New Delhi: For much of the past three years, the global AI race has been seen as a battle over chips. The United States appeared firmly in control, thanks largely to Nvidia, whose advanced processors power everything from ChatGPT-like models to corporate AI systems.
That lead remains intact. Nvidia still dominates the global AI chip market and controls roughly half of the world’s installed AI chip stock, accounting for nearly two-thirds of computing capacity used for AI. China’s Huawei remains its closest challenger but still trails Nvidia in performance and efficiency.
Yet focusing only on chips misses a much bigger story. China has realised that winning the AI race requires more than hardware. It also requires software platforms, developer communities and applications that can create a complete AI ecosystem.
This broader shift is highlighted in the report released by Brussels-based think tank Bruegel yesterday (June 22). It argues that the competition is no longer just about who makes the best chips but about who controls the entire AI technology stack.
China’s strategy combines state subsidies, policy support and a huge domestic market to help local firms scale up despite American export restrictions. The result is that Huawei and other Chinese companies are becoming increasingly competitive, even if they remain technologically behind US rivals.
According to the authors of the report, Alicia García-Herrero and Bertin Martens, “Despite Chinese progress, the United States remains for now ahead in the race for dominance over the so-called AI hardware stack.” However, they also highlighted that Huawei has emerged as “the most credible foreign challenger”.
Software Becomes Crucial
The real challenge to America may come not from chips but from software.
Nvidia’s dominance rests not only on its hardware but also on CUDA, a software platform that allows developers to efficiently build and run AI applications. Over the years, CUDA has become the industry standard, creating a powerful ecosystem that keeps customers tied to Nvidia products.
The Bruegel report notes that “Nvidia’s durable advantage also rests on CUDA,” describing it as the platform that has become “the de-facto standard for AI development.” The authors also point out that CUDA creates powerful network effects by linking chip makers, developers, companies and users into one ecosystem.
These network effects have given Nvidia enormous pricing power, with profit margins on some AI chips estimated at 70% to 80%.
China’s answer is to create its own software ecosystem.
Huawei has developed CANN and MindSpore, software tools designed to compete with CUDA. More importantly, it has made it easier for developers to shift their AI applications from Nvidia systems to Huawei hardware. In 2025, Huawei open-sourced key parts of its software platform and encouraged universities, research institutes and technology firms to contribute to its development.
The report highlights a particularly important innovation, ‘torch_npu’, a backend plugin that lets standard PyTorch code run on Ascend processors.” This means developers can continue using familiar tools while running applications on Huawei chips.
Huawei has also teamed up with DeepSeek, one of China’s fastest-growing AI companies. DeepSeek’s models work on both Nvidia and Huawei hardware, helping Chinese software gain wider acceptance among developers worldwide.
According to the authors, “The software gap, like the hardware gap, is no longer fixed or unbridgeable.” They argue that China has made significant progress in reducing the advantages that once seemed to guarantee US dominance.
Europe Falls Behind
While the US and China are battling for AI leadership, Europe appears to be slipping out of the race.
The continent still possesses important strengths. European companies manufacture critical semiconductor equipment and its universities remain major centres of AI research. Yet Europe lacks a globally important AI chip company and has no software platform comparable to Nvidia’s CUDA or Huawei’s CANN.
The Bruegel report states that “Europe has no AI chip designer of global significance and no software layer analogous to CUDA or CANN.” As a result, Europe risks becoming a supplier to the AI revolution rather than a leader of it.
The authors argue that Europe could focus on specialised areas such as automotive AI, climate modelling and trustworthy AI systems. But they also warn that the value in AI increasingly belongs to those who control both chips and software platforms.
The message for India: Building AI capacity will require more than investing in data centres and semiconductors. The next phase of competition will depend on creating strong software ecosystems, developer communities and home-grown AI platforms.
The global AI race is no longer just about who builds the fastest chips. It is increasingly about who controls the entire technology stack. America still leads, but China is closing the gap faster than many expected. Europe, meanwhile, risks becoming a spectator in one of the defining technology contests of the century.
(Cover photo by BoliviaInteligente on Unsplash)