Structured finance backed by semiconductor chips is emerging as the central funding engine driving global AI expansion, rapidly evolving into a new asset class that could surpass one trillion dollars in size.
The nascent form of this market began in 2023 with a $2.3 billion private credit facility provided by Blackstone to CoreWeave, marking the first landmark deal in chip-collateralized debt.
Today, Broadcom is in talks with Wall Street institutions to arrange $60 billion in chip financing for Anthropic, while SpaceX is reportedly seeking $40 billion in financing for NVIDIA chips, and NVIDIA itself has announced a $500 billion cooperation framework with six major lenders.
JPMorgan analysts forecast that total chip spending could exceed US$3 trillion by 2030, with the majority financed through debt.
However, this debt‑driven expansion also entails significant risks. Because most transactions are private‑placement‑based, investors lack convenient exit mechanisms; moreover, cross‑exposures among chipmakers, cloud‑computing giants, and emerging cloud service providers are highly concentrated.
CoreWeave's corporate‑grade junk bonds have recently plunged, with some falling to 80 cents on the dollar and yields climbing to around 15%, reflecting deep market skepticism about the commercial returns of AI.
A financing round ushers in a new paradigm.
The starting point of this revolution was the Blackstone Tactical Opportunities team's prudent assessment of an unfamiliar concept. According to Brannin McBee, a CoreWeave co-founder who participated in the due diligence at the time, Blackstone executives once directly asked, "What is a GPU?" and "Does AI really exist?"
After receiving a satisfactory answer, Blackstone spearheaded a $2.3 billion private credit deal for CoreWeave in 2023, setting a precedent for chip-collateralized debt financing.
The core logic of this structure is as follows: a special-purpose vehicle (SPV) is established to acquire and hold the chips, with debt service financed by the cash flows generated under contracts with target customers. The debt's maturity is aligned with the chips' five- to six-year depreciation cycle, thereby confining hardware‑depreciation risk within a predictable range.
CoreWeave's model is rooted in commodity market experience—similar to the development approach in the liquefied natural gas sector, which involves "signing contracts first, then financing and building plants." The initial deal was collateralized by a Microsoft contract, with financing costs at the time reaching mid-double-digit interest rates. McBee admitted that this was "the most challenging due diligence he has ever conducted in his career."
As CoreWeave's client base continues to expand, its financing costs have declined accordingly. In March this year, the company secured $8.5 billion in debt financing backed by its Meta contract, bringing the borrowing rate down to around 6%—Wall Street priced the risk based on Meta's high credit rating rather than CoreWeave's own junk rating.
Chip costs far exceed those of data centers.
To grasp the scale of this financing revolution, one must first appreciate the magnitude of chip‑related expenditures. Chip costs can be two to four times the capital required to build the data centers that house them, yet they depreciate completely within five to six years, creating a massive funding gap that traditional capital markets struggle to absorb on their own.
Take Nscale, an AI cloud service provider, and its flagship project in West Virginia as an example: the accompanying power plant is projected to cost $6.5 billion, the data center and related equipment are estimated at $18 billion, while GPU procurement alone amounts to a staggering $47 billion.
Alphabet's Google, Microsoft, Amazon, Meta, and other investment-grade cloud computing giants can leverage their ample cash reserves or low-cost bond financing to cover the aforementioned hardware costs, but OpenAI, Anthropic, and a host of emerging cloud service providers do not possess such strong balance-sheet capabilities.
Mike Liberatore, Chief Financial Officer of OpenAI, admitted: "The scale of our infrastructure build is unprecedented. There's no playbook that tells you how to transition from scratch to general-purpose artificial intelligence in a highly capital-intensive manner using conventional approaches."
To this end, the industry has adopted mechanisms from project finance and structured finance: each transaction's risks are segregated through a special-purpose vehicle (SPV), and debt is raised based on the credit quality of the transaction itself and the borrower's creditworthiness rather than the issuer's own rating, often remaining off the company's balance sheet.
Chip manufacturers are entering the market and lending their endorsement.
Broadcom and NVIDIA's deep involvement have propelled this market to a new scale. The two companies have provided "guarantee" commitments for the related debt transactions:
Once the borrower defaults, the chips will be re-leased or sold to repay the outstanding debt, with the chip manufacturer responsible for covering any shortfall. This mechanism enables the SPV‑issued senior bonds to be linked to the chip manufacturer's investment‑grade rating, thereby significantly reducing financing costs.
