Wall Street giant Morgan Stanley's senior strategist team, led by Andrew Sheets, recently noted that while the U.S. government has accumulated roughly $40 trillion in federal debt, the rising debt burden has so far failed to halt the historic wave of corporate borrowing from AI-related tech companies, nor has it shaken the resilience of U.S. household spending. In the view of the Morgan Stanley strategists, America's $40 trillion federal debt has not yet triggered a "credit crowding-out" effect on the corporate and household sectors—corporate leverage remains stable with issuance still potentially hitting records, household debt burdens are lower than in 2000 and 2019, and increasingly robust consumer spending and corporate earnings continue to outweigh the negative impact of rising interest rates and steeper Treasury yield curves on U.S. and global equities.

The 6.2% yield on long-term U.S. investment-grade bonds and the roughly 300 basis point premium of 30-year Treasury yields over expected inflation are steadily raising the opportunity cost of owning stocks. What could truly end the "stocks and bonds both strong" dynamic is not the debt total itself, but rather the eventual slowdown in corporate earnings growth driven by persistently high financing costs. The $40 trillion U.S. debt mountain has not yet sealed off corporate credit channels, meaning the AI computing power supercycle currently faces "rising financing costs" rather than "exhausted financing supply."

U.S. corporate debt as a share of GDP is roughly unchanged from a decade ago and remains significantly below pre-pandemic levels, with Morgan Stanley still projecting record corporate bond issuance this year. Household debt stands at about 67% of GDP, nowhere near its limit, and consumption resilience continues to underpin corporate earnings. As a result, hyperscale cloud providers, top-tier AI labs like Anthropic and OpenAI, and AI data center developers led by "new cloud" players are still able to tap investment-grade bonds, project financing, and private credit to tightly link near-limitless future token revenue expectations—along with actual cash flows from AI subscriptions, API calls, and cloud services—to the robust order pipelines across the entire AI computing supply chain, including GPU compute racks, HBM/DRAM/NAND storage components, data center CPUs, optical interconnect systems, power infrastructure, and liquid cooling systems.

Total U.S. debt has surpassed $40 trillion, but it has not yet produced the classic private-sector crowding-out effect. In other words, the $40 trillion debt pile has not ended the AI computing investment boom; instead, it has pushed AI infrastructure construction into a second phase characterized by "credit still open, but capital more expensive," with the next 2-3 years likely to remain a period of high-intensity buildout. With capital markets still open, hyperscale cloud providers can issue investment-grade bonds backed by their balance sheets, AI labs can secure funding through equity capital, compute purchase commitments, and third-party guarantees, while data center developers can obtain project financing or private credit backed by long-term leases, power contracts, and project assets. These capital flows ultimately convert into orders for GPUs, HBM, optical interconnect, power distribution, and cooling systems, with debt repayment sourced from actual cash flows generated by AI subscriptions, API calls, and cloud services—not merely from speculative future token revenue growth expectations.

Another Wall Street bank, Goldman Sachs, projects that large tech companies will cumulatively spend approximately $5.3 trillion on capital expenditures between 2025 and 2030. The share of debt financing in AI capex is expected to rise from roughly 33% in 2026 to 35% in 2027, with direct bond issuance by hyperscale cloud providers alone potentially reaching around $250 billion in 2026, excluding project financing. These data points and AI infrastructure trends indicate that the $40 trillion U.S. debt burden has not yet closed the corporate credit gate, with investment-grade bonds, project financing supported by long-term leases, private credit, and supplier guarantees continuing to jointly drive AI computing asset deployment.

Corporate Financing Wave Shows No Signs of Ebbing: Morgan Stanley Warns the Real Danger for Stocks Lies in Earnings, Not Rates

Approximately half of U.S. federal debt has been added over the past decade, and government debt as a share of GDP has also risen across major global economies. Except for the U.K., fiscal deficits in most countries are expected to remain elevated for an extended period, keeping government borrowing high. Despite rising benchmark rates and long-dated Treasury yields, corporate balance sheets continue to show relative resilience. U.S. corporate debt as a percentage of GDP is little changed from a decade ago and remains well below pre-pandemic levels. Meanwhile, Morgan Stanley's credit strategist team continues to forecast record corporate bond issuance this year. Sheets stated that investors should not expect higher yields to halt the current unprecedented wave of AI-computing-related financing.

U.S. household finances also remain relatively healthy. Household debt currently stands at about 67% of GDP, lower than the roughly 70% seen in 2000 and below the record levels near 74% that emerged in 2019. Sheets pointed out that despite persistently rising rates and long-term Treasury yields, household spending has maintained its resilience. For financial markets, the bigger and more serious question is whether yields and benchmark rates will eventually rise high enough to trigger a large-scale rotation of investor asset allocation from equities toward bonds. Morgan Stanley data shows that the 30-year U.S. Treasury yield curve currently offers a premium of roughly 300 basis points over expected inflation, while long-term U.S. investment-grade bond yields have reached 6.2%.

