However, despite the intense investment activity, the emergence of DeepSeek, a new, more efficient open-source AI platform, has sparked discussions about whether all these new facilities are truly necessary. If the latest AI models consume significantly less energy and storage space, then estimates regarding the need for massive investments in data centers may be excessive.
Investment Boom and Gigantic Data Centers
The development of AI has doubled the average size of proposed data centers, from 150 megawatts in 2023 to 300 megawatts in 2024. Meta announced the construction of a $10 billion data center in Louisiana, covering a total area of 4 million square feet, which will consume energy equivalent to two large nuclear reactors. At the same time, the Virginia Data Center Market continues to expand, with Ashburn considered the "capital of data centers," while large investments are also being made in Europe, primarily in Germany, Sweden, and Ireland.
K.R. Sridhar, founder and CEO of Bloom Energy, highlights the huge increase in demand for computing power but also the challenges this growth brings. "In 2010, it took around 1,500 watt-hours to transfer one gigabyte of data. Last year, this decreased to a tenth... but the overall increase in traffic, and thus chips and energy, was double."
This raises a concern: Is the rapid expansion of data centers sustainable, or is it an "infrastructure bubble" that could lead to oversupply?
The Role of DeepSeek and the Challenges of Oversupply
DeepSeek, a new AI platform developed for maximum energy efficiency, raises doubts about whether the growing development of data centers is truly necessary. This technology consumes less energy and requires smaller infrastructure, which could lead to a reassessment of investments in the data center sector.
According to Victor Leung, Senior Vice President of Morningstar DBRS, "DeepSeek's approach to building and training its model can be seen as an effort toward more efficient models, which is expected as companies seek to reduce the cost of AI development."
This raises the question of whether the market has underestimated how quickly AI will improve the efficiency of computing systems, potentially reducing the need for more data centers in the future.
Impact on Real Estate and Infrastructure Markets
Data centers are a significant part of the real estate market, as they require vast land areas, specialized infrastructure, and substantial investments in energy systems. Especially in the U.S. and Europe, locations near industrial zones and areas with low energy costs have become hotspots for new investments.
However, if technologies like DeepSeek reduce the need for large facilities, the demand for new data centers could stabilize or even decrease, affecting investor returns.
Ashish Nadkarni, Vice President at IDC, notes that tech giants won't immediately change their strategies, as investing in more computing power is considered essential for their long-term survival. "The need for increased AI capacity is not something that can be ignored. Large companies will continue to invest, even if the technology becomes more efficient."
Development or "Bubble"?
The construction of new data centers is at the peak of investments, but the evolution of artificial intelligence and technological efficiency could lead to a reversal of current predictions.
DeepSeek may just be the beginning of a new generation of AI that requires less energy and less infrastructure, raising concerns about the future of properties designated for data centers. If developments continue at the same pace, the investment frenzy could turn into an oversupply of space, pushing prices down.
The data center industry is at a critical crossroads. Technology is evolving rapidly, but investors are being asked to predict whether the market will continue to grow or if we are approaching a saturated market of overbuilt infrastructure.
