AI wars: How US and China compete for tech dominance, and where Ukraine fits in

Key questions:

• How are the US and China waging the chip war?
• Why has data center construction turned into a modern-day gold rush?
• Who leads the race for data collection to train artificial intelligence?
• And why does Ukraine need its own AI?

Artificial intelligence was named the word of the year for 2023 by the authoritative Collins English Dictionary. That was the moment when AI truly went mainstream. Two years later, it has spread into nearly every aspect of our lives — from American ChatGPT to China's DeepSeek.

Artificial intelligence is, above all, a multiplier of effort. Now that most countries have learned to collect vast amounts of data about nearly everything, AI enables them to process and utilize that data to their advantage.

"AI can identify unexpected connections that a human simply couldn't see, just because no one can process such enormous amounts of information. These insights and the ability to apply AI across different areas are changing the rules of the game,” explained Dmytro Dubov, an expert at Ukraine's National Institute for Strategic Studies, in a comment to RBC-Ukraine.

AI is already directly shaping the quality, comfort, and safety of everyday life. It improves healthcare systems, makes public transportation more efficient and safer, and enhances urban infrastructure.

"Soon we'll see AI traders making money on the market. Whoever has the better model will do it faster. AI will also move deeply into medicine — and the country that builds better models will live longer and treat diseases more effectively,” said Volodymyr Kulinich, head of the AI Lab at the Kyiv School of Economics.

So it's no surprise that in the ongoing political and technological rivalry between the world’s leading powers — the US and China — AI may not always be the loudest issue, but it's always there in the background. In fact, it could prove decisive for global dominance in the coming decades.

At the same time, smaller nations — including Ukraine — must still find their place in this new reality. To understand what that could look like, RBC-Ukraine explored what AI actually runs on, what it depends on, and where it's heading next.

Chip war

To develop AI technologies, countries need not only skilled experts but also three key components, and each of them has become a battlefield of global competition, some more intense than others.

The first is microchips. In the context of AI, chips play a crucial role: they don't just perform calculations. They determine how fast AI models can be trained.

When it comes to chip production, both the US and China have their own national champions — companies capable of developing cutting-edge products. American giants Nvidia and AMD control over 80% of the AI chip market, while China's Huawei and ZTE, despite sanctions, are gradually catching up. To support domestic producers, both Washington and Beijing are using every possible tool — from export restrictions to massive state subsidies.

The United States has imposed strict export controls on the supply of chips to China. According to expert Dmytro Dubov, this defined the first half of this year, creating significant challenges for Beijing. But by late summer, the situation began to shift: in August and September, the Chinese government instructed domestic companies to stop buying American chips — including Nvidia's.

One reason is that Huawei and ZTE have learned to produce their own chips. There are no open test results yet, but according to Chinese officials, they are powerful and efficient enough to replace American products.

"As a result, Nvidia has lost a huge share of the market and revenue. It's unclear how the company plans to regain its foothold in China,” Dubov told RBC-Ukraine.

Nvidia office in Taiwan (Photo: Getty Images)

Meanwhile, both China and the US continue to create obstacles for each other in the chip supply chain.

In September, the US added 32 Chinese companies involved in semiconductor production to its trade blacklist. In response, China announced stricter export controls on rare earth metals and related technologies — resources it dominates, supplying about 90% of the global market.

"Each side is leveraging its own strengths. For the US, that's the companies making chips; for China, it's access to the raw materials they're made from,” explained Dubov.

This standoff has left many other countries uneasy, realizing they're dependent on both superpowers. The European Union, while maintaining close ties with the US, is also expanding cooperation with the United Arab Emirates. Meanwhile, India and Japan are launching their own government-subsidized projects to strengthen domestic chip production.

"This battle will likely remain intense for at least the next two years. It's still hard to say who will win — the US or China. I think it will eventually come to some kind of parity," Dubov predicted.

Yet in this race, Ukraine could also play a role. Thanks to an agreement signed this year on the supply of rare earth materials, which are vital for chip manufacturing.

Data center gold rush

For most users, artificial intelligence is just a program on a smartphone or a browser tab on a computer. But behind the scenes, it requires massive physical infrastructure — data centers. These facilities store terabytes of textual data that power services like ChatGPT, Gemini, and other similar platforms.

