Written by @_kabat_
The Infrastructure Gap: Financing the AI Frontier
In 2025, Meta, Google, and Oracle, collectively issued over $90 billion in debt — a dramatic acceleration from prior years' issuance by these companies. For decades, hyperscalers funded their buildouts from operating cash flow, treating the bond market as a convenience rather than a necessity. That era ended. The AI infrastructure buildout has crossed a threshold where even the most cash-rich balance sheets on earth cannot self-fund, and what began as a technology capex story has become a credit market event with implications extending far beyond Silicon Valley.
Simultaneously, the same forces reshaping traditional fixed-income markets — a shift from credit scarcity to a buyer's market, a pivot toward collateral-based lending, and widening K-shaped dispersion between economic winners and losers — are accelerating the adoption of onchain lending by regulated financial institutions. The AI financing gap creates demand for more credit and, critically, for new credit rails capable of handling the volume, speed, and asset types that the buildout requires.
Three of the world's largest alternative asset managers have independently arrived at the same diagnosis. KKR calls it "High Grading." Apollo describes a "Buyer's Market." PIMCO frames it as a "Compounding Opportunity." All three describe a credit regime where selectivity has replaced beta as the source of returns and collateral-based cash flows have become the winners. When institutions managing trillions in combined assets converge on the same conclusion through independent analysis, the signal warrants examination.
What follows explores what happens when the AI financing gap and the maturing of onchain lending pull two historically separate worlds onto the same rails.
Bridging the $2.7 Trillion Financing Gap
Macroeconomic Neutrality and the K-Shaped Economy
By late 2025, the Federal Reserve had completed its pivot to neutrality — cutting the policy rate by 1.75 percentage points since mid-2024, halting quantitative tightening in December, and initiating reserve management Treasury bill purchases. Fed Vice Chair Jefferson characterized the stance as "cautious optimism" on January 16, 2026: a baseline where US GDP trends near 2%, unemployment sits at 4.4%, and inflation's last mile is complicated by tariff pass-through effects that the Fed treats as a one-time price-level shift rather than entrenched persistence.
Beneath aggregate resilience, a K-shaped economy — where high-earners thrive while low-earners struggle — has hardened into a durable feature rather than a cyclical anomaly. Apollo data shows the top 10% of consumers are now driving nearly half of all spending, while lower-income real wage growth has decelerated to 1.4% year-over-year. KKR documents a stark performance gap: S&P 500 inflation-adjusted revenue per worker has risen 5.5% since ChatGPT's launch, while for the S&P 600 — representing smaller, less tech-integrated companies — it shows the opposite trend. PIMCO observes that AI-related wealth effects sustain asset-holder consumption, while high costs are crowding out housing and other investments for those without significant portfolios. The post-covid recovery is real, but narrow — and this narrowness matters for credit dispersion, the widening performance gap between healthy and struggling borrowers.
The Hyperscaler Funding Gap
Since 2023, hyperscaler capital expenditure has nearly tripled. Institutional research from major investment banks aggregated by Apollo, project cumulative AI-related spending exceeding $2.7 trillion between 2025 and 2029, while KKR data shows capex now consuming 60-70% of operating cash flows and approaching 78% by late 2026. As capex consumes an ever-larger share of operating income, the AI investment cycle's transformation from a self-funded story into a broad-based financing event has become the primary catalyst reshaping supply dynamics.
At the heart of this buildout sits a timing mismatch of extraordinary scale. The massive upfront cash spent for chips, servers, data centers, and energy are front-loaded, while the actual profit from AI services is expected years from now. Internal cash flows cannot bridge the gap, forcing hyperscalers into public debt markets at unprecedented scale and through off-balance-sheet structures like Meta's Beignet SPV, which finances data center capacity through asset-level private debt invisible in public issuance statistics.
According to Apollo's projections, funding just 20% of AI capex through investment-grade markets would propel Amazon to the third-largest issuer of corporate bonds in the index by 2030 and move Google from 67th to 8th. Five companies concentrating index exposure around a single correlated AI trade would create a diversification mirage — portfolios that appear spread across technology, utilities, real estate, and industrials running identical directional risk.
