SemiAnalysis: Not Bearish on NVIDIA, the 'AI Central Bank' Could Leverage a $7 Trillion Debt Snowball

Original Author: Zhao Ying

Original Source: Wall Street News

On July 6th, renowned semiconductor research firm SemiAnalysis posted a series of six tweets on platform X, revealing that NVIDIA's Kyber NVL144 rack is delayed by over 12 months due to manufacturing challenges with the PCB mid-plane. Asian AI hardware supply chain stocks slumped in response.

NVIDIA subsequently responded, stating its "roadmap remains unchanged", but did not disclose specific progress details.

The controversy remains unresolved. On July 7th, SemiAnalysis released another paid long-form article, this time turning its "critical lens" towards NVIDIA. However, this time it didn't play the role of a "bear."

SemiAnalysis predicts: By 2029, the global scale of AI debt financing will exceed $7 trillion. What does $7 trillion mean? It will become the second-largest asset-backed debt market globally, trailing only the U.S. residential mortgage market (approximately $13 trillion).

What is NVIDIA's role in this? SemiAnalysis disclosed a strategic move by NVIDIA—the "backstop plan". NVIDIA is leveraging its AA/Aa2 investment-grade credit rating to provide minimum revenue guarantees for AI compute rental providers, thereby incentivizing bank lending. In other words, NVIDIA is acting as the lender of last resort and insurer for the entire AI ecosystem, booking significant sales while shouldering part of the risk from insufficient downstream demand. SemiAnalysis likens NVIDIA to the "Central Bank of AI."

Regarding the discussion on platform X about whether SemiAnalysis is "bearish on NVIDIA," the firm stated:

It has not published any bullish or bearish views on NVIDIA's stock, but is merely accurately capturing supply chain and technical details. The market can trade based on that information itself.

The AI Debt Snowball: Reaching $7 Trillion by 2029, Approaching the U.S. Mortgage Market

SemiAnalysis believes that AI infrastructure construction is forming a multi-trillion dollar credit market. By 2029, outstanding AI-related debt could reach approximately $7.1 trillion, a scale that would surpass all other U.S. asset-backed debt markets except for the mortgage financing market.

This debt primarily stems from two types of capital expenditure. One is AI IT CapEx, including GPUs, networking, storage, and supporting CPUs; the other is AI Data Center CapEx, including the facilities, power, and cooling infrastructure needed to house these GPUs.

Historically, cloud giants like Google, Amazon, Meta, Microsoft, and Oracle primarily relied on their own cash flow to build AI clusters. However, in the past year, Oracle, Meta, and even Google have started using more debt. As project scales continue to grow, the market constraint is no longer just about getting GPUs or finding facilities, but whether sufficient, cheap, long-term capital can be borrowed.

SemiAnalysis concludes that the financing method for AI CapEx is changing. The balance sheets of cloud giants are not infinite. If all AI clusters depend on endorsements from a few investment-grade cloud vendors, new projects will inevitably face credit bottlenecks sooner or later.

The "Trilemma": Capital, Customers, and Data Centers Are All Indispensable

SemiAnalysis breaks down AI project financing into a "trilemma": Capital, Underwriting Contracts, and Data Centers.

First is Capital. Lenders typically require long-term take-or-pay contracts from investment-grade cloud vendors, or similar credit guarantees, before they are willing to lend. In other words, lenders truly value the credit of the underlying customer, not the Neocloud itself.

Second is Underwriting. To secure customers, a Neocloud often needs to prove it can pay GPU deposits and lock in equipment. But to secure equity funding, it needs to prove it already has customers and loans. This creates a circular dependency for projects in the early stages.

Third is Data Centers. Neoclouds must either use customer contracts and financing to convince data center operators to lease capacity, or build their own data centers. The latter carries greater capital pressure and longer timelines.

This model locks the market into a template of "5-year terms backed by cloud giants." The problem is that many VC-backed AI startups and inference service providers need short-term, massive-scale compute power, not 5-year long-term contracts. Inference service providers, in particular, are unwilling to bear long-term price and demand risks, often preferring to forgo compute power rather than sign leases exceeding one year.

