Brookfield anticipates AI-oriented data centers will reach roughly 15 GW of capacity by the end of 2025, up from about 7 GW at the end of 2024
In sum – what to know:
$7 trillion AI infrastructure spend forecast (2025–2034) – $4T for chips, $2T for data centers, $500B each for power/transmission and supporting tech.
75GW of new AI data center capacity expected – From 7 GW in 2024 to 82 GW by 2034, alongside a GPU base rising from 7M to 45M units.
Inference to dominate AI workloads – 75% of compute demand by 2030, fueling GPU-as-a-service growth from $30B in 2025 to $250B in 2034.
Brookfield Asset Management projects that AI-related infrastructure will experience a huge expansion over the next ten years, with the company estimating a total spend exceeding $7 trillion and “low risk of overbuilding” due to sustained AI demand.
The investment firm, which has recently unveiled a dedicated AI infrastructure strategy, outlined its projections in a newly published white paper. Brookfield holds significant stakes in several data center operators, including Compass, Centersquare, Data4, Ascenty, Digital Connexion, and DCI, as well as extensive energy investments.
According to Brookfield’s new research, about $4 trillion of the total investments will go toward chips — covering semiconductor fabs and supply chains — while $2 trillion will be directed to AI data centers. Meanwhile, power and transmission infrastructure is expected to require a total investment of $500 billion, with another $500 billion earmarked for supporting technologies such as dedicated fiber networks, advanced cooling, and robotics manufacturing.
Brookfield anticipates AI-oriented data centers will reach roughly 15 GW of capacity by the end of 2025, up from about 7 GW at the end of 2024. Over the following decade, the firm expects an additional 75 GW to be built, bringing total capacity to around 82 GW by 2034—more than ten times the current level.
Brookfield’s white paper noted that the installed base of GPUs is forecast to grow from about 7 million units in 2024 to 45 million by 2034, a sevenfold increase. Brookfield also expects around three-quarters of AI computing needs to come from inference workloads by 2030, driven by the rise of complex AI agents that use multiple modes and data requests to complete tasks.
This trend will shift data center design priorities toward handling high-volume inference traffic, rather than focusing solely on large-scale training jobs. The company sees GPU-as-a-service providers—such as CoreWeave—expanding significantly, with the market growing from roughly $30 billion in 2025 to over $250 billion by 2034 as organizations seek on-demand access to AI compute capacity.
The company also stressed that rapid advances in hardware require modular data center designs, enabling quick upgrades to power and cooling systems as new chips and form factors emerge.
Earlier this month, Brookfield announced a dedicated investment strategy aimed at developing infrastructure to support AI as it seeks to ride the wave of accelerating global demand for data centers.
In a letter to investors shared alongside its quarterly earnings report, CEO Bruce Flatt and President Connor Teskey said the initiative will cater to the growing needs of hyperscalers, enterprises, and governments for scalable and integrated AI infrastructure solutions.
Brookfield emphasized that it is well-positioned to extend its leadership position in the field fo AI infrastructure, saying: “We have already deployed tens of billions into this area and see a significant pipeline of opportunities that fit our capabilities and integrated approach.”
Brookfield previously committed up to SEK95 billion ($9.8 billion) for a long-term AI infrastructure development project in Sweden, expected to span 10 to 15 years.
The investment committed in Sweden is one of Brookfield’s largest AI investments in Europe, and extends its partnership with the government, public authorities, academia and businesses in the region.
Brookfield also said that this investment will be chiefly focused on a new large AI facility in Strängnäs, Sweden, creating a strategic infrastructure asset to support the country’s national AI strategy.
Earlier this year, Brookfield announced a €20 billion ($22.8 billion) infrastructure investment program in France, which includes a €10 billion investment into the first AI factory in the country.
