KKR launches Helix with Nvidia and Vistra to build $10B in AI data centers

KKR, Nvidia, and Vistra are building a $10 billion AI infrastructure company called Helix. The private equity giant announced the new venture on June 20, 2026, alongside a power supply partner and the chipmaker whose GPUs will fill the data centers Helix plans to build. Helix will finance, own, and operate the next generation of AI-ready data centers, with secured power contracts as a starting point rather than an afterthought.

The company starts with real assets, not a press release. Per the announcement reported by Reuters, Helix will deploy more than 1 gigawatt of IT capacity across its initial portfolio, with sites that already have grid interconnect and power purchase agreements in place. The reported target is to scale toward $10 billion of cumulative project investment over the next several years. KKR is anchoring the equity, Nvidia is supplying the GPU and networking stack, and Vistra is providing the behind-the-meter power that hyperscale AI sites need to come online on schedule.

A 1 GW starting footprint puts Helix in the same conversation as CoreWeave, Crusoe, and Nebius, the neoclouds that have built multi-billion dollar AI compute businesses by securing scarce power. The difference is structure. Helix is a developer and owner, not a renter. It will finance the buildings, the substations, and the long-term power contracts, then lease capacity to AI cloud providers, enterprises, and sovereign customers. Nvidia's involvement is the other tell. When a chipmaker co-builds a data center platform, the reference architecture tends to follow.

For broader context on how the AI infrastructure market is consolidating, the enterprise AI resource page tracks the largest private capital commitments and the corporate partnerships shaping who owns the next wave of AI compute. The Helix deal fits that pattern: a single vehicle that brings together private capital, chip IP, and power generation in one stack.

Why the Helix structure matters now

For the last 18 months, the AI infrastructure buildout has run into the same wall: there are far more willing buyers of AI compute than there are shovel-ready sites. Hyperscalers like Microsoft, Google, and AWS can fund their own campuses, but they want optionality outside their own regions and grids. Sovereign buyers want compute that lives inside their borders. Enterprises running regulated workloads want a third-party alternative that does not require a five-year deal with one of the three big clouds.

The shortage is not really about chips. Nvidia can ship Blackwell systems on a roughly 12-month forward order book. The bottleneck is the data center itself: a 100 MW AI campus takes 18 to 36 months to plan, permit, energize, and fill. Most projects slip because they cannot get a power interconnect from the local utility, not because they cannot get a GPU.

A similar dynamic played out earlier this year when Nvidia committed $2.1 billion to IREN for 5 gigawatts of AI infrastructure, a deal that paired chip supply with a buildout partner. Helix follows the same template but with private equity capital and a power generator in the seat the IREN deal gave to a publicly traded neocloud.

Helix is built to attack that bottleneck. The Vistra partnership matters because Vistra owns and operates one of the largest competitive power generation fleets in the United States, with sites in Texas that already have multi-hundred-megawatt grid connections and the cooling and water infrastructure for hyperscale campuses. Several reported Helix anchor sites are at or adjacent to existing Vistra plants, which means the company is starting from a place where the hardest part of the build, the utility interconnect, is already in place.

KKR's role is the capital and the financing structure. The firm has been one of the most active private equity investors in digital infrastructure since 2022, with prior positions in companies like FiberCop, Telecom Italia NetCo, and Avantus. Helix lets KKR package the AI infrastructure thesis into a single vehicle that can co-invest with sovereign wealth funds, pension funds, and corporate strategics. The reported $10 billion cumulative target is the size of a mid-tier infrastructure fund, and KKR is well placed to syndicate pieces of it to other long-duration investors.

Inside the Helix operating plan

Nvidia's stake goes beyond supply. The chipmaker has been increasingly active in the data center real estate stack, including a reported $6.9 billion investment in CoreWeave announced earlier this year. By co-developing the Helix platform, Nvidia locks in a reference deployment for its Blackwell and Rubin product lines and gets a seat at the table on how those systems are cooled, networked, and operated. It also means Helix is more likely to get early access to next-generation silicon, which is the real moat for any AI cloud.

The first Helix sites are expected to break ground later in 2026 and come online in 2027, with most of the early capacity allocated to a mix of U.S. enterprise customers, AI cloud resellers, and at least one reported sovereign deployment outside North America. The company is hiring an operating team separate from KKR's infrastructure arm, with the CEO slot reportedly already filled by a former data center executive.

For enterprise buyers, the practical impact is more competition in the AI compute market and more flexibility on contract structure. Today, a large enterprise that wants 50 MW of dedicated AI compute has only a handful of realistic counterparties. Helix adds another option, and one that is structured to offer long-term, fixed-price power contracts rather than the spot-pricing model that some neoclouds use. For the bigger hyperscalers, Helix is a sign that the AI buildout is going wider rather than deeper, with more independent capacity outside the top three clouds.

For Vistra, the deal is a hedge against the long-term risk that AI workloads and data centers start to compete with the broader grid for power, rather than ride alongside it. By owning the behind-the-meter generation, Vistra captures the margin on the power sale without depending on utility rate cases.

For Nvidia, the strategic question is whether the chipmaker becomes a permanent fixture in AI real estate. The CoreWeave investment, the reported discussions with Applied Digital, and now Helix suggest the answer is yes. Nvidia is acting less like a chip supplier and more like a vertically integrated AI infrastructure company that just happens to sell its compute to other people.

The bigger pattern is that AI infrastructure is no longer a build-it-yourself market. Hyperscalers, neoclouds, sovereign buyers, and now private-equity-backed developers are all trying to capture parts of a stack that used to be invisible. The companies that figure out how to package power, real estate, financing, and silicon into a single product will own the next decade of AI deployment. KKR is betting $10 billion that it can be one of them, with Nvidia and Vistra along for the ride.