NVIDIA, the "AI Central Bank": Leveraging AA-rated Credit to Stimulate the Entire Market

NVIDIA proposed the "backstop plan" to fill this financing gap.

According to SemiAnalysis, NVIDIA provides a GPU rental revenue backstop to Neoclouds. If third-party customer demand is insufficient, NVIDIA commits to purchasing compute power at a predetermined price; if the Neocloud rents out compute power at a higher price, NVIDIA shares a portion of the excess revenue.

These arrangements are typically for 6-year terms, providing a minimum revenue guarantee for the underlying GPU capacity based on an agreed-upon price curve. The Neocloud can still rent compute power to any customer and offer more flexible lease terms. NVIDIA's backstop is triggered only when market demand is insufficient and capacity cannot be rented at market prices.

This is the origin of the "AI Central Bank" analogy. NVIDIA is not literally issuing currency, but acting as a quasi-buyer of last resort and credit endorser within the AI compute credit system. Lenders can evaluate the worst-case scenario for a project based on NVIDIA's AA/Aa2 credit rating, making them more willing to lend.

For NVIDIA, this helps expand the buyer base for its GPUs. If the market could only rely on a few mega-cloud vendors signing 5-year underwriting contracts, GPU demand would quickly hit financing constraints; and these cloud vendors are also using their own custom chips to hedge against NVIDIA's systems. By supporting Neoclouds and more enterprise customers, NVIDIA opens new financing channels for GPU demand.

Anatomy of the "Backstop Plan": How Much NVIDIA Earns, How Much NeoCloud Earns

SemiAnalysis emphasizes that Neoclouds do not use NVIDIA's credit for free. Under the backstop structure,Neoclouds must sacrifice a portion of their upside in exchange for project bankability.

In an illustrative price curve, the 6-year average backstop price is $2.36 per GPU per hour. Assuming a 1-year lease price for GB300 is $6.75 per hour in the first year, and the first-year backstop price is $3.68 per hour, the difference between the customer price and the backstop price is $3.07. If NVIDIA takes 40% of the amount exceeding the backstop price, NVIDIA gets $1.23, the Neocloud gets $1.84, and the Neocloud's actual first-year revenue is $5.52 per hour, lower than the $6.75 without a backstop.

Over six years in this scenario, NVIDIA takes an average cut of approximately 18%. The Neocloud's project IRR also decreases. With NVIDIA's backstop and primarily 1-year short-term leases, the project IRR is 25.4%; without a backstop but able to secure financing and lease out capacity, the IRR could reach 40.7%.

The key is the worst-case scenario. If demand is insufficient, the Neocloud can only lease compute power back to NVIDIA, resulting in near-zero or slightly negative project returns. Lenders do not require the project to be profitable in the worst case; they only require it to still service its debt. Therefore, whether the debt structure works ultimately depends more and more on the reliability of NVIDIA's backstop.

This is also the core concern for investors: NVIDIA's arrangements help drive GPU sales and Neocloud expansion in the short term, but if compute demand falls short of expectations, the revenue shortfall will be absorbed by NVIDIA. The debt may not be directly on NVIDIA's books, but the safety net of the financing model is increasingly concentrated on NVIDIA's credit.

GPU Financing Pricing Ultimately Depends on Who Provides the Backing

SemiAnalysis states that current pricing in the GPU financing market is primarily determined by who signs the long-term underwriting contracts, not the Neocloud's own credit standing.

CoreWeave serves as a reference. Its 5-year unsecured bond yields approximately 10%; however, for the $8.5 billion DDTL 4.0 delayed draw term loan backed by Meta, the fixed-rate portion costs approximately 5.9%, only 90 basis points higher than Meta's 5-year bond yield of about 5.0%. This 90 basis points roughly represents the market's pricing of CoreWeave's execution risk.

