Nvidia is mobilizing Wall Street's heaviest hitters for a $500bn artificial intelligence infrastructure financing package, a move that would fundamentally change how the AI supply chain is funded. Apollo Global Management, Blackstone, BlackRock’s Global Infrastructure Partners, Brookfield Asset Management, Goldman Sachs and KKR are all working with the chipmaker on the deal, according to multiple reports. Sources familiar with the negotiations said an announcement could come as early as Monday, though no final agreement has been reached.
The sheer scope of the proposed package—half a trillion dollars—has sent ripples through financial markets. Nvidia shares fell as much as 3.2% in Monday trading and were down 2.2% in the early New York afternoon, changing hands at $219.01. A deal that should, at first glance, be unambiguously positive for the company instead stoked fear among equity and credit investors. The reason for that fear lies not in the size of the number, but in the structure of the transaction and the troubling precedent it sets.
Almost nothing is settled
Bloomberg’s sources could not say which projects or companies the $500bn funding would back, nor what form the financing would take. They could not even confirm whether the entire amount represents new money. That last point matters more than the headline figure. Nvidia has already announced hundreds of billions of dollars in commitments across the AI supply chain this year. A package that repackages existing pledges is a very different object from one that injects fresh capital into the ecosystem.
Investors have learned to parse the difference between announced frameworks and fully funded projects. Corporate press releases often present memorandums of understanding as if they were binding contracts, and the AI sector has become particularly prone to such inflation. The $500bn figure could be a ceiling, a floor, or a repackaged sum of previous announcements. Until the details are released, the market cannot properly price the risk.
Nvidia's quiet role as the AI industry's banker
For most of this year, Nvidia has been effectively financing its own demand. The company has been in talks over a $250bn backstop for OpenAI to lease compute at a 10-gigawatt Ohio campus. SoftBank subsidiary SB Energy is developing that site. In a separate negotiation, Nvidia has discussed financing around $350bn of OpenAI’s chip purchases. It has expanded its partnership with South Korea’s SK Group to more than $500bn of mutual business, a relationship spanning memory chips, networking and data centre construction. It also made a substantial investment in Ilya Sutskever’s Safe Superintelligence, the AI safety startup that has drawn some of the industry's brightest researchers.
Add it up and the pattern is unmistakable: Nvidia guarantees the customer, the customer buys Nvidia chips, and the revenue eventually lands back on Nvidia’s income statement. This circularity is precisely what unsettles investors. Circular financing describes an arrangement where a supplier funds the buyer that in turn buys from the supplier. Critics argue that this structure inflates demand and valuations across an entire sector, creating a facade of organic growth where none exists. If Nvidia is financing the very customers that justify its own astronomical valuation, then the $2 trillion market capitalisation rests on a precarious foundation.
A familiar charge
The charge has followed Nvidia all year. When the company announced $750bn of deals earlier in 2026, its own credit market flinched rather than cheered. Bond investors, typically more conservative than equity traders, saw the circular dependency as a risk they were unwilling to price in. The credit market reaction was a warning that the company's growth, while real in terms of revenue, was partially self-funded.
Critics have also drawn comparisons to the years before the 2008 financial crisis, when financial engineering made risk appear manageable until it was not. The Bank for International Settlements, the global financial watchdog, has already flagged the resemblance between today's AI financing structures and pre-2008 credit structures. Private credit funds, infrastructure vehicles and complex debt arrangements can obscure where the ultimate risk sits. When Nvidia is both the supplier and the guarantor of demand, the loop is closed in a way that makes traditional underwriting difficult.
Six external balance sheets as an answer
Bringing in six outside balance sheets is a direct answer to that criticism. Spreading the capital load beyond the chipmaker reduces Nvidia's own exposure. It also puts independent underwriters between Nvidia and the projects. Apollo, Blackstone, BlackRock's infrastructure arm, Brookfield, Goldman Sachs and KKR all have deep expertise in funding large-scale physical assets. They have decades of experience structuring deals for power plants, pipelines, ports, and data centres. If they are willing to commit capital, it signals that the projects have been scrutinized by parties with skin in the game.
That answer only works if the underwriting is real. There is a meaningful difference between arranging a financing package and committing one's own capital. The six firms are highly skilled at earning fees from arranging deals that redistribute risk to other investors. Whether they are putting their own money at stake, or merely serving as intermediaries selling debt to pension funds and insurance companies, is a critical distinction. The market will be watching the term sheets closely.
Wall Street was already inside the tent
None of the firms involved in the current talks are strangers to AI infrastructure. Apollo and Blackstone built a $35bn vehicle around Google's TPUs, a custom chip designed specifically for machine learning workloads. Morgan Stanley arranged a $917m loan secured against Lambda's Nvidia GPUs, using the hardware itself as collateral. BlackRock, Global Infrastructure Partners, Microsoft and MGX launched the AI Infrastructure Partnership in September 2024. That vehicle targets $30bn of equity and up to $100bn including debt. Nvidia and xAI joined in March 2025, with Nvidia serving as technical adviser rather than capital partner.
Jensen Huang's vision
Jensen Huang framed that earlier arrangement in characteristically broad terms. "The global buildout of AI infrastructure will benefit every company and country that wants to achieve economic growth and unlock solutions to the world's greatest challenges," Nvidia's chief executive said at the time. The rhetoric matches the scale of the ambition. A $500bn package would be roughly five times the size of the AI Infrastructure Partnership. It would also place Nvidia much closer to the money, not as a technical adviser but as a principal in a massive financial engineering exercise.
Huang has consistently argued that AI is not just a technological shift but an industrial revolution. In his telling, data centres are the new factories and Nvidia chips are the new steel. If that vision is correct, then the world needs trillions of dollars of infrastructure investment, and creative financing structures are essential. If it is overblown, then the same structures become liabilities, channeling capital into assets that may not produce the promised returns.
The scale of the investment wave
Big Tech as a whole is on track to spend more than $730bn on AI this year, according to industry estimates. Microsoft, Google, Amazon and Meta have all signaled that they will continue to invest heavily in data centres and intelligence capacity. Nvidia itself returned to the US bond market in June, its first debt sale since 2021. That move signaled that even the most profitable semiconductor company in history wants to lock in cheap capital before rates shift. Against that backdrop, half a trillion dollars of arranged financing is large but not implausible.
The open question is narrower and harder. The AI buildout requires data centres, power plants, cooling systems, networking gear and, above all, chips. Every major technology company has announced ambitious plans, yet the funding structure behind those plans remains opaque. Banks are eager to collect fees. Asset managers are eager to deploy capital. Nvidia is eager to maintain its dominance. None of these motivations necessarily align with the long-term health of the AI industry.
The $500bn package, if it materializes, will be a landmark in the intersection of technology and finance. It will dwarf previous deals, reshape the balance sheets of the firms involved, and set the tone for how AI infrastructure is funded for the rest of the decade. But until the details are released—which projects, what terms, how much new money—investors will be justified in treating the headline number with suspicion.
Nvidia has been carrying the buildout on its own shoulders for most of the year. The six financial firms could take some of that weight, but only if they are truly underwriting the risk. The market will be watching closely to see whether this is a genuine diversification of risk or just a more sophisticated form of the circular financing that has haunted the AI trade from the start. If the data centres get built and the demand does not follow, somebody eats the loss. Nobody outside the room yet knows who.
Source: TNW | Nvidia News