The Global AI Race: A Five-Layer Cake

Over the past few years, dominance in Artificial Intelligence has rapidly become a primary focus of the most competitive nations across the globe. Governments are investing hundreds of billions of dollars, tech giants are building data centers the size of cities, and companies are churning out more and more AI models. The competition between countries to develop and utilize advanced Artificial Intelligence technologies is now referred to as the “AI race.”

By a wide margin, the two major players in the AI race are currently the United States and China. Of course, this raises the question central to the future of the world: who will dominate the AI era?

While the competition is often framed as a race for the best AI models, that view is far too narrow. Training models, such as OpenAI’s GPT-5 or Google’s Gemini 3, are only the visible tip of a much larger technological iceberg.

NVIDIA CEO Jensen Huang describes the AI ecosystem as a “five-layer cake,” a simple yet effective framework to summarize the industry. At the bottom sits energy, followed by semiconductor chips, cloud infrastructure, AI models, and finally AI applications.

Each layer builds on the ones beneath it. And when we view the U.S. - China competition through this lens, a new idea appears: the struggle for AI dominance is not simply one race… it’s built upon five different races happening simultaneously!

First Layer: Energy

Energy is the invisible foundation of AI: without power, nothing else in the “five-layer cake” works.

This is because all other layers of the cake require enormous amounts of electricity, particularly AI infrastructure. For example, Amazon’s Project Rainier (activated in late 2025) built a massive data center in Indiana with a power capacity of over 2.2 gigawatts - enough power for around 1.6 million homes! As such, the global demand for power is only accelerating: some projections estimate that AI could double data center electricity consumption by the end of the decade.

In this foundational layer, China holds a massive advantage.

China has spent decades constructing massive energy infrastructure. The country leads the world in renewable energy, nuclear construction, and power transmission networks. Furthermore, the speed at which China builds infrastructure is astonishing. As Jensen Huang noted, “If you want to build a data center in the United States, from breaking ground to standing up … is probably about 3 years. [China] can build a hospital in a weekend. That’s a real challenge.”

The Chinese Communist Party’s centralized rule allows for coordinated efforts to grow energy infrastructure. Meanwhile, the United States faces slower energy development due to more complex regulatory processes and decentralized decision-making.

Each year, China produces over twice the amount of electricity as the U.S. To compensate, American companies are taking matters into their own hands. Tech giants are investing in:

• Nuclear energy partnerships

• Renewable power purchase agreements

• Dedicated energy projects near data centers

Energy is perhaps the most underrated layer of the AI race, and the area where the U.S. must invest more aggressively to power the rest of the AI ecosystem and compete with China.

Second Layer: Semiconductor Chips

Chips, also called microchips or integrated circuits, are tiny electronic parts made from semiconductor materials. If energy is the fuel of AI, then chips are the engine. Training modern AI systems requires special processors known as GPUs (Graphics Processing Units). These chips perform the massive calculations needed for machine learning.

The center of gravity in the chip industry is NVIDIA, an American company that produces the most powerful AI chips in the world. Other major players in the field include American companies AMD and Broadcom. (Fun fact: NVIDIA CEO Jensen Huang and AMD CEO Lisa Su are distant cousins! I guess it runs in the family.)

Building these chips requires a deep global supply chain, though. While the aforementioned American companies focus on designing and selling the chips, they all rely on TSMC, the Taiwanese semiconductor giant responsible for the most advanced wafer manufacturing in the world. Additionally, the Dutch company ASML produces the world’s only machines capable of extreme ultraviolet (EUV) lithography - a technology required to manufacture the most advanced chip designs.

Because these machines and chips are so critical for the AI race, the United States and its allies have placed severe export controls restricting China’s access to cutting-edge manufacturing technology.

However, Chinese companies like Huawei are developing domestic AI chips, and the country is investing heavily in building its own semiconductor ecosystem - including attempts to develop alternatives to ASML’s EUV technology.

Also, Taiwan is a major wildcard in current affairs. If China were to invade Taiwan, and therefore gain control of TSMC, the balance of power in the AI race would shift dramatically towards China. Many analysts believe such a scenario would be catastrophic for the global semiconductor supply chain.

For now, however, this second layer of the cake is dominated by the United States and represents its most significant advantage over China in the AI race.

Third Layer: Cloud Infrastructure

AI chips need somewhere to live! That’s where “cloud infrastructure” comes in.

This layer includes the massive data centers that host GPU clusters and computing platforms used to train and run AI models. Building them requires huge capital investments and technical expertise. Here, the United States again has a major advantage.

The global cloud ecosystem is dominated by American hyperscalers (large-scale cloud service providers that offer extensive computing and storage through data centers):

• Amazon (AWS)

• Microsoft (Azure)

• Google (Cloud)

• Meta

These companies and others are collectively investing hundreds of billions of dollars in AI infrastructure. For example, Project Stargate was announced in early 2025 by companies OpenAI, Oracle, and SoftBank as a 500 billion dollar initiative to build data centers across the U.S.