Apollo and Blackstone recently jointly closed a $35 billion "Big Sky" deal to support Anthropic. This also marks the first transaction under the "AI XPV" platform co-built by Broadcom, Apollo, and Blackstone—Bank of America strategists estimate that if the platform achieves a financing capacity of 20 gigawatts, it could require $235 billion in backed debt by mid-2029.
In this Anthropic deal, the two senior bonds backed by Broadcom received investment-grade ratings, with the highest tranche carrying a coupon of 5.75%; the riskier $4.5 billion junior tranche was priced at 8.5%—still well below Anthropic's standalone financing costs as an unrated private company.
In August this year, NVIDIA brought together Apollo, Blackstone, Blackrock, Brookfield, KKR, and Goldman Sachs to jointly provide funding for AI infrastructure, with NVIDIA itself committing to underwrite up to 25% of the total.
From Private Placements to the Syndicated Loan Market
To achieve true-scale deployment, chip‑related financing must transcend the private‑equity market and tap into syndicated loan markets and even investment‑grade public bond markets—these latter representing one of the world's largest and lowest‑cost pools of capital.
This year, CoreWeave was the first to close a chip-backed syndicated leveraged loan; both transactions received strong subscription from institutional investors and traded at prices close to par, in stark contrast to the company's corporate bonds.
Rich Myers, Global Head of Structured Products and Private Capital Markets at Morgan Stanley, stated: "These financings are directly linked to the most valuable asset in the market—namely, the computing power generated by chips. A wealth of evidence indicates that demand for computing power will continue to grow, outpacing supply."
Chris Edson, Head of Global Asset Allocation at Apollo, stated that the traditional public debt market alone cannot meet the financing needs of digital infrastructure; instead, it requires the participation of diverse capital sources, including project finance, private placements, insurance funds, and commercial real estate investors. Securing an investment-grade rating is the key to unlocking access to the largest pools of capital, such as insurance funds.
The accumulation of risks cannot be ignored.
However, as capital flows in at an accelerating pace, risks are simultaneously mounting.
Jason Pompeii, a Senior Director at Fitch Ratings, pointed out that the current scale of vendor financing is unprecedented and that risks are highly concentrated. "This will only increase one possibility: once AI ultimately falls short of expectations, the entire supply chain will be hit."
The private‑equity nature of the transaction structure raises questions about liquidity. Since most chip‑financing deals are not traded on public markets, investors find it difficult to exit easily, and the deals tightly tie a handful of chipmakers, cloud‑computing giants, and emerging cloud service providers together, resulting in highly concentrated cross‑exposures. Dave Snyderman, Managing Director at Magnetar, which specializes in credit and fixed‑income products, warned: "The big players will only get bigger. Only they can access financing channels."
The legal validity of customer contracts is likewise in doubt. Akhil Bansal, head of asset-backed financing at Carlyle, stated bluntly: "Everyone assumes these contracts are unbreakable, but I'm focused on the practical reality: if only seven institutions provide all the credit support for GPUs, whose leverage really holds the upper hand?"
Laura Parrott, head of private fixed-income at Nuveen, said she is currently holding off on participating in large investment-grade chip transactions, pending further maturation of the market structure. "This is an entirely new structure, and in the event of a shock, we still don't know how the outcomes will unfold."
In addition, macro-level uncertainties also persist: local opposition in Texas has led to a moratorium on new projects, Oracle's project in New Mexico has been forced to delay payments due to regulatory hurdles, and OpenAI and Anthropic are also discussing slowing the pace of deployment in response to safety concerns.
Moving Toward a "Computing Power Asset Class"
Despite the coexisting challenges, market participants remain optimistic about the long-term potential of this paradigm.
John Watson, Senior Managing Director of the Blackstone Tactical Opportunities team, likened the current moment to the early days of the automotive industry: back then, those who financed Ford's Model T also faced the question, "Will people really drive cars?" Today, auto finance has grown into a large-scale, well-established market.
In terms of the evolution of the chip market, early research and real-world usage data indicate that the actual service life of chips is exceeding the useful lives reflected in current accounting practices, which will enhance their financability. Karen Fang, Global Head of Infrastructure and Sustainable Finance at Bank of America, anticipates that semiconductor valuation tools similar to used-car price guides will gradually emerge, thereby fostering a more active market for chip leasing and hedging.
Watson's ultimate vision is to transform a unit of computing power into an asset class that can be traded and financed globally: "We understand the economic value of computing power, and therefore we finance it."
Achieving this goal still depends on the AI commercialization process truly meeting market expectations.
Editor/Liam