On September 29, Samsung and five of its affiliates committed $1 billion to Helix Digital Infrastructure, the AI infrastructure platform KKR launched in June with the Kuwait Investment Authority, NVIDIA, and Vistra. The new money builds on more than $10 billion in long-duration capital committed at launch, pushing Helix’s capital base past $11 billion. Samsung Electronics accounts for $500 million of the commitment; Samsung C&T, Samsung SDS, Samsung SDI, Samsung Life Insurance, and Samsung Fire & Marine Insurance supply the other $500 million.
That headline — another billion for another AI vehicle — understates what is interesting about Helix. The interesting part is not the size of the capital. It is the composition of the table. A private equity manager, a sovereign wealth fund, the dominant AI chipmaker, a Korean industrial conglomerate, and one of America’s largest competitive power generators have all bought into the same vehicle, each contributing something the others cannot. Read it as a structure story first and a funding story second: Helix is the clearest attempt yet to fuse the data center, the chip, and the power plant into a single financed package.
What Helix actually is
Helix, announced June 11, 2026, describes itself as a single coordination point for hyperscalers’ data centers, power, connectivity, and related needs. It is a company — led by former Amazon Web Services CEO Adam Selipsky, with KKR’s global head of digital infrastructure, Waldemar Szlezak, as chief investment officer — funded through a long-duration capital fund, KKR Helix C L.P., an open-ended investment fund managed by KKR affiliates. It will seek to invest in and manage hyperscale data center development and operations, baseload and flexible power generation, transmission and distribution infrastructure, and fiber and connectivity infrastructure.
The scope matters. Most data-center vehicles own buildings and lease them to tenants, who then solve power separately. Most power vehicles sell electrons and let someone else worry about the building. Helix proposes to own and coordinate the whole stack: the site, the servers’ home, the generation, the wires, and the fiber. NVIDIA serves as a strategic partner, supporting deployment of NVIDIA DSX AI factory-aligned infrastructure aimed at maximizing tokens per watt and accelerating time to first token. Vistra is the preferred power provider for Helix investments. Samsung brings capabilities Helix says it expects to explore across advanced technology, construction, energy storage, and cooling.
Four months after launch, Helix has publicly announced no specific site or project. What has been assembled so far is capital, partnerships, and a management team — not yet a pipeline the public can evaluate. The $11 billion-plus is committed capital, not deployed capital, and the distance between those two things is where this story will be decided.
The capital stack, consortium edition
Read Helix through the capital-stack lens developed in this series’ earlier explainers, and it looks like a project financing scaled up to the level of an entire sector. In a classic project financing, a single-asset company raises non-recourse debt secured by contracted revenue, with equity beneath it absorbing first losses. Helix generalizes that logic: a platform with a pool of committed equity intended to fund a portfolio of assets, each of which will presumably carry its own contracts and its own leverage in time.
The committed capital is the equity layer. KKR, the Kuwait Investment Authority, NVIDIA, Vistra, and now Samsung have all contributed or committed long-duration capital — money with a horizon measured in years to decades, not quarters. That duration is the point. Data centers and power plants are 20-to-30-year assets; financing them with capital that needs an exit in five years creates exactly the refinancing risk that has sunk infrastructure strategies before. Sovereign wealth and insurance-affiliate capital is patient by design — a genuine advantage in a buildout of this size.
What each investor contributes beyond money is what distinguishes Helix from a plain infrastructure fund. KKR contributes the platform, the management team, and the deal flow. The Kuwait Investment Authority contributes scale and a sovereign’s tolerance for long-dated illiquid assets. NVIDIA contributes the technology stack and a view into what hyperscalers will need next. Samsung contributes manufacturing depth: semiconductors, construction, batteries, cooling, and insurance capital in one corporate family. And Vistra contributes the one input none of the others can manufacture: electricity, at scale, from an operating fleet.
Vistra’s two-tranche commitment is the key mechanism
The most revealing document in the Helix story is not a press release. It is Vistra’s second-quarter 10-Q, filed with the SEC, which discloses the precise structure of the generator’s commitment. Vistra committed up to $1 billion to the Helix fund, consisting of an initial $500 million commitment plus an additional $500 million subject to the fund meeting certain commercial power-supply milestones for Vistra’s existing generation assets — or, at Vistra’s election, regardless of whether those milestones are achieved.
Pause on that structure, because it is doing real work. The second $500 million is conditioned on Helix actually buying power from Vistra’s existing fleet. Vistra has tied half of its equity commitment to its own commercial success as Helix’s power supplier: the fund gets the second tranche when — and, in effect, because — Helix projects start consuming Vistra electrons. Vistra retains an escape valve, the right to fund the second tranche at its own election even if the milestones are not met. But the default logic is alignment: capital follows contracts. That is a materially different commitment from a passive limited-partner stake, whose check depends only on the manager finding good deals.
Preferred provider is not a PPA
It is worth being precise about what “preferred power provider” means, because it is easy to overread. It is not a power purchase agreement. A PPA is a binding contract for a defined quantity of power from a defined asset at defined terms — the instrument this series examined in the “Data-Center Power as an Asset Class” explainer, and the instrument Vistra used in January 2026 when it signed 20-year PPAs with Meta covering 2,609 MW, including the full 433 MW of planned nuclear uprate capacity across its Beaver Valley, Davis-Besse, and Perry plants. A PPA creates revenue. A preferred-provider designation creates a right — typically a first look, a seat at the table, a presumption that the provider’s offer gets serious consideration before others are solicited.