According to analytics firm Brightlio, as of September 2025, there are around 11,000 data centers worldwide. Their total number hasn't grown dramatically in recent years, but their capacity keeps increasing. The US leads by far, with 5,426 centers, followed by Germany with 529 and the UK with 523. China, with 449 centers, is rapidly catching up, investing heavily to support its technological leap. Despite the war, Ukraine operates 58 data centers, mainly in Kyiv and Lviv.

In 2025, the seven largest US tech companies are expected to invest a record $364 billion in building and upgrading data centers — almost double China's investment. Europe, with its regulatory focus and green energy priorities, lags, though EU countries plan to increase capacity by 20% by year-end.

The global capacity of data centers is projected to grow 15% annually, according to JLL, but even that may not meet the demand driven by AI.

One key challenge is electricity consumption. Data centers need power not only for computing but also for cooling. The Electric Power Research Institute (EPRI) estimates that a single AI query to a model like ChatGPT consumes ten times more energy than a standard Google search.

Today, for every $1 billion invested in data centers, over $600 million is expected to be spent on electricity over the next decade — about 60% of total costs.

This raises a new question: where will the extra electricity come from? In the US, data centers already consume 4% of national electricity, expected to rise to 8–10% by 2030. Existing power grids also need upgrades, as they weren't designed for such loads and are often outdated.

"For AI and data center needs, the focus isn't on expanding fossil fuel extraction but on nuclear energy. Nuclear plants provide a stable, low-cost power source. In the US, there is a push to restart retired plants and build small modular reactors,” said Oksana Ishchuk, executive director of the Strategy XXI Center for Global Studies, in a comment to RBC-Ukraine.

For example, in September 2024, Microsoft and Constellation Energy signed a 20-year agreement to purchase electricity from a restarted reactor at the Three Mile Island nuclear plant to power Microsoft's data centers. In this area, climate goals aren't being compromised.

Data center in Beijing, China (Photo: Getty Images)

American tech companies also locate data centers abroad, where electricity is cheap — in South Africa, Malaysia, Chile, and Indonesia. Cold climates are particularly advantageous, allowing natural cooling, as in Sweden, Norway, Canada, and Ireland.

"The main uncertainty around data centers is predicting how much energy they will need in the next five years. Even respected institutions like the IEA (International Energy Agency - ed.) produce projections as multiple scenarios," added Ishchuk.

This uncertainty slows private investment in power generation. On one hand, demand for data centers grows; on the other, servers used for AI or cryptocurrency mining are being improved to reduce energy consumption using liquid cooling and AI-optimized chips that cut power use by 30–40%. Some data centers may end up incomplete or even closed.

"It's a bit like a gold rush. Not everyone who goes looking for gold finds it. Some data centers will be built, some unfinished, some will launch, some will eventually be fully utilized," said Dmytro Dubov to RBC-Ukraine.

China is also prioritizing nuclear power for data centers.

"Currently, nuclear ranks fourth in China's energy mix after coal, oil, and gas, but nearly 30 new reactors are under construction — almost half of the global total," Ishchuk explained.

By the end of this year, China plans to create eight national computing hubs and ten data center clusters, redistributing computational power from densely populated eastern regions to western provinces with abundant solar and wind energy. ResearchAndMarkets estimates that Chinese data center investments will reach $97.3 billion by 2030.

However, China still faces significant challenges: many data centers remain underutilized.

Overall, the Chinese model contrasts with the US approach. While the US relies on private giants, China combines state control with rapid scaling. According to RAND Corp, this gives China a potential long-term advantage.

Data challenge

The third essential component for AI development is data — the fuel for training models. Right now, it's difficult to say who leads in this area: the US or China. China has a more unified ability to collect data, but many questions remain about its quality. For effective AI training, data quality is more critical than sheer quantity.

"No one knows the quality of the datasets used to train AI in China. You can collect huge amounts of data, but if it isn't properly labeled, it effectively doesn't exist for AI," explained Dmytro Dubov.

According to the expert, China has extensive data for training AI on social behavior and human interactions. A well-known example is the social rating system implemented in several Chinese cities.

By early 2025, the system covered 990 million citizens and over 30 million corporate entities. The Chinese government has access to vast databases — from medical records to genetic information — which can be used to train specialized AI models.