If a Neocloud operates without long-term cloud vendor underwriting contracts, its financing costs increase significantly. For top-tier Neoclouds, unsecured financing might require paying approximately 10% interest, about 4 percentage points higher than backed financing. At a 70% to 80% loan-to-value ratio, financing costs rise from 5.62% to 10%, causing pre-tax profit margins to drop from 14.8% to 5.4%.

NVIDIA's backstop would position pricing between these two levels: higher than the approximately 5.9% all-in yield on current cloud vendor-backed deals, but lower than the approximately 10% yield on CoreWeave's unsecured bonds. The key metric for banks is the Debt Service Coverage Ratio (DSCR). For projects with a NVIDIA backstop, loan sizes are typically calculated based on a scenario where the backstop is triggered, requiring a DSCR of at least 1.3x in the early years, corresponding to a loan-to-value ratio usually between 70% and 80%.

Public Projects Amplified in Asia-Pacific, Backstop Model Begins Implementation

The publicly disclosed NVIDIA backstop projects are currently concentrated in the Asia-Pacific region.

The first is SharonAI's 72MW AI factory in Australia. Announced in June 2026, the project plans to expand to up to 40,000 GB300 GPUs under a 6-year backstop. SharonAI disclosed a total backstop value of $4.88 billion, translating to a six-year average floor price of approximately $2.33 per GPU per hour.

Another is Firmus's 360MW AI cluster in Batam, Indonesia, potentially located within DayOne's Kabil Industrial Tech Park facility. Announced on June 29, 2026, this indicates NVIDIA's backstop is moving towards larger scales.

Firmus estimates the project's six-year customer revenue at $25 billion to $30 billion, targeting AI-native companies, enterprise clients, and inference service providers, offering various lease terms. However, before deploying GPUs, Firmus still needs to finalize the data center provider or continue building its own.

SemiAnalysis also points out that NVIDIA is not the only GPU manufacturer using backstop arrangements. AMD has offered similar arrangements to customers like AWS, OCI, DigitalOcean, Vultr, Tensorwave, and Crusoe last year: customers purchase more AMD GPUs, and if the Neocloud cannot fully sell the capacity, AMD is willing to lease back a portion under a long-term contract for internal software development.

SemiAnalysis Denies Being Bearish, But the Market is More Sensitive to Its Signals

The publication of this article comes amidst controversy surrounding SemiAnalysis itself.

On the morning of July 6th, SemiAnalysis posted a series of tweets on platform X, stating that NVIDIA's Kyber NVL144 rack architecture faced significant delays, pushing it back over 12 months to 2028. This news caught attention before the market opened and caused declines in several AI hardware supply chain stocks in Japan, South Korea, and Taiwan, China. NVIDIA subsequently responded that its product roadmap had not changed, denying that core progress was affected.

This made SemiAnalysis's follow-up article more susceptible to market interpretation as either bearish or bullish on NVIDIA. In response, SemiAnalysis stated on platform X that it had not published any bullish or bearish views on NVIDIA's stock, but was merely sharing the company's supply chain and technical details.

Crackerjack Finance countered the "bearish" interpretation, stating that SemiAnalysis's charts indicated actual data for the second half of the year was 20% higher than market expectations, and derived an estimate of approximately $15 in earnings per share for the next year, suggesting a stock price between $300 and $400. THE Grand Poobah commented that "the three-party circular financing seems insufficient," pointing to market concerns about the increasing complexity of financing structures.

The issue is that AI-related assets have experienced years of gains, with valuations and expectations both at high levels. Any supply chain risk signal or change in financing structure can be rapidly amplified. SemiAnalysis's clarification shows it did not directly offer a stock opinion, but following the Kyber NVL144 incident, the market influence and credibility debate surrounding its supply chain revelations will continue to coexist.

For investors, the true meaning of this "long-form article" is: AI competition is no longer just about "who has the GPUs," but "who can simultaneously piece together GPUs, debt, customer contracts, and data centers." NVIDIA's backstop mechanism may continue to amplify GPU demand, but it could also concentrate more of the tail risk from the AI debt cycle onto NVIDIA's own credit.