Buyer’s Market
Combined with a reacceleration in M&A activity — North American deal volume reached $2.4 trillion in 2025, the highest since 2021 — AI-driven supply has shifted credit markets from a seller's regime defined by scarcity to a buyer's market defined by selection. Pricing power has returned to lenders for the first time in years.
Credit dispersion — not distress — is the defining mechanism of this regime. Aggregate high-yield spreads (the average risk premium investors demand to hold junk bonds - rated BB+ down to D) were essentially unchanged through 2025, yet CCC spreads (the extra interest paid by the distressed, riskiest companies) widened roughly 85 basis points from 550 bps to 635 bps. KKR frames the selection logic using an education analogy: "a B is an A in this environment, but a C could be an F." Aggregate metrics mask a fork in the economy that rewards precision and punishes undifferentiated exposure.
Against this high-supply, high-dispersion backdrop, the three institutional frameworks converge on a single strategic conclusion: collateral-based cash flows — secured against specific, identifiable assets with clear income profiles — offer superior risk-adjusted returns compared to unsecured corporate credit. KKR identifies a $7.7 trillion investable universe in Asset-Based Finance (ABF). Apollo favors residential mortgages and NAV-backed capital solutions. PIMCO explicitly recommends secured lending in areas such as ABF, real estate credit, and well-structured infrastructure debt. The consensus on collateral-based lending creates a natural bridge to onchain finance. The same logic — securing loans against specific assets with programmable enforcement and continuous monitoring — can be implemented through smart contracts rather than traditional legal documentation, at lower friction and with around-the-clock availability.
DeFi as the New Credit Backend
Modular Rails Driving Institutional Adoption
In January 2025, Coinbase launched crypto-backed loans powered by Morpho, a decentralized lending protocol on the Base Layer 2. Millions of users can borrow USDC against Bitcoin collateral within the Coinbase app — a product where the user sees a familiar fintech-like interface while the liquidity and lending logic execute onchain.
Morpho's positioning as what the company calls the "internet protocol for lending" — a neutral, immutable base layer upon which institutions build customized lending experiences — mirrors the modular logic that traditional asset managers seek. Institutions retain control over risk parameters and user experience without ceding those parameters to a third-party intermediary. Morpho provides the rails; the institution provides the product.
During 2025, Morpho's growth trajectory offered a concrete measure of institutional adoption velocity: users grew from 67,000 to over 1.4 million, deposits expanded from $5 billion to $13 billion, real-world asset deposits climbed from nearly zero to $400 million by Q3, and the integration list expanded beyond Coinbase to include crypto.com, Gemini, Bitget, and Société Générale Forge. The last entry warrants emphasis: a major European bank's digital asset arm participating directly in decentralized lending infrastructure.
Morpho V2, under development for over a year and described as the company's core execution priority for 2026, aims to externalize rate pricing in addition to risk. Where V1 embeds formula-driven rates in the protocol, V2 would enable market-discovered rates where participants express views on risk and return directly. Such an evolution would make onchain lending markets significantly more comparable to the credit markets that KKR, Apollo, and PIMCO analyze — less algorithmic, more reflective of genuine price discovery.
Regulatory Clarity Helping Scale Onchain Lending
In late 2025, the SEC officially closed its four-year investigation into Aave — the dominant DeFi lending protocol with approximately 61% active loans market share — without enforcement action. The investigation had included a Wells Notice, and its resolution without charges removed the existential regulatory uncertainty that had kept many institutions on the sidelines. Aave illustrates a different facet of the integration: scale legitimized through regulatory process.
Aave's operating figures provide context that most fixed-income analysts have not yet absorbed: cumulative deposits since launch exceeding $3.33 trillion, close to $1 trillion in loans issued, approximately $885 million in 2025 fee revenue. Measured against traditional finance, those figures place the protocol in the range of a mid-size regional bank's lending operation — executed entirely through smart contracts without a traditional balance sheet.
