KKR, NVIDIA Launch $10B Helix to Bankroll AI Buildout

KKR on June 11 unveiled Helix Digital Infrastructure, a new platform seeded with more than $10 billion in long-duration capital commitments to finance the next wave of AI data centers, power generation and fiber connectivity. The founding investors are KKR, the Kuwait Investment Authority, NVIDIA and Vistra Corp. (NYSE: VST), with former Amazon Web Services CEO Adam Selipsky returning to cloud infrastructure as Helix’s chief executive (Yahoo Finance).

The launch is less a single transaction and more a new financing architecture for the AI buildout: a coordinated bundle of compute, power and capital underwritten by a sovereign wealth fund and a private equity sponsor rather than the traditional investment-grade bond market.

Inside the deal

Helix is structured as a stand-alone operating company with a dedicated management team and board. Selipsky — who ran AWS from 2021 to 2024 and was named a KKR senior technology adviser in September 2025 — will lead the business as co-founder and CEO. Waldemar Szlezak, KKR’s global head of digital infrastructure, takes the chief investment officer role.

The mandate spans the full AI-infrastructure stack:

Helix founder Role in the platform
KKR (NYSE: KKR) Sponsor, capital, M&A & structuring; CIO seat
Kuwait Investment Authority Sovereign wealth co-anchor; long-duration capital
NVIDIA (NASDAQ: NVDA) Strategic partner; deploys NVIDIA DSX AI-factory-aligned infrastructure
Vistra (NYSE: VST) Preferred power provider; baseload and flexible generation
Source: KKR/Helix joint announcement, as reported by Yahoo Finance, June 11, 2026.

Helix will invest in and operate hyperscale data centers, transmission and distribution assets, fiber connectivity, and — critically — baseload and flexible power generation. Vistra brings the muscle on the power side: the company operates across 18 states with 5 million retail customers and is expected to reach nearly 50,000 megawatts of generation capacity by year-end 2026 (Vistra IR).

Why the structure matters

The capital markets puzzle the hyperscalers face is not that money is unavailable. It is that the financing they need does not match the duration, complexity or risk profile of what investment-grade bond desks or speculative VC typically write.

J.P. Morgan estimates the data center buildout will require roughly $1.5 trillion of investment-grade bonds over the next five years, against a longer-term spending need above $5 trillion when power and related infrastructure are included (Data Center Dynamics, citing J.P. Morgan). Morgan Stanley’s most recent hyperscaler capex track now points to about $800 billion in 2026 and $1.16 trillion in 2027 — up from a prior $450 billion estimate (I/O Fund analysis of Morgan Stanley).

Hyperscaler AI capex trajectory, 2024-2027E ($ billions) Morgan Stanley’s most recent estimate of combined hyperscaler capex rising from roughly $250 billion in 2024 to over $1.1 trillion in 2027. 0 300 600 900 1200 2024A $250B 2025E $480B 2026E $800B 2027E $1.16T Hyperscaler AI capex (estimated, $B)
Source: Morgan Stanley estimates as summarized by I/O Fund; 2024 actual figures referenced by J.P. Morgan via Data Center Dynamics.

That funding stack splits into three binding constraints: compute (GPUs and AI-factory design), power (gigawatts, increasingly behind-the-meter), and capital (long-dated, patient, multi-billion-dollar tickets). Helix is engineered to address all three under one roof — NVIDIA supplies the silicon architecture, Vistra supplies the electrons, KIA and KKR supply the balance sheet. The pitch to hyperscalers is fewer counterparties, faster sequencing, and a financing partner that can write equity-like checks at infrastructure-fund duration.

Selipsky’s framing

“Large users of digital infrastructure have an urgent need to reduce complexity and unlock new capacity,” Selipsky said in the launch statement, telling CNBC that hyperscaler buildouts are getting “harder with AI and the pace of the buildouts and the scale of the buildouts” (GeekWire). NVIDIA CEO Jensen Huang called AI “the largest infrastructure buildout in modern history.” Vistra CEO Jim Burke said Helix puts “data center, infrastructure and power capabilities under a single umbrella.”

How it slots into the capital markets

For KKR, Helix is a portable platform — one that can absorb additional sovereign and institutional commitments as the program scales. The $10 billion+ in committed founder capital is positioned as seed, not ceiling: the press release explicitly notes the platform is open to additional eligible institutional investors after founder commitments close.

For NVIDIA, the play is downstream demand-pull: every gigawatt of Helix-financed data center is, in effect, a forward order book for DSX-aligned reference designs. For Vistra, it locks in a multi-decade offtake counterparty in a power market where merchant pricing has been volatile but data-center demand is structurally rising. For KIA, it is the kind of long-duration, dollar-denominated real-asset allocation sovereign wealth funds have been chasing as energy revenue is recycled away from concentrated oil-equity exposure.

For the broader capital markets, Helix is the latest sign that the AI capex cycle is moving beyond pure hyperscaler balance-sheet financing. J.P. Morgan argues that 2026 funding needs near $700 billion can still be covered by hyperscaler cash flow and the high-grade bond market, but that 2030 needs above $1.4 trillion will require contributions from “all capital-providing markets” — private credit, infrastructure equity, sovereign wealth, and structured finance included (J.P. Morgan via DCD).

What to watch

  • First announced project. Helix has not named a flagship site. The location, the hyperscaler customer, and the power source on the first project will signal the platform’s true scope.
  • Follow-on capital. Watch for additional sovereign or pension commitments. A second close materially larger than $10B would reset KKR’s positioning against Brookfield, Blackstone and DigitalBridge in the AI-infrastructure capital league table.
  • NVDA and VST as read-throughs. Helix is not a public security, but its order flow is. Watch Vistra disclosures for new AI offtake contracts and NVIDIA’s quarterly commentary on partner-aligned capacity.
  • Grid and permitting friction. Power interconnection queues and regulatory approval timelines are the real binding constraint at scale — capital is necessary, not sufficient.

Bottom line

Helix is a bet that the financing layer of the AI buildout is the next bottleneck after silicon and electrons. It also signals where private equity is taking its biggest swing: not in buying out yesterday’s businesses, but in originating the operating companies that will carry tomorrow’s capital. Whether $10 billion proves to be a down payment on a $50–100 billion platform — or a high-profile launch that struggles to deploy — will depend less on the press release and more on the first three projects Helix actually closes.

Sources

Disclosure: This article was produced with AI assistance and reviewed before publication. It is for informational purposes only and is not investment advice.

Leave a Comment