Another advantage for China in data collection is legal flexibility. In the past two years, the US recorded 52 lawsuits related to AI data usage, while China had only five. In China, the interests of the Communist Party can take priority over existing laws, offering more flexibility in handling legal disputes.

China also collects information abroad more aggressively. A striking example is Zhenhua Data, which secretly compiled detailed digital profiles on over 2 million people worldwide, including EU politicians, violating platform terms of service and European law. Despite international pressure, no public sanctions were announced, and the company likely continues to operate.

However, the US has its own strengths. American platforms — Google, YouTube, X, Facebook — dominate the global market, enabling them to collect user data from almost every country. The nearest Chinese competitor is TikTok, but its reach is likely smaller than the combined datasets gathered by American platforms.

"From a technological standpoint, the US probably has more opportunities right now, partly thanks to data-sharing with partner countries. It's not clear who would share data with China," Dubov noted.

Nuclear power plant in California, USA (Photo: Getty Images)

Both countries face common challenges regarding data. One major issue is bias: if datasets aren't representative, AI models can reproduce demographic or social distortions.

Looking ahead, there will also be a shortage of high-quality data. AI models are advancing so rapidly that the industry is approaching a situation where the entire corpus of quality textual content on the internet may be exhausted for training purposes. This will push the use of synthetic data — information generated by AI models themselves. However, the quality of synthetic data remains a separate concern.

The US and China are simultaneously trying to expand their own access to data while limiting their rivals. This creates a situation where information circulates less freely. For the rest of the world — including Ukraine — this means finding a careful balance between the two poles.

Ukraine in the global AI race

Beyond the US and China, few countries have the full capacity to develop artificial intelligence. Most other players lack at least one of the three key components.

The European Union has green-powered data centers and requires that citizen data remain within its territory, but it does not produce its own chips. Japan manufactures its own chips and data centers, but has limited data due to its smaller population. India, conversely, collects vast amounts of data thanks to its population and has data centers, but lacks sufficient chip production.

Against this backdrop, Ukraine is carving out its own role. The country has specialists capable of implementing AI technologies and is negotiating with American companies for access to advanced chips. Energy shortages, however, make large-scale development of domestic data centers a lower priority — though progress is being made where possible.

"We are moving toward becoming the third country in the world in terms of AI implementation in the public sector. Overall, we are moving toward agent-based governance. The Ministry of Digital Transformation is working to provide more services and to test open AI models that already exist in partnership with Google and Meta, retraining them on Ukrainian data," said Volodymyr Kulinich, also secretary of the expert-consultative subcommittee on AI regulation at the Ministry of Digital Transformation.

Currently, Ukraine is developing its own large language model (LLM), which will serve as the foundation for AI-based products such as chatbots, analytical systems, automated document processing tools, and voice assistants.

This project is being carried out by Kyivstar in collaboration with the Ministry of Digital Transformation and the WINWIN AI Center, without government funding. The model is expected to be presented in December 2025. Having a domestic LLM is critically important for national security, unlike situations where Ukraine has relied on imported solutions without developing its own alternatives.

"It's different from using operating systems like Windows, where no sensitive data is transferred, and it's relatively safe. With LLM models and AI technologies, the risks are much higher," Kulinich told RBC-Ukraine.

For example, in energy, foreign AI systems for predicting power grid loads could access confidential data about infrastructure vulnerabilities — a significant risk during a hybrid war.

In healthcare, imported diagnostic AI models may carry biases based on data from other regions, potentially leading to misdiagnoses or even the leakage of citizens' medical information.

***

Artificial intelligence is steadily becoming the backbone of global progress, penetrating every sphere of human life and determining the winners in a new era. Its role will only grow, reshaping economies, societies, and even international politics. At the same time, competition over chips, data, and energy resources will intensify.

For Ukraine and other smaller players, this presents not just a challenge but a window of opportunity: by investing in talent, partnerships, and domestic AI models, the country can not only survive in this new reality but actively shape it. This ensures independence and prosperity in a world where AI is no longer just a tool — it is the foundation of the future.

Sources: AI Report 2027, RAND Corp, JLL, and comments from Dmytro Dubov, Oksana Ishchuk, and Volodymyr Kulinich.