The $1.3B Bet That AI Needs Its Own Power Grid
Image: The $1.3B Bet That AI Needs Its Own Power Grid
Brookfield just launched Radiant by merging with UK cloud firm Ori Industries. The goal is not to build another AI startup. It is to become the electrical grid that all AI runs on.
Brookfield Asset Management is not a tech company. It manages $1 trillion in real assets: power grids, toll roads, pipelines, ports. That background is the whole point.
On February 24, Brookfield officially launched Radiant, a new AI infrastructure company, by merging it with Ori Industries, a London-based GPU cloud provider. The combined entity was valued at $1.3 billion. The announcement was quiet. The implications are not.
The pitch is this: the world is about to spend decades building AI compute capacity, and someone needs to own the physical layer underneath it. Not the model, not the application, not even the data center in the conventional sense - the vertically integrated stack of land, power, hardware, and software that makes AI physically possible at national scale.
Radiant wants to be that stack.
What Ori Brought
Ori Industries spent seven years building what its founder Mahdi Yahya describes as "software that could enable infrastructure for AI at scale." In practice, that means a distributed GPU cloud with on-demand inference, Kubernetes orchestration, model registries, fine-tuning pipelines, and bare metal access. It is an NVIDIA Cloud Partner, running the latest Blackwell and GB200 NVL72 chips, with Rubin access planned.
Ori was a credible but small player in a crowded compute market. What it lacked was capital at infrastructure scale. A startup cannot build sovereign compute for governments on venture money.
Brookfield can.
What Brookfield Brings
The parent fund, Brookfield's AI Infrastructure Fund (BAIIF), has a $100 billion investment pipeline earmarked for AI infrastructure. Radiant is its first operating deployment. The money comes from the same playbook Brookfield used to build power grids and pipelines - long-term capital, long-term contracts, infrastructure-grade returns.
The structure matters: Radiant signs long-term contracts with sovereign governments, telecom providers, and large enterprises. That is not how SaaS pricing works. It is how a toll road works. Brookfield is treating GPU compute the way it treats a power transmission line - an essential utility that governments will pay for under decade-long agreements rather than monthly subscription cycles.
NVIDIA is onboard as a reference design partner. Radiant's data centers follow the NVIDIA DSX blueprint - a standardized "AI factory" design that NVIDIA has been pushing as the physical template for national AI infrastructure. That alignment gives Radiant preferential access to the latest silicon at scale.
The Sovereign AI Angle
The customer list Radiant is targeting - sovereign governments, telecoms, select enterprises - tells you what problem they are really solving. Countries that do not want their AI workloads running on US hyperscaler infrastructure (Azure, AWS, Google Cloud) need alternatives. France does not want Paris's AI running in Virginia. The UAE, Saudi Arabia, India, and dozens of others are all in various stages of building national compute capacity.
Radiant's pitch is that it can deliver NVIDIA-grade AI factories to these governments without them having to negotiate directly with Big Tech. It is a B2G play (business to government) dressed in infrastructure clothing. Brookfield's credibility with governments in emerging markets - built over decades of power and transport deals - is the sales channel that no GPU startup has.
The Second-Order Effect
Here is what the coverage mostly missed: this deal signals that the AI infrastructure buildout is entering the asset class legitimization phase.
When traditional infrastructure capital - the kind that funds highways and water treatment plants - starts entering a technology sector, it means institutional money has decided the risk profile is manageable. Pension funds and sovereign wealth funds follow Brookfield's playbook, not Sequoia's. Their entry sets floor valuations, stabilizes long-term demand signals, and crowds out speculative players who cannot compete on capital cost.
For NVIDIA, this is also quietly excellent news. Long-term government contracts lock in Blackwell and Rubin purchases years in advance, providing demand visibility no hyperscaler purchase order can match. NVIDIA's partnership with Radiant is not a marketing arrangement - it is revenue backlog.
"We have created a unique platform that lowers the cost of compute while raising the standard of performance. It's an approach designed for the realities of the AI economy - efficient, resilient, and built to scale." - Vishal Padiyar, Managing Director, Brookfield
Whether Radiant actually executes at the scale it is claiming is a separate question. Building sovereign AI factories for governments across multiple continents involves regulatory complexity, local partnerships, and energy sourcing challenges that no press release addresses. The gap between infrastructure vision and delivered gigawatts is where most of these plays quietly fail.
But Brookfield has built things before. The $1.3 billion valuation is a starting number. The real story will be what BAIIF's $100 billion pipeline looks like in five years - and whether Radiant ends up as the backbone nobody talks about because it just works, the way power grids do.