So far, this kind of capital migration has not occurred on a massive scale. Despite the 10-year U.S. Treasury yield rising about 50 basis points year-to-date, the S&P 500—the benchmark index for U.S. equities—has still surged 13% this year. Since the AI boom erupted in 2023, robust tech earnings growth has helped equities remain competitive in the face of higher bond yields. For the Morgan Stanley strategist team led by Sheets, the most critical question is whether equity market earnings growth can maintain its resilience amid persistently elevated borrowing costs. If earnings growth decelerates significantly in the future, the relationship between rising long-dated Treasury yield curves (10 years and beyond) and overall equity market valuations will become significantly more important to markets.

Token Consumption Poised for 24-Fold Surge: $10 Trillion Institutional Capital Spreading Across AI Computing Bottlenecks

Goldman Sachs' latest $10 trillion institutional holdings report reveals not that "AI conviction" has fully faded, but rather that capital is beginning to rotate from core GPU leaders toward "irreplaceable infrastructure bottlenecks." The Goldman report covers 991 hedge funds with a combined $5.4 trillion in equity positions, as well as 504 large active mutual funds managing $4.6 trillion in equity assets. Both categories of funds have been increasing positions in companies like Bloom Energy, Flex, and Seagate Technology, reflecting a cross-institutional consensus forming around storage product lines, data center power infrastructure, server manufacturing, and other bottleneck segments of the AI computing supply chain.

Nine of the top ten hedge fund holdings are AI-related, with Amazon ranking first for the eleventh consecutive quarter. Mutual funds remain underweight Nvidia and AMD relative to their benchmarks by approximately 100 and 60 basis points, respectively, suggesting that the AI trade has not yet reached full market positioning limits—but any potential increases must be driven by earnings delivery rather than pure thematic enthusiasm. The underlying logic of this rotation and diffusion is AI's evolution from a "core computing chip supercycle" into a full-stack AI computing capital cycle: increases in GPU/TPU/AI ASIC counts simultaneously amplify demand for servers, memory, enterprise NAND storage components, Ethernet switches, optical modules, data center optical communications/interconnects, high-speed connectors, server rack slides, cooling, and power distribution. The larger the training and inference clusters, the higher the port counts, per-rack value, and interconnect complexity.

Morgan Stanley projects that by 2028, nearly $3 trillion in AI-related infrastructure investment will flow through the global economy, with over 80% of that spending still ahead. Goldman Sachs' latest calculations show that the baseline model for global AI capex expects growth from $765 billion annually in 2026 to $1.6 trillion annually by 2031, with cumulative capex of approximately $7.6 trillion from 2026 to 2031. U.S. data center electricity demand is projected to rise from 31 GW in 2025 to 66 GW in 2027, which will directly spill AI infrastructure investment into server CPUs, DRAM/NAND/HBM, advanced packaging, liquid cooling, power equipment, transformers, gas turbines, grid connection equipment, data center REITs, and construction engineering.

On the demand side, the core driver of AI computing needs comes from AI's transition from "answering questions" to "executing workflows": agents require continuous planning, tool invocation, result verification, and failure retries, with token consumption per task potentially reaching 10x, 20x, or even 50x that of traditional chat queries. World models further expand computing boundaries from text to robotics, industrial simulation, and physical systems. Goldman Sachs officially projects that global token consumption could grow 24-fold by 2030, reaching 120 quadrillion tokens per month. Meanwhile, inference token unit costs are declining 60%-70% annually, creating a Jevons Paradox dynamic of "falling costs—expanding applications—growing total demand."

The $40 trillion debt burden has not ended the AI computing investment boom; instead, it has pushed the industry into a second phase of "credit still open, but capital more expensive." With 30-year Treasury yields around 5.23% and long-term investment-grade corporate bond yields at 6.2%, every AI project must now prove that its internal rate of return (IRR) can cover the rising cost of capital. Consequently, subsequent funding will not flow indiscriminately into all concept stocks, but will instead prioritize leaders across the AI computing supply chain that possess long-term orders, free cash flow, pricing power, and critical bottleneck positions—including GPU clusters, HBM/DRAM/NAND storage, data center high-speed networking and optical interconnect, and new cloud providers. If token commercialization continues to deliver, the corporate financing wave will extend the AI capex cycle; if earnings decelerate or long-end yields spiral out of control again driven by "bond vigilantes," highly leveraged new cloud providers and projects reliant on external guarantees will be the first to face valuation and credit double blows.