China’s major cloud players such as Alibaba, Tencent, and Huawei operate on a much smaller scale. However, China retains their key advantage, just like in energy infrastructure: speed of construction. The implication is clear; while the United States currently leads in cloud infrastructure, China’s ability to rapidly deploy massive projects could allow it to start closing the gap.

Still, the U.S. currently boasts over 5400 data centers, the most in any country by far. China lags behind with less than 500. Similarly, the U.S. controls around 75% of the world’s AI compute power, while China holds second place with only around 15%. The large gaps in AI infrastructure between the U.S. and China, as well as American companies’ interest in investing further, suggest that the U.S. will likely keep its lead in this layer of the AI cake for the time being.

Fourth Layer: AI Models

AI models are the software systems that actually perform tasks: writing text, generating images, analyzing data, or controlling robots. This is the layer most people default to when they think about “AI.”

In terms of raw capability, the United States currently leads, though not by a huge margin. The world’s most advanced AI labs are mostly American:

• OpenAI

• Anthropic

• Google DeepMind

• xAI

These organizations produce the frontier models that push the boundaries of AI, and their capabilities are accepted worldwide as the best of the best.

But of course, China has quickly developed its own competitive systems. Major Chinese tech companies (including Alibaba, Baidu, ByteDance, and DeepSeek) have launched increasingly capable AI models. In some benchmarks, such as competitive mathematics, the performance gap between Chinese and American models has shrunk dramatically.

China also takes a different strategic approach: while American companies often keep their most advanced models closed-source due to competition concerns, many Chinese firms emphasize open-source and free-to-use models. China focuses on cheaper technology that “gets the job done” efficiently. This strategy is nothing to be sneezed at. For many real-world applications, people don’t need the most advanced AI model… They need one that is cheap and efficient. In many cases, Chinese models are becoming “good enough” at far lower cost.

So, while the United States leads the frontier today, China’s focus on efficiency and accessibility could prove powerful over time. The U.S. should prioritize open-source models to enhance collaboration and innovation if it wishes to keep its competitive edge in this layer of the AI cake.

Fifth Layer: AI Applications

The final layer of the cake is where AI actually touches the real world.

Potential applications are widespread throughout many domains:

• Robotics

• Military systems

• Logistics optimization

• Healthcare diagnostics

• Financial analysis

• Legal research

…and so much more. In this layer, the winner is far less clear.

The United States dominates in consumer-facing AI software. Tools like ChatGPT and AI coding assistants have largely been developed by American companies. The U.S. also leads in entrepreneurship and innovation. It hosts the vast majority of the world’s AI unicorns (startup companies valued at over $1 billion USD).

China, however, excels at large-scale technology deployment. Because regulatory barriers are lower and the government can coordinate major initiatives, China can integrate new technologies into society very quickly. AI surveillance systems, industrial robotics, smart cities, and autonomous transportation projects are already being deployed at large scale.

In other words, the United States may lead in inventing AI applications while China may lead in diffusing them into society. This fifth layer of the cake will become the most vital part of the AI race in the coming years, but AI applications require the support of the four layers underneath.

Who is Winning the AI Race?

The answer depends on which layer you examine. As of early 2026, there is no clear winner, although the United States maintains an overall advantage. Right now, the U.S. appears to lead most of the AI stack: particularly in chips, cloud infrastructure, frontier models, and consumer-based software. China, however, holds major advantages in energy infrastructure and large-scale technological deployment. It is also aggressively trying to close gaps across the other layers.

Both nations are pouring money into the AI race, and several future outcomes are possible.

The United States could maintain its lead, driven by innovation and entrepreneurship. Or maybe China could overtake it by scaling infrastructure and deploying applications faster. Perhaps the world could split into two parallel AI ecosystems, with competing standards, technologies, and alliances, leading to an Artificial Intelligence Cold War.

Why does this matter? Because the stakes are enormous! At the very least, AI leadership will shape:

• Global economic dominance

• Military power

• Technological standards

• Ethical standards

• Political influence

The country with the biggest piece of the five-layer AI cake will control the next era of technological civilization. And the race has only just begun.

Sources

1. https://www.wsj.com/tech/ai/nvidia-ceo-says-ai-needs-more-investment-in-defiance-of-bubble-fears-9dabba63

2. https://www.weforum.org/stories/2026/01/nvidia-ceo-jensen-huang-on-the-future-of-ai/

3. https://fortune.com/2025/12/06/nvidia-ceo-jensen-huang-ai-race-china-data-centers-construct-us/

4. https://lunabase.ai/blog/the-5-levels-of-ai-complete-guide-to-the-7-trillion-ai-ecosystem-2025

5. https://time.com/7358519/ai-china-us-race-graphs/

6. https://time.com/7379419/ai-race-us-china/

7. https://www.whitehouse.gov/articles/2025/07/white-house-unveils-americas-ai-action-plan/

8. https://www.aboutamazon.com/news/aws/aws-project-rainier-ai-trainium-chips-compute-cluster