Google and Meta Bet on Energy to Support AI Growth

The Rise of AI Investments and Their Energy Challenges

In recent years, venture capital investors have heavily backed startups specializing in artificial intelligence, injecting over $500 billion into the sector. However, a recent report from Sightline Climate suggests that the future of investments may well shift towards the energy sector. Indeed, researchers have highlighted that up to 50% of announced data center projects are at risk of significant delays, primarily due to difficulties in accessing energy.

Among the 190 gigawatts of data centers tracked by Sightline, only 5 gigawatts are currently under construction. Last year, about 6 gigawatts of projects came online, but a much larger number, approximately 36%, have seen their deadlines pushed back to 2025. These delays could have repercussions for large companies and other firms integrating AI into their daily operations.

Tech Giants Turn to Renewable Energy

In the face of these challenges, major tech companies such as Google and Meta have taken proactive measures by investing heavily in renewable energy projects, including solar, wind, and nuclear. These companies are not only directly funding these initiatives but are also collaborating with utilities to accelerate the adoption of emerging technologies like the 100-hour battery developed by Form Energy.

Many startups are entering the race to solve energy supply issues. For example, Amperesand, DG Matrix, and Heron Power are working on new energy conversion technologies, while Camus, GridBeyond, and Texture are developing software to optimize the management of electron flow.

Energy remains one of the main obstacles for data centers, and this constraint is unlikely to disappear anytime soon. According to Goldman Sachs, energy consumption by data centers could increase by 175% by 2030, largely due to the rise of AI.

Towards Alternative Energy Solutions

Companies like Amazon, Google, and Oracle are exploring ways to reduce their reliance on the traditional power grid. Several data centers are being planned with on-site energy sources or hybrid approaches combining on-site energy and grid connection.

The largest data centers are at the forefront of this transition. Less than a quarter of projects that have identified an energy source will opt for an on-site or hybrid solution, but these projects still represent 44% of total capacity. This shift is partly due to shortages of energy production equipment, including gas turbines, and an aging power grid, paving the way for alternative energy sources.

An example of this approach is Google's recent agreement to power a new data center in Minnesota. The company plans to combine wind and solar energy with a massive 30-gigawatt-hour battery from Form Energy. Google has also collaborated with Xcel Energy to create a new pricing structure that it believes will encourage the adoption of new technologies in utility planning.

Energy Management: A Crucial Challenge

Energy supply is just one part of the equation. Once energy reaches the grid or the data center, it must be managed efficiently, a task that primarily relies on transformers.

Most current transformers are based on a 140-year-old technology, using large blocks of iron wrapped in copper wire. While reliable, this technology becomes cumbersome as the energy needs of data centers increase. When a server cabinet reaches 1 megawatt of power density, the equipment required to operate it occupies twice the space of the cabinet itself, according to an expert cited by TechCrunch.

This is why investors are increasingly interested in startups developing solid-state transformers, which hope to replace the old technology with silicon-based power electronics. Although these new transformers are more expensive, they offer sufficient flexibility to replace multiple pieces of equipment in a data center, potentially making them cost-competitive.

Overall, investments in battery and transformer companies remain modest compared to the spectacular fundraising seen in the AI sector. However, this situation could be advantageous for investors, as these funding rounds are more manageable. Moreover, as the world moves towards widespread electrification, the demand for energy will continue to grow, providing a hedge against a potential collapse of the AI market. Ultimately, the best investment in AI may well be found in energy itself.