AI infrastructure is getting messy, expensive, and increasingly difficult to coordinate. The industry has spent the last few years talking about models, chips, etc., but behind the scenes, a fun new challenge has been piling up.
Actually building the infrastructure needed to support all of that demand is getting complicated.
For those who have been monitoring the situation in any capacity, this isn’t exactly new, but it’s quickly becoming a hard-to-ignore problem. So, on today’s episode of “hey maybe we should hit pause all of this for a minute…”
KKR just launched Helix Digital Infrastructure, a new company backed by more than $10 billion in committed capital, designed to bring data centers, power, connectivity, and related infrastructure under one roof. The company launched with backing from KKR, the Kuwait Investment Authority, NVIDIA, and power provider Vistra.
Former Amazon Web Services CEO Adam Selipsky will lead the company as co-founder and CEO.
Why AI infrastructure is a coordination problem
The basic idea behind Helix is that AI infrastructure projects often involve multiple moving parts that don’t always move at the same pace. Again, this has been a snag for a while, but it’s starting to unravel the whole sweater.
Data centers need power (lots of it). Power projects need grid connections. Connectivity needs to be in place before any of it becomes useful.
Large cloud providers can manage most of this themselves, but coordinating everything at the scale AI demands has become harder as projects grow and timelines tighten.
“Large users of digital infrastructure have an urgent need to reduce complexity and unlock new capacity,” Selipsky said in the company’s launch announcement. “Helix combines significant long-term capital with the capabilities and expertise to deliver holistic AI infrastructure solutions with speed and scale.”
Helix isn’t launching as an investment fund. The company plans to build and manage the infrastructure behind AI growth, including data centers, power, transmission, fiber, and connectivity.
Its goal is to bring more of those pieces together for hyperscalers that are trying to add capacity quickly.
The growing AI power issue
One thing Helix keeps coming back to is power. A lot of the AI conversation still revolves around chips, but plenty of projects are running into a more basic problem… getting enough electricity where it’s needed, when it’s needed.
Data center demand keeps getting more, well, demanding, while new generation and grid upgrades can take years to bring online.
That’s where Vistra comes in. The company will be Helix’s preferred power provider, giving it access to existing generation assets and experience in building and operating energy infrastructure. NVIDIA’s role is different. It will work with Helix on infrastructure deployments built around its AI systems and designs.
“Useful AI has arrived, and demand for AI factories is extraordinary,” NVIDIA CEO Jensen Huang said in the announcement. “AI is driving the largest infrastructure buildout in modern history.”
More money, more infrastructure
The launch is also a rather big signal of where a growing share of AI investment is heading.
For a while, the bulk of the attention centered on foundation models and chipmakers. Now, though, investors are looking further down the stack at the physical infrastructure needed to support those systems over the long term.
Helix has come out of the gate with plenty of resources behind it, including major investors, power expertise, and leadership that has spent years inside the cloud industry.
It hasn’t put any specific projects on the board yet, but its entire premise is that AI growth depends on much more than just better models. Somebody still has to build the data centers, find the power, and connect it all together.
NVIDIA has spent much of the past year talking about AI factories and the infrastructure needed to support them at scale. Helix brings together several of the same ingredients, including NVIDIA technology, power generation, and long-term capital, in an effort to help get those projects built faster.