The difference is the difference between contracted cash flow and pipeline visibility. Vistra’s Meta PPAs are bankable: a lender can underwrite them. The Helix preferred-provider role is not bankable in the same way, because no megawatt is yet contracted under it. What it gives Vistra is earlier and better information about where Helix projects will be sited, how much power they will need, and when — and a favored position in the negotiation that follows. In a market where interconnection queues and generation scarcity make timing everything, early information has real value. But investors should not confuse a favored position with a signed contract.
The milestone structure bridges that gap. By conditioning the second $500 million on power-supply milestones, Vistra has effectively said: the preferred-provider role is worth paying for in full only if it converts into actual power sales. If it does, the equity and the power revenue arrive together, and Vistra owns a piece of the platform it supplies. If it does not, Vistra has limited its exposure to $500 million and kept the option to fund the rest anyway. It is a hedged bet with a built-in performance test.
From selling electrons to owning the load
Step back, and Helix represents the logical endgame of a trend this series has tracked all year: the generator moving up the value chain from selling power to owning a stake in the load itself. The traditional merchant model is transactional — build a plant, sell its output into the market or under a PPA, repeat. The Helix model is positional: own a piece of the vehicle that decides where data centers get built, what power they buy, and on what terms.
The comparison that clarifies the shift is Talen Energy’s arrangement at Susquehanna, where Talen sold its data-center campus adjacent to the nuclear plant to AWS and signed a long-term PPA to supply the site. That deal made Talen the supplier to a site it once owned — a deep, durable relationship, but still a supplier relationship. Helix goes one step further. Vistra is not just supplying a Helix site; it is an equity owner of the platform that will choose the sites.
Why would a generator want that? Because the scarcest input in the AI buildout is not capital — capital is abundant, as the $11 billion attests — and it is not chips, which NVIDIA supplies. The scarcest input is firm power, delivered to the right site on the right timeline, with the interconnection to match. Whoever controls the power decision controls the gating item for every project. A generator that merely sells power captures the commodity margin on the electrons. A generator that co-owns the platform captures a share of the development margin on the whole project — the land, the building, the fiber, the financing — plus the power margin. That is the vertical-integration endgame: the power plant becomes not just an input to the data center, but a shareholder in it.
The questions a structure like this raises
Enthusiasm should be tempered by the questions Helix has not yet answered. The first is the pipeline question: with no announced site or project four months after launch, there is nothing yet to underwrite except the partners’ reputations and the capital’s patience. A platform with $11 billion and no projects is a promise; a platform with one financed, sited, interconnected project is a business.
The second is the governance question. When a supplier is also an owner, conflicts follow. Vistra sits on both sides of the power negotiation: as Helix’s preferred provider selling power, and as a Helix investor whose returns depend on Helix’s project costs, which include the price of that power. That tension is manageable but real, and the terms on which Helix buys power from its own investor deserve scrutiny when they are eventually disclosed. The milestone structure sharpens this tension: it gives both sides a reason to transact with each other, which is the point, but it also means the arm’s-length test for those transactions matters more, not less.
The third is the demand question that hangs over every data-center power investment. Helix is premised on hyperscale AI demand absorbing a very large buildout of data centers and power infrastructure over the coming decade. If that demand grows more slowly than expected — because of efficiency gains, architectural shifts, or simple overbuilding — the platform holds long-duration capital in long-lived assets with fewer buyers than planned. The committed-capital structure cushions this risk: unlike a leveraged developer that must service debt on a fixed schedule, Helix’s equity can wait. But patience is not immunity.
The fourth is the concentration question. Helix’s pitch is integration — one coordination point for the whole stack. Integration’s mirror image is single-point exposure: a platform whose data centers, power, and connectivity are all bound together shares risks that a fragmented market diversifies. A permitting delay on the power plant becomes a delay on the data center; a technology shift in cooling or chips ripples through the whole portfolio.
What to watch
Two milestones will determine whether Helix becomes a template or a footnote. First, the first project: a named site, a named hyperscaler customer, and a power plan with a source, a size, and a timeline. Second, Vistra’s second $500 million tranche: whether it funds through power-supply milestones or through Vistra’s election will reveal how quickly the preferred-provider role is converting into real power sales.
The bottom line
Helix is best understood not as a data-center company or a power deal, but as an attempt to reprice the boundary between them. The old boundary was a contract: the data center bought power from the generator at arm’s length, and each side kept its own margin and its own risk. Helix erases that boundary by putting both sides inside one capitalized vehicle, with the generator’s equity commitment conditioned on the power relationship materializing.
For power investors, the lesson is broader than one platform. The asset class this series described — data-center power as contracted infrastructure — is evolving from contracts into ownership. The PPA was the first step: long-term revenue certainty for the generator. The preferred-provider stake is the second: a share of the platform that creates the demand. If Helix’s first projects pencil out, expect the industry’s next question to be not “who signed the PPA?” but “who owns the platform?”
None of this is investment advice, and nothing in this analysis recommends buying or selling any security. What it does suggest is a structure worth watching: committed capital, milestone-linked alignment, and a generator that stopped merely selling electrons and bought into the buildout itself. Whether that structure earns its keep will be decided project by project, megawatt by megawatt.