Aave's 2026 roadmap targets what its founder describes as "the next trillion dollars" in onchain assets through three pillars: Aave V4 — a unified liquidity layer eliminating cross-chain fragmentation; Horizon — permissioned pools where institutions transact with KYC/AML-compliant counterparties using smart contract efficiency; and a simplified consumer-facing application. Horizon directly addresses the compliance requirements that traditional asset managers identify as prerequisites for institutional credit allocation: counterparty verification, regulatory traceability, and permissioned access within a programmable lending environment.
These two examples, together with growing institutional demand for DeFi seen in strategic directions that Sky and Euler are taking, demonstrate that the overlap between traditional and onchain finance is an observable process rather than a speculative projection. Euler has pivoted toward becoming the 'credit layer' of the internet, leveraging its modular vault technology and ERC-4626 compatibility to provide the predictable risk frameworks and 'invisible infrastructure' required by asset managers. Simultaneously, the Sky Ecosystem has transitioned from purely crypto-native yield to a diversified capital deployment strategy, launching specialized Sky Agents like Grove and Obex to bridge billions in liquidity into real-world verticals, including energy infrastructure and tokenized structured credit. Sky and Euler will be covered in more detail in Stablewatch Memo #2.
Regulated institutions are using DeFi protocols as backend systems for products serving millions of users. DeFi protocols are building compliance layers specifically designed to meet regulatory requirements. And the logic connecting both sides — secured lending against identified assets with programmable enforcement — is the same thesis that the TradFi world has independently endorsed.
Navigating the $1.5 Trillion Revenue Hurdle
The Systemic Credit Risk of AI Spend
Both the TradFi regime change, and the DeFi integration built upon it, face the same fundamental risk: AI investment does not generate sufficient returns to justify its scale. Apollo estimates current AI application revenue at roughly $40-60 billion, while at least $1.5-2 trillion in annual revenue may be needed by 2030 to support projected cumulative AI and data-center capital expenditure exceeding $5 trillion through 2030. A monetization gap of that magnitude represents the possibility that the central assumption driving credit supply dynamics proves wrong.
Should AI capex slow or reverse because returns disappoint, the supply dynamics reshaping credit markets would shift dramatically. The buyer's market thesis depends on continued high issuance. Financing structures built to fund AI buildouts — whether traditional project finance or experimental onchain GPU-backed lending facilities — would face simultaneous stress. A credit supply contraction triggered by AI disappointment would ripple through both traditional and digital lending channels, because both have oriented themselves around the same underlying demand.
PIMCO's cyclical outlook adds a second layer of concern with its warning about later-cycle behavior across credit markets: payment-in-kind provisions masking stress, overreliance on rating agency assessments, vehicles promising more liquidity than their strategies can deliver, and rapid growth in private credit accompanied by weaker underwriting standards. These cautions apply with equal force to DeFi lending protocols that have expanded rapidly during a period of rising asset prices and benign default conditions.
Solving the AI Economy Funding Problem
While the monetization gap presents a systemic risk, a structural counterbalance is emerging through "InfraFi" — onchain ABF specifically engineered for the hardware constraints of the AI economy. Traditional corporate bonds frequently rely on general balance sheet strength, an exposure that ignores the specific lifecycle of the underlying assets. By contrast, "GPU Bonds" utilize smart contracts to create non-recourse, collateral-specific lending. A landmark validation of this model occurred recently with the $500M facility for QumulusAI, structured via the USD.AI protocol. In this arrangement, the debt is secured directly by H100 GPU clusters, with the compute-based revenue flowing through a programmable waterfall that prioritizes senior debt service before any equity distributions.
The utility of USD.AI in these structures solves the critical timing mismatch inherent in high-depreciation tech assets. USD.AI-powered facilities enable rapid, algorithmic amortization that tracks the physical utility of the silicon. This creates a capital-light path for AI firms, unbundling hardware ownership from model operation. For credit investors, it represents the "High Grading" KKR describes: moving away from the speculative beta of AI software and toward the senior secured cash flows of the physical compute layer.
