from Google (TPUs) and others. On the other side, China is rapidly building its own stack of AI hardware and software – from chips like the Ascend series to frameworks like MindSpore – largely incompatible with or isolated from Western supply chains. If this trend continues, we could witness a world where AI development in China is built on Chinese processors running in Chinese data centers, while the rest of the world runs on Western chips.
Beijing’s encouragement for firms to use domestic tech and Washington’s bans on chip exports are together driving this wedge deeper. The global AI race, in effect, may splinter into separate lanes: each side racing with its own technology, rules, and standards.
Such a divide carries profound implications. In the near term, China’s pivot to self-reliant AI hardware ensures it can pursue cutting-edge AI research (from large language models to advanced computer vision) without begging Silicon Valley for tools. This is vital for China’s aspirations to lead in AI by 2030 – a goal enshrined in its national strategy.
In the longer term, however, a decoupling of AI ecosystems could lead to reduced interoperability and knowledge exchange between East and West. Today, a machine learning model developed in one country can often be shared and run elsewhere, assuming the hardware is available; tomorrow’s bifurcated landscape might complicate that flow. For instance, engineers proficient in Nvidia’s software may not easily transition to programming Huawei’s Ascend chips, and vice versa. Companies and researchers may have to specialize for one ecosystem, potentially limiting collaboration.
On the flip side, competition between two AI superpowers can spur innovation: each side will be driven to outdo the other, possibly accelerating advancements in chip design and AI capabilities at a blistering pace. We might see divergent approaches to AI computing emerge – perhaps novel architectures or optimizations in China that differ from those in the West – enriching the global innovation pool, but also creating technical barriers between the two spheres.
For the global supply chain, this split means adaptation. Manufacturers, cloud service providers, and even smaller nations will face choices about which ecosystem to align with, or how to bridge both. It could lead to duplicate investment in parallel infrastructures – costly, but seen as necessary for strategic autonomy. Countries in Europe or Asia-Pacific not directly involved in the U.S.-China standoff may try to stay neutral or support standards that allow some interoperability, but they too may eventually lean one way or another for critical technologies.
In essence, Huawei’s new AI chips are a strategic statement. They signal that the balance of power in AI computing is beginning to shift, however gradually, and that we are entering an era where technological power is more distributed. The coming years will reveal whether this marks the start of a truly divided tech world or simply a more competitive one. Either way, Huawei’s Ascend chips have ensured that the global AI race will not be a one-horse race – and that geopolitics will remain intertwined with who leads in AI. The world will be watching as these chips roll out, for what they mean not only for China’s tech trajectory but for the future shape of AI innovation everywhere.