In the first wave of artificial intelligence, the world competed over models. In the second wave, it competed over data. Now, the contest has moved deeper — into the silicon, cloud infrastructure, developer tools, and national policy frameworks that decide who can build AI at scale.
At the center of this race stand two names: Nvidia, the American company whose GPUs have become the default engine of modern AI, and Huawei, the Chinese technology giant trying to prove that sanctions cannot stop a country determined to build its own stack.
This is no longer just a corporate rivalry. It is a sovereignty race.
Nvidia’s latest financial results show the scale of the global AI buildout. The company reported first-quarter revenue of $81.62 billion, up 85% year over year, driven by extraordinary demand for AI chips and data-center infrastructure. CEO Jensen Huang described the AI infrastructure expansion as one of the largest infrastructure buildouts in history.
“The AI race is no longer about having the smartest algorithm. It is about having guaranteed access to the machines that can train, run, secure, and scale intelligence.”
For governments, that dependency is becoming uncomfortable. A nation that relies entirely on foreign chips, foreign cloud providers, foreign model APIs, and foreign software frameworks may have AI capability — but not AI sovereignty.
That distinction is now reshaping policy from Washington to Beijing, Brussels, New Delhi, and the Gulf.
The Nvidia Advantage: Not Just Chips, But an Entire AI Operating System
Nvidia’s power does not come only from the GPU. It comes from the ecosystem around it.
CUDA, Nvidia’s developer platform, has become deeply embedded in AI research, machine learning frameworks, enterprise deployments, and high-performance computing. For many AI teams, Nvidia hardware is not merely faster; it is familiar, supported, optimized, and surrounded by years of tooling.
That creates a powerful lock-in effect.
Even when alternative chips exist, switching away from Nvidia can mean rewriting software, retraining engineers, rebuilding optimization pipelines, and accepting performance uncertainty. This is why Nvidia remains central to global AI infrastructure even as governments worry about overdependence.
Reuters Breakingviews recently noted that China’s domestic AI chips from Huawei, Cambricon, and others still face a gap against Nvidia, while CUDA remains a major barrier to migration.
“In the AI economy, sovereignty is not achieved by buying servers alone. It requires control over the full stack — silicon, networking, power, cloud, software, models, data, governance, and talent.”
This is why Nvidia is not simply selling processors. It is selling the rails on which much of the AI economy runs.
Huawei’s Countermove: China’s Sanction-Driven AI Stack
Huawei has become the symbol of China’s effort to reduce dependence on American technology.
Years of U.S. restrictions have cut Chinese firms off from the most advanced Nvidia chips. In response, China has accelerated domestic alternatives, with Huawei’s Ascend AI chips becoming central to that strategy. Demand for Huawei’s Ascend chips has reportedly surged as Chinese companies look for local substitutes for Nvidia hardware.
On May 25, 2026, Huawei disclosed a new chip-design strategy called Tau Scaling Law, along with a LogicFolding architecture. The company said this approach focuses on improving system-level efficiency, data movement, and latency rather than depending only on transistor shrinkage. Huawei also said the architecture would feature in future Kirin smartphone chips and Ascend AI chips.
The announcement matters because China still faces a manufacturing gap. Reuters reported that China’s current domestic chipmaking capability is around 7 nanometers, while global leaders such as TSMC are targeting 1.4 nanometers by 2028. Huawei’s message is therefore strategic: if China cannot immediately match the West in chipmaking equipment, it will try to compete through architecture, packaging, system design, and domestic scale.
“Huawei’s challenge is not only to build a chip. It must build confidence that a Chinese AI stack can survive without permission from Washington.”
The pressure is already visible in the market. Reuters reported in April 2026 that major Chinese technology firms were scrambling to secure Huawei AI chips after the launch of DeepSeek V4. Another Reuters report in March 2026 said Huawei’s newer AI chip had tested well with major Chinese customers including ByteDance and Alibaba, which planned orders.
The signal is clear: China does not want to remain an Nvidia-dependent AI economy.
Export Controls Turned Chips Into Geopolitical Assets
The U.S. has used export controls to slow China’s access to the most advanced AI chips. Nvidia responded by designing China-specific products such as the H20, but even those became politically sensitive.
In April 2025, Nvidia said the U.S. government would require a license for H20 exports to China, and the company expected a charge of up to $5.5 billion tied to H20 inventory and commitments. Later in 2025, the U.S. licensed Nvidia to resume some chip exports to China, while also pursuing an unusual arrangement under which Nvidia and AMD would give the U.S. government 15% of revenue from certain China chip sales.
