September 30, 2026 · Brendan Hogan, Founder, GridPermit

Samsung Just Put $1 Billion Into a Data Center Company That Hasn't Named a Single Site.

Helix Digital Infrastructure launched in June with $10 billion and Vistra as its "preferred power provider." Samsung's September 29th investment pushed that past $11 billion. Helix still hasn't announced where any of it gets built.

On September 29th, Samsung Electronics and five of its affiliates announced a combined $1 billion investment in Helix Digital Infrastructure, the AI-infrastructure company KKR launched in June with the Kuwait Investment Authority, Nvidia, and the power generator Vistra. The check pushes Helix's total committed capital past $11 billion. It's real money from a serious list of backers, run by a genuinely credible operator — Adam Selipsky, who spent three years as CEO of Amazon Web Services before taking this job. It's also, as of this writing, capital committed to a company that has not announced a single data center site.

What Helix is actually selling

KKR launched Helix in mid-June with more than $10 billion in long-duration capital commitments from its three founding partners, and the pitch was specific: not just financing, but hyperscale data centers with power pre-arranged. Nvidia signed on as a strategic partner to align Helix's builds with its DSX AI-factory reference architecture. Vistra — which runs a 50-gigawatt generation portfolio across 18 states — became what both companies are calling Helix's "preferred power provider." Selipsky's own framing leaned hard into the pain point everyone building AI infrastructure right now actually has: "Large users of digital infrastructure have an urgent need to reduce complexity and unlock new capacity," he said at launch. "Helix combines significant long-term capital with the capabilities and expertise to deliver holistic AI infrastructure solutions with speed and scale." Strip the word "unlock" out of that sentence and the underlying claim is a real one worth taking seriously: a hyperscaler or large enterprise tenant could, in theory, sign with Helix and skip years of separately chasing land, permits, and an interconnection queue position, because Helix's financial sponsors already include the company that would otherwise be on the other side of the power purchase agreement.

Why Samsung wanted in, split six ways

Samsung's $1 billion isn't coming from one corporate treasury. Samsung Electronics is putting up $500 million; the rest is split across Samsung C&T, Samsung SDS, Samsung SDI, Samsung Life Insurance, and Samsung Fire & Marine Insurance. That's an unusual capital structure for a single strategic bet, and it tells you something about what Samsung is actually buying. For decades Samsung's role in any data center or AI buildout was as a component supplier — memory chips, primarily, sold to whoever was building. This investment is explicitly framed, by Samsung's own newsroom announcement, as a shift from hardware vendor to what the company is calling an AI-infrastructure ecosystem architect: a seat at the table where construction, energy storage, and cooling contracts get decided, not just a purchase order for DRAM once the building is up. Spreading the check across six affiliates — including two insurers — reads less like caution about Helix specifically and more like Samsung structuring a long-duration infrastructure bet the way an insurer's balance sheet is actually built to hold one, spread across entities with different capital and liability profiles rather than concentrated on one electronics maker's balance sheet.

"Secured power" is doing a lot of work in that sentence

Here's the part worth sitting with. Vistra being Helix's preferred power provider is a corporate-level relationship between a generator and a capital platform — it is not a signed interconnection agreement, a specific plant assignment, or a queue position at any grid operator. It's a real advantage; a developer with a standing relationship to a 50-gigawatt generation fleet starts a power conversation from a materially better position than one starting cold. But it answers a different question than the one a specific site actually has to answer, which is whether a specific parcel, in a specific utility's territory, has a specific megawatt allocation it can point to in writing. Helix hasn't named a site yet, which means none of its capital — not KKR's, not Nvidia's, not Kuwait's, not now Samsung's — has been tested against that question anywhere. The company is, by its own account, still assembling the platform. The power story is a corporate-level premise the market is pricing in advance of any project where it gets proven out.

The diligence angle

This is the same evidence-ladder problem we keep running into from the other direction — Texas's new interconnection deposit rule exists because regulators stopped trusting a developer's claimed megawatt figure until it's backed by real collateral, and PJM's capacity shortfall is what happens when a grid operator's own planning assumed load growth would show up on schedule. Helix inverts the failure mode: instead of a developer overstating demand, it's a capital platform whose top-line pitch — power is handled — is true at the corporate level and unverifiable at the site level, because there isn't a site yet. In a Kill Test, we'd grade "preferred power provider from a company backed by our own power-generation investor" exactly where it belongs on the ladder: better than nothing, well short of a utility's written load-specific confirmation, and nowhere near a signed, collateralized interconnection agreement. None of that makes the $11 billion fake or the bet a bad one — Selipsky and his backers clearly believe a pre-arranged power relationship is worth paying up front for, and they may be right. It just means the interesting diligence question about Helix doesn't start until the first site does. Whichever state gets Helix's first announced project will be the one where "preferred power provider" either converts into a real agreement on the same terms the pitch implies, or doesn't — and that's the number worth watching for, not the size of the capital stack behind it.

← Back to the blog

Evaluating a specific parcel for a 50–300 MW load? GridPermit's Preliminary Site Kill Test grades the power, zoning, water, and commercial-readiness evidence before you buy the land.

Start a Qualification Review

See state-by-state coverage · Read the methodology · See a sample report