Nvidia says Wall Street lined up $500bn to finance AI “compute” infrastructure
Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR are backing data centres and chip factories.
Nvidia said it has assembled $500bn (£370bn) of capital for AI infrastructure with a roster of marquee Wall Street firms, pitching “compute” as a new investable asset class. The company framed the financing as a way to expand data-centre capacity and chip manufacturing as AI spending continues to compound across big tech.
Key Takeaways
- Nvidia said it raised $500bn (£370bn) for AI infrastructure through deals with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR.
- The company framed AI hardware and infrastructure — “compute” — as an investable asset class that long-term capital providers can underwrite.
- The financing is intended to support both Nvidia-built and partner-built projects, including new data centres and new factories to manufacture AI chips.
- Major tech and AI firms have spent over $1tn in three years on AI projects and infrastructure, alongside Nvidia’s market value rising fivefold over the same period.
Nvidia Pitches $500bn for “Compute” — and Names the Backers
Nvidia said it has lined up $500bn (£370bn) in capital for artificial intelligence infrastructure, naming Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR as counterparties. The company’s framing matters as much as the number. It is explicitly trying to turn “compute” into something allocators can underwrite like infrastructure, rather than treating AI buildout as a series of one-off corporate capex decisions.
“Compute,” in Nvidia’s usage, is the full stack that makes AI workloads run at scale: GPUs (graphics processing units), the servers they sit in, and the data centres that provide power, cooling, and networking. Nvidia CEO Jensen Huang put the monetisation claim in plain terms: “In AI, compute is revenue,” he said, adding: “We are bringing the world's leading long-term capital providers together to independently underwrite AI infrastructure.”
For traders, the mechanical read-through is that Nvidia is trying to widen the funding base for the AI capex cycle that has been propping up tech-beta sentiment. If compute becomes a financeable asset class, the marginal buyer of AI capacity is no longer just hyperscalers writing internal budgets. It is also credit and infrastructure capital looking for contracted cashflows.
Where the Capital Is Supposed to Go: Data Centres and Chip Factories
Nvidia said the financing will go toward its own projects and those built by partners. The stated targets are the two bottlenecks that keep showing up in AI deployment: physical data-centre capacity and chip supply.
On the data-centre side, Nvidia described projects that would “house, operate, and cool miles of stacked computer chips” running AI workloads. That is a reminder that AI scaling is constrained by real-world inputs, not just model releases. Power delivery, cooling, and space are the gating items when utilisation spikes.
On the manufacturing side, Nvidia said the financing would back new factories to manufacture AI chips and “increase their availability to buyers.” That is the supply-side counterpart to the demand story that has driven Nvidia’s valuation. The report also noted that essentially every major technology and AI company uses Nvidia GPUs to power services and chatbots, naming Google, Meta, Amazon, Microsoft, SpaceX, Tesla, OpenAI and Anthropic.
The broader backdrop is already extreme. Major tech and AI firms have spent over $1tn in just three years on AI projects and infrastructure, with more spending expected, and demand for Nvidia’s chips and services has driven the company’s market value up fivefold over the same period.
The Bull Case vs. Execution Risk: ‘Compute Is Revenue’ Meets ‘Delivery Is the Hard Part’
The bull case Nvidia is selling is straightforward: compute is scarce, demand is durable, and the cashflows can be underwritten like infrastructure. Apollo president Jim Zelter called it directly: “Modern compute has emerged as a scarce, mission-critical asset class.” He added that it is “positioned to drive significant long-term economic growth and productivity gains.”
KKR’s leadership endorsed the category shift while flagging the operational trap. “Compute has become a critical infrastructure asset,” co-CEOs Joe Bae and Scott Nuttall said, before adding: “As we've scaled our approach to digital infrastructure, we've learned that delivery, not ambition, is the hard part.” That line is doing work. Data centres and manufacturing expansions fail in familiar ways: permitting and grid interconnect delays, cost overruns, supply-chain slippage, and demand forecasts that look clean until pricing compresses.
Public-market investors are already sensitive to the overbuild question because the AI capex cycle has been so front-loaded. Jane Sydenham, senior investment manager at Rathbones, described Nvidia’s centrality to the stack, saying: “Nvidia is absolutely enormous and produces these chips that everybody needs for AI and it needs to keep facilitating the growth of AI,” before landing the key risk: “The worry is that more and more money is going into these projects. Are they all going to earn the right return for the future?”
The report pointed to adjacent deals that fit the same pattern of capital moving into AI infrastructure. BlackRock last month entered an individual deal with Meta to finance and take a majority ownership stake in one data centre in Texas. Anthropic also recently entered into a deal with Macquarie Asset Management and GIC for investment in AI infrastructure, and said more financing was needed because Claude’s popularity means “demand requires significant new compute.”
What Traders Still Don’t Know About the $500bn Figure
The $500bn headline number is directionally clear but structurally vague, and that is the part that will decide how tradable this is beyond the initial narrative. The report does not specify whether the $500bn represents committed capital, a maximum financing capacity, or a multi-year target. It also does not break down the debt-versus-equity mix, which determines both the cost of capital and how quickly projects can be pushed through.
There is also no project-level disclosure: no named builds, no geographies, no partner-by-partner allocation, and no stated underwriting terms or return targets from the participating firms. Without those, it is hard to map the announcement to a timeline for incremental compute supply.
The next set of signals is therefore mechanical, not rhetorical. A clarification on whether the $500bn is committed or merely available capacity, plus a deployment timeline and capital-structure details, would tighten the market’s read. Concrete announcements of funded data-centre projects, including locations and counterparties, would show whether this is a financing framework or an active buildout pipeline. Updates tied to new chip-manufacturing capacity — factory plans and production timelines — matter because Nvidia explicitly positioned factories as part of the use of proceeds. More large financings like the BlackRock–Meta deal and Anthropic’s Macquarie/GIC arrangement would also confirm that “compute as infrastructure” is broadening beyond a single Nvidia-led package.
My Read: ‘Compute as an Asset Class’ Is a Liquidity Narrative With Crypto Spillover Potential
The part that matters here is not the $500bn headline. It is the attempt to standardise compute into something long-term capital can underwrite, which is what Huang is pointing at when he says “compute is revenue” and talks about independent underwriting. If that framing sticks, it extends the AI capex cycle by widening who can fund it, and that tends to support the same risk-on tape that crypto trades against when tech-beta is the marginal driver.
The real test is whether this turns into named projects with timelines and a disclosed capital structure, not just a financing umbrella with a big round number. If the disclosures show committed capital deploying into data centres and chip factories on a defined schedule, “compute as an asset class” starts to look structural rather than narrative-driven, and that is when the spillover into crypto AI and infrastructure narratives becomes more than a sympathy trade.