This revealed a new reality: AI chips are no longer treated like normal commercial products. They are strategic assets, subject to national-security calculations, diplomatic bargaining, and industrial policy.
“A GPU shipment today can carry the geopolitical weight once associated with oil tankers, telecom networks, or defense systems.”
China, meanwhile, has pushed back. Chinese state media and regulators raised concerns about Nvidia’s H20 chips, including security concerns, even after U.S. export approvals. That creates a double squeeze for Nvidia: Washington controls what it can sell, while Beijing can influence whether Chinese customers should buy.
Why Every Nation Now Wants Its Own AI Stack
The phrase “sovereign AI” is often misunderstood. It does not mean every country must manufacture every chip domestically. That is unrealistic for most nations.
Instead, sovereign AI means a country wants enough control over the critical layers of AI capability to avoid being strategically helpless.
A national AI stack includes:
Compute: GPUs, accelerators, supercomputers, data centers, energy, cooling, and networking.
Cloud: domestic or trusted cloud infrastructure that can host sensitive workloads.
Data: national datasets, language corpora, public-sector records, industrial data, and privacy rules.
Models: locally trained or adapted foundation models for language, health, education, law, defense, and public services.
Software: frameworks, deployment tools, monitoring systems, security layers, and developer ecosystems.
Governance: laws, auditability, model safety, procurement rules, and national-security safeguards.
India is already moving in this direction. A Press Information Bureau release in February 2026 said the IndiaAI Compute Portal provides access to more than 38,000 GPUs and 1,050 TPUs at subsidised rates under ₹100 per hour. The official IndiaAI Compute Capacity page describes the initiative as public-private AI compute infrastructure intended for academia, startups, researchers, government, public-sector agencies, and approved entities.
Europe is taking a different path through AI Factories. The European Commission says AI Factories use EuroHPC supercomputing capacity to develop trustworthy cutting-edge generative AI models. The European Parliament has described AI factories as combining supercomputers, data, and human capital.
“The country that owns the AI stack owns more than technology. It owns the ability to decide what intelligence is built, where it runs, whose laws govern it, and who can switch it off.”
The Market Is Splitting Into AI Blocs
The global AI economy is beginning to fragment into blocs.
The U.S. bloc is anchored by Nvidia, AMD, Intel, hyperscalers, cloud platforms, venture capital, and model companies such as OpenAI, Anthropic, Google, Meta, and xAI.
China’s bloc is built around Huawei, Alibaba, Tencent, ByteDance, Baidu, SMIC, domestic cloud platforms, and state-backed self-reliance programs.
Europe wants trusted, regulated, industrial AI infrastructure that reduces dependency on American cloud and Chinese hardware.
India wants affordable compute, local-language models, public-sector AI, startup access, and strategic independence without fully disconnecting from global suppliers.
Russia, under sanctions, is increasingly looking to Chinese chips. Reuters reported that Sberbank is seeking Chinese-made chips to power its GigaChat AI model, while competing with Chinese firms for Huawei’s high-end Ascend chips.
This is how the AI map is changing. The world may still use global technology, but governments want national fallback options.
The Real Battle: AI Infrastructure as National Power
The future of AI will not be decided only by model benchmarks. It will be decided by infrastructure capacity.
Who has enough chips?
Who has enough electricity?
Who controls the data centers?
Who owns the software layer?
Who can train frontier models without foreign approval?
Who can run sensitive AI workloads inside national borders?
Who can survive sanctions, shortages, or export bans?
These are now boardroom questions, defense questions, and cabinet-level questions.
Nvidia remains the dominant commercial force. Huawei has become the strongest symbol of China’s technological resistance. India, Europe, and others are trying to build sovereign capacity without isolating themselves from the global AI economy.
“The AI sovereignty race is not a rejection of globalization. It is a recognition that dependence without control is a strategic risk.”
For enterprises, this means future AI procurement will become more complex. Companies may need to choose not only the best model or cheapest cloud provider, but also the jurisdiction, supply-chain exposure, compliance posture, chip availability, and long-term portability of their AI systems.
For nations, the lesson is sharper: AI capability without infrastructure control is borrowed power.
The Bottom Line
Nvidia proved that AI compute can become the most valuable industrial layer of the digital economy. Huawei is proving that restricted access can accelerate national alternatives. Governments are now learning that AI is not just a software revolution — it is an infrastructure race.
The next decade will not be defined only by who builds the most intelligent AI model. It will be defined by who owns the stack beneath it.
Because in the age of sovereign AI, the question is no longer just “Who has the best AI?”
It is:
“Whose chips run it, whose cloud hosts it, whose laws govern it, and whose nation can still operate when the supply chain breaks?”
