Microsoft announced it has signed a major new renewable energy framework deal with Brookfield to secure sustainable energy for its data centers amid soaring power demands driven by AI.
The hyperscaler committed to invest an estimated $10 billion in renewable energy projects to be developed by Brookfield, which it said will unlock a further 10.5 gigawatts of generating capacity between 2026 and 2030 in the US and Europe.
The deal, which BloombergNEF described as the largest corporate agreement to purchase renewable energy to date, also includes the potential to deliver additional capacity to regions such as Latin America, India, and Asia.
Brookefield president and CEO of its renewable energy arm, Connor Teskey, described this agreement as the first of its kind in terms of scale.
“This first of its kind agreement, which is almost eight times larger than the largest single corporate PPA ever signed, is a testament to our ability to reliably deliver clean power solutions at scale to our corporate partners and accelerate the energy transition.”
Why Microsoft is ramping up alternative energy investment
The move from Microsoft marks the latest in a string of deals from the tech giant to tap into alternative energy sources amidst soaring data center infrastructure power demands.
In October 2023 Microsoft signed a similar PPA agreement in Japan with Shizen Energy that uses energy generated by a 25 megawatt solar park.
Microsoft is not the only hyperscaler exploring reliable sources of clean energy for its cloud infrastructure, either. Both AWS and Google Cloud have made similar moves in recent months.
In January, Google announced it had started building a $1 billion data center north of London that will be powered by renewable energy generated by Scottish offshore wind.
The new data center builds on a previous deal made by Google with French energy ENGIE in 2022 that comprised a 100 megawatt corporate power purchase agreement to provide it with over 5 TWh of green power from the Moray West wind farm.
Amazon also signed a PPA with ENGIE in January, increasing its total share of the output from the Moray wind farm to 473 megawatts, with the site set to become operational later in 2024.
With this agreement in place, Amazon said it will be the largest purchaser of renewable energy worldwide for the fourth year running.
AI explosion puts real pressure on an already constrained energy grid
One of the early movers in the generative AI race over the last 18 months, Microsoft has struck a lead over competitors in the space.
However, the sharpened enterprise focus on the power-hungry technology has placed significant strain on data center infrastructure globally, with industry analysts predicting a rapid surge in power needs over the next decade.
Research from JLL in February suggested data center operators and hyperscalers could face serious challenges meeting power demands in years to come, with regional power limitations expected to exacerbate a “scarcity of data center colocation supply”.
Niklas Lindqvist, general manager for the Nordics at IT infrastructure specialist, Onnec, said that big tech companies face a dilemma between ramping up data center energy capacity while balancing environmental goals.
Microsoft’s recent deal appears to be focused on offsetting the potential hit to ESG targets the company has set while still meeting surging demand.
“The boom in artificial intelligence (AI) and machine learning will pose significant challenges to data center operators attempting to stay in line with ESG commitments and upcoming regulations,” he said. “For example, the EU Corporate Sustainability Reporting Directive (SDS) will emphasize the essential nature of ESG reporting.”
“The GPUs that power AI at scale require radically different infrastructure from traditional CPU-powered compute. As AI’s power requirements surpass that of existing data center workloads, navigating this new energy landscape requires strategic planning and readiness for the evolving demands of the digital era.”
Moving forward, Lindqvist believes data centers will need to totally overhaul their operations in the future, with sites designed specifically for AI workloads.
“This will mean many operators must overhaul existing power, cooling, and cabling infrastructure. To ensure they’re prepared to handle AI’s energy and cooling demands, operators must design sites holistically, while also using AI to create energy efficiencies and identify opportunities to invest in materials that minimize waste.”
Alex McMullan, International CTO at Pure Storage, told ITPro these new data centers will place additional strain on an already struggling energy grid, noting that hyperscalers are trying to get around these complications by making deals directly with energy producers.
“[B]uilding more data centers compounds the overarching issue around broad AI adoption. That issue is that the new power-hungry GPUs, housed in newer and larger data centers, will place enormous burdens on existing power supply networks,” McMullan said.“This was highlighted by the National Grid CEO, who recently warned of AI’s demands on an already ‘constrained’ grid. In fact, Microsoft has already taken steps to bypass this problem by purchasing a nuclear power plant to power its own facility.”
McMullan said the long-term solutions to drive power-hungry AI workloads are not mature enough yet, but that the industry needs to find ways to alleviate the pressure already being put on the electricity grid if we are to keep up as these power demands scale.
“It’s going to take years to develop alternative energy sources to deliver long-term power to AI, and that will come in time, but the pressure on the electricity distribution grid is another critical element that needs to be solved in parallel”, he argued.
“The IEA is projecting that data center power consumed globally could double by 2026 at current growth rates. In the meantime, data center owners and their customer organizations are asking, ‘How can I absorb that increase within my current capability?’ as they consider how soon electricity capacity will be maxed out.”