An honest assessment must acknowledge meaningful uncertainty on both sides. TradFi credit dynamics depend on AI economics that remain unproven at the required scale. Onchain lending depends on technical resilience and regulatory stability untested under genuine stress. If either side disappoints, the integration thesis does not merely slow — it potentially reverses.
Asset-Based Finance and the New Credit Stack
Each section of this memo has traced a different thread of the same pattern: an AI financing gap too large for any single channel to fill, a consensus favoring secured lending against identifiable assets, the maturity of onchain protocols into invisible backend for major financial products, and a normalizing regulatory posture toward established DeFi platforms. Taken together, these threads point toward credit formation occurring across a continuum rather than within separate traditional and digital compartments.
In the near term — traditional institutions using onchain protocols as backend while maintaining familiar customer interfaces — is the expansion path already underway. Morpho V2's market-discovered rate pricing and Aave's Horizon compliance layer provide the two missing pieces that institutional adoption has historically required: pricing mechanisms resembling traditional credit markets, and counterparty controls satisfying regulatory frameworks. Together, they lower the integration barrier from experimental pilot to scalable architecture.
Over a longer horizon, as tokenized real-world assets grow beyond their current footprint and the collateral thesis endorsed by KKR, Apollo, and PIMCO finds expression through programmable, continuously auditable structures, the distinction between traditional and digital ABF begins to dissolve. Not because one replaces the other, but because the underlying credit logic is identical — and the most efficient rails will carry the volume. As USD.AI structures specialized hardware debt, the plumbing of the financial system is being re-piped in real-time. The era of easy index returns is giving way to a regime of active selection, where the winning portfolios will be those that own the rails of the AI economy.
A $2.7 trillion financing gap will not be filled by traditional channels alone. Onchain lending built over the past four years was designed for this specific moment. Yet the overlap of massive demand and maturing supply is observable in the data, the integrations, and the independent conclusions of the firms closest to these markets. Where this overlap leads — and how it survives its first genuine stress test — remains to be seen.
Related People, Projects:
@Morpho
@USDai_Official
@aave
@StaniKulechov
@0xKolten
@PIMCO
@KKR_Co
@apolloglobal
@ApolloCryptoFM
@melloninvests
You can see the full article and many more about stablecoins on our blog:
https://app.stablewatch.io/blog
Sources:
https://www.federalreserve.gov/newsevents/speech/jefferson20260116a.htm
https://www.mellon.com/insights/insights-articles/record-breaking-ai-related-debt-issuance-in-2025.html
https://www.pimco.com/eu/en/insights/compounding-opportunity
https://www.apolloacademy.com/2026-credit-outlook/
https://www.kkr.com/insights/outlook
https://www.bis.org/publ/bisbull120.pdf
https://www.spglobal.com/ratings/en/regulatory/article/-/view/sourceId/101651795
https://www.janushenderson.com/en-us/advisor/article/mega-issuance-and-the-ai-arms-race-big-techs-impact-on-credit-spreads/
https://aave.com/blog/aave-2025-recap
https://morpho.org/blog/morpho-2026/
https://x.com/0xJHan/status/2014754594253848955
https://insights.skyeco.com/insights/sky-ecosystem-q4-update-and-2026-outlook-summary
https://www.stablewatch.io/blog/usd-ai-deep-dive
Disclaimer
This material is provided by Stablewatch and the authors for informational/educational use only. Opinions are as of publication. Content may include forward-looking statements and third-party data; accuracy, completeness, and timeliness are not guaranteed. Nothing herein is investment, legal, tax, accounting, or other professional advice, nor an offer, solicitation, recommendation, or endorsement of any asset, protocol, product, or service. Do your own research and consult qualified professionals. Stablewatch and the authors may hold positions or relationships in the mentioned projects and may change them without notice. Subject to applicable law, by accessing you agree to release and hold harmless Stablewatch, its affiliates, contributors, and the authors from any loss or damages (inclusing direct, indirect, incidental, consequential, special, or punitive) arising from use of or reliance on this content or on any projects/products described, including errors, omissions, delays, security incidents, or performance outcomes. You assume all risk. Stablewatch and/or the authors have no duty to update. If you disagree, do not access or rely on this material.
