AI Computing Debt Soars: Can On-Chain Credit Fill the Banking Gap?

AI computing financing is spilling over from banks to on-chain, with GPU collateral loans becoming a window to observe bubbles and opportunities.

Written by: @PinkBrains_io

Compiled by: AididiaoJP, Foresight News

AI Computing Debt Soars, On-Chain Credit Fills Banking Gap

Elon Musk poured cold water on the AI computing boom. In response to Chamath, he stated that the debt market has almost closed its doors to a certain type of company in recent weeks: those whose revenues heavily depend on AI data centers and lack other significant income sources.

SpaceX is an exception. It signed a $40 billion financial lease to procure NVIDIA's Vera Rubin chips, with lead investors including Apollo and Goldman Sachs, expected to be completed by the end of the year. SpaceX claims this is the largest GPU financing in history. It has an investment-grade rating and revenue from both Starlink and rocket launches.

Musk's assessment is more severe: over 200 neocloud companies will face large-scale liquidation and consolidation. Neocloud refers to cloud companies that only rent out GPUs, including CoreWeave, Nebius, Lambda, Crusoe, and hundreds of smaller firms. Amazon, Google, and Microsoft are the

GLX launched on October 1, breaking the one-to-one correspondence between deposits and loans. sUSDai funds a loan, which is nurtured through months of timely repayments before being sold, with cash then used to issue new loans. External lenders can also join at the initiation stage. On October 29, Silver Point Capital and Victory Park Capital, in collaboration with USD.AI, provided a $106 million, four-year loan to NexGen Cloud for 1,728 B300 units in Finland. The cumulative circulation of GLX has exceeded $200 million.

In terms of pricing, the minimum sale price is set at par value of 100, meaning no discount and sold at the original price. Prices above 100.5 are considered the "commercial guidance price," encouraging sales to this level, but it is not a strict minimum price. A price of 99.5 is the floor price, allowed only under one condition: if redemption requests exceed available cash after a 30-day redemption queue, meaning liquidity is tight, then a slight discount sale is permitted.

In terms of revenue, 60% of the proceeds from sales go to USD.AI Inc., which operates GLX, while 40% goes to Yield Reserve. Yield Reserve is a reserve fund used to boost the yield of sUSDai to a target of 7%. Discounted sales do not participate in revenue sharing.

Regarding buyers, they cannot have economic ties with the proposers to prevent self-dealing.

For example, if a $10 million loan is sold at 100.5, the $50,000 profit is split into $30,000 and $20,000. With an initiation fee of 2% to 3% based on USD.AI's quote, re-lending this $10 million could earn $200,000 to $300,000. The profit comes from re-lending, not from the sale price.

Currently, the only loan that has exited comes from Duos Edge AI.

On June 5, 2026, USD.AI issued a $98.1 million, three-year loan to Duos Edge AI, a subsidiary of Nasdaq-listed Duos Technologies, with collateral of 2,304 B300 chips. If the loan defaults, USD.AI can only pursue the SPV and the chips, not Duos itself—thus isolating the risk within the SPV.

In September, another GPU operator, Axe Compute, purchased the entire SPV, paying off approximately $87.8 million in loans. Axe acquired a computing cluster with a signed five-year lease, expecting revenues of $364.6 million.

For USD.AI, this loan recovers most of the funds in less than four months, allowing for continued lending.

V. CHIP Buyback and Scale Ceiling

For CHIP holders, the above business ultimately results in a 10% cut of the interest on sUSDai. The more GLX lends out, the larger the cut.

GLX has two sides. Selling loans takes away interest, and the initiation fee goes to USD.AI Inc.; however, cash returns are then lent out at approximately 12%, rather than sitting at treasury bond rates.

The buyback has already started. The net interest margin of sUSDai, after deducting CHIP Foundation costs, is 100% used for programmatic purchases of CHIP. USD.AI has not set a fixed amount or timetable. The initiation fee remains with USD.AI Inc., and any surplus can be used to buy CHIP; loan sale proceeds are distributed at 60% and 40%. The fee switch for USDai has not yet been activated.

As of October 26, based on the buyback price, the circulating market value of CHIP is approximately $90 million, corresponding to 2 billion units at $0.0448 each.

How much can be bought back in a year depends on the calculation basis: based on the average level of the past 12 months, about 2.5% of the circulating volume can be bought back in a year; at the current rate, about 6.5%.

On October 26, just two days before the buyback routing was initiated, USD.AI Inc. manually purchased about 22 million CHIP for $1 million USDC, at an average price of $0.044756, accounting for approximately 1.1% of the circulating volume.

This purchase is equivalent to running the program for 5 to 6 months based on past metrics, or 2 months at the current speed—indicating a rush to buy.

The long-term goal is for the buyback volume to exceed the reward issuance volume. Emissions have decreased from 3% in the first quarter to 2.25% in the second quarter; once purchases exceed new issuances, rewards will shift to a monthly, income-funded arrangement. Most DeFi buybacks rely on trading fees, while CHIP buybacks depend on GPU loan interest, which can only grow with the scale of loans.

The scale is still very small. In 2026, GPU-supported debt issuance is projected to reach $48.4 billion, with on-chain lenders holding less than 1%. At an approximate 12% loan interest rate and 10% protocol cut, capturing 1% of the market would fund about $5.8 million in CHIP buybacks annually, about 6.5% of the current circulating market value; capturing 5% would be about one-third. USD.AI holds 93% of the on-chain portion, with GAIB, Forum Markets, and Falcon Finance following behind.

VI. Growth Boundaries of On-Chain GPU Financing

GPU loans are viable because the collateral can generate rental income, and rental prices are still rising. However, risks also lie with the assets: GPUs depreciate quickly, and new generations of Nvidia products may simultaneously lower rental and resale prices, while the small neocloud services provided by on-chain lenders are among the most likely to be integrated.

However, there are also many reasons for growth. Banks have largely closed their doors to small operators, and regulatory rules vary by state, giving lenders who can complete loans in a few days a competitive edge. Data center restrictions that may be introduced after the U.S. midterm elections could shift construction towards smaller sites, more suited to rapid lending by clusters.

Demand outside the U.S. is also rising. Nvidia's sovereign AI business revenue has more than doubled this year, exceeding $30 billion, with Canada, France, the Netherlands, Singapore, and the UK being major contributors. A portion of this capacity will be realized through local new cloud providers like NexGen Cloud and Sharon AI, which are currently targeted by on-chain lenders as medium-sized borrowers.

Currently, on-chain GPU finance accounts for less than 1% of the total GPU-supported debt in 2026. It fills the gap that banks are unwilling to cover, rather than replacing banks. The growth of the protocol itself does not equate to an increase in token prices.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.