The AI data center race is now a race to power

The AI data center race has shifted focus from the best chips and models to the energy that powers them. As AI moves from research to real-world applications, access to reliable electricity and the speed of data center grid connections (known as "time to power") are becoming crucial factors in determining where data centers, or "AI factories," are built and who benefits economically.

Investors now require more than just assurances of future electricity availability; they need reliable, competitive, and timely power. Three key factors influence where AI investments are made: the regulatory environment, energy availability, and energy prices. However, investors are increasingly seeking certainty that permits will be secured and grid connections delivered on schedule, with power costs remaining affordable.

China and the United States lead global AI data center capacity but are pursuing divergent strategies. China is investing approximately $295 billion in a nationwide AI computing network and associated energy infrastructure and renewables. In contrast, the US is grappling with increasing electricity demand and evolving grid-connection timelines, which are prompting interest in on-site generation and storage solutions.

Europe presents a more complex landscape, with some markets offering favorable conditions and opportunities, while others face challenges such as grid congestion and lengthy permitting processes. Despite these hurdles, Europe possesses a wealth of resources, including nuclear power and renewables, established infrastructure, and deep industrial expertise, making it well-positioned to become a hub for the next wave of AI infrastructure.

To capitalize on the transformative potential of AI, three key measures should be implemented. First, speed must be prioritized. Energy systems are struggling to keep pace with the rapid growth in electricity demand, with an estimated 2,500 gigawatts of generation capacity currently waiting in connection queues worldwide. In Europe's data center hubs, data centers can face waiting periods of up to a decade for grid connections, while a facility might only take four years to build.

Governments and grid operators must expedite permitting and connection processes, moving from a "first come, first served" approach to prioritizing "first ready, first connected" projects. The UK's efforts to streamline these processes, reducing waiting times from up to 15 years to shorter intervals, demonstrate the effectiveness of such reforms.

Second, flexibility is essential for quickly unlocking existing energy capacity. While major grid upgrades could take up to 15 years, energy-management technologies can optimize energy use in various sectors, including buildings, transportation, and industry. By shifting non-critical demand away from peak periods and utilizing battery storage, grid capacity can be expanded more rapidly.

Data centers themselves can also serve as active grid assets through intelligent energy management, potentially doubling the available grid capacity with just 30 hours of flexibility per year.

Lastly, making flexibility the most cost-effective path is crucial. In regions with existing grid flexibility, such as France, Germany, Italy, and the Nordic countries, households and businesses can be incentivized to use power during periods of abundant renewable energy and reduce demand during peak times. Data centers can benefit from lower tariffs or faster staged connections in exchange for investing in battery storage and automated demand response systems.

By adopting these measures, countries can still reshape the map of AI infrastructure investment, ensuring that the economic benefits of AI are widely distributed.