NextFin News - The credit rating agencies are no longer treating the hyperscaler AI buildout as a manageable growth story. Moody's Ratings warned this week that the sector's spending surge is "unprecedented" and is eroding free cash flow and balance-sheet quality across six of the world's largest technology companies, just as capital expenditure heads toward $785 billion in 2026 and roughly $1 trillion in 2027. Standard & Poor's has already acted, cutting Oracle one notch toward junk and naming OpenAI as a key credit risk. After a decade of rating cloud computing as an asset-light cash machine, the agencies are drawing a line: the bill for the AI boom has arrived, and the companies that treated balance sheets as optional are about to find out which of them can actually pay.
The Numbers Behind the Warning
Moody's research note, released this week, tracks six companies: Microsoft, Amazon, Alphabet, Meta, Oracle and CoreWeave. Direct debt across the group has reached $460 billion, but the larger figure sits off the balance sheet. Data-center lease commitments have ballooned to $1.2 trillion, up from $662 billion in February, and more than $820 billion of that total relates to facilities still under construction. Because accounting rules do not recognize a lease as a liability until the data center is handed over or operating, those obligations do not appear as debt today. Moody's treats them as debt-equivalent liabilities anyway - an explicit attempt to close the gap between reported leverage and economic leverage.
The spending pace is historic. Combined 2026 capital expenditure for the group is projected at $785 billion before reaching about $1 trillion next year. Morgan Stanley puts the 2027 figure for the five largest hyperscalers at roughly $1.1 trillion, or about 3.2% of projected U.S. GDP - larger than the annual U.S. defense budget. CreditSights charts the acceleration: combined capex for the top five hyperscalers rose from about $256 billion in 2024 to an estimated $443 billion in 2025 and roughly $602 billion in 2026, with about three-quarters of the 2026 total going to AI infrastructure. To fund it, companies are leaning on debt, stock sales and off-balance-sheet leases, the same tools that, in Moody's words, "threaten credit quality."
The market has already begun to react. Alphabet reported Google Cloud revenue up 82% to $24.8 billion in the second quarter, yet posted negative free cash flow of $5.9 billion - the first negative free-cash-flow quarter since its 2004 IPO. Shares slipped about 4% after hours despite beating headline estimates. Alphabet has since moved to shore up its funding position, raising approximately $85 billion in June in what was the largest equity offering in corporate history, backed by a $10 billion private placement from Berkshire Hathaway, and issuing debt including a rare 100-year sterling bond earlier in the year. The equity raise is itself a signal: a company that spent two decades buying back shares is now selling them to fund infrastructure.
S&P Global Ratings cut Oracle's long-term issuer rating from BBB to BBB- on July 9, leaving it two notches above junk, and also lowered its short-term issuer and commercial paper ratings to A-3 from A-2. The agency cited deteriorating cash flow and named OpenAI concentration as a key risk: nearly half of Oracle's $638 billion backlog is tied to the AI startup. Oracle's fiscal 2026 numbers, for the year ended May 31, tell the story - revenue rose 17% to $67.4 billion, but capital expenditure jumped 162% to $55.7 billion, producing a free-cash-flow deficit of $23.7 billion. S&P now expects that deficit to widen to $42 billion in fiscal 2027, with adjusted leverage above 4x for the next two years. The agency also spelled out the tripwires for another cut: leverage exceeding 4.5x, no path to positive free operating cash flow by fiscal 2029, or a negative view of Oracle's AI strategy or the AI industry as a whole.
From Asset-Light to Asset-Heavy
The core of the rating agencies' concern is a change in the business model itself. For two decades the investment case for big technology rested on asset-light economics: software, intellectual property and scalable cloud services requiring modest capital. That model is gone. Capital intensity has reached levels previously unthinkable for software-era companies - 45% of revenue at Microsoft and 57% at Oracle in the most recent quarter, according to CreditSights. At those levels, technology companies look less like software businesses and more like utilities or telecom operators - capital-intensive industries where credit ratings have always mattered more.
"Previously, these companies relied on asset-light structures centered on software, intellectual property, and scalable cloud services that required modest capital investment," Moody's said in its note. "The transition from asset-light to asset-heavy..."
The mechanism is straightforward, and it is why the warning matters. A data-center lease signed today creates a real economic obligation, but under accounting rules it does not hit the balance sheet until the facility is operating. That delay lets reported leverage look benign while the true commitment grows. Moody's calculation that lease commitments reached $1.2 trillion, with $820 billion not yet commenced, is the agency's way of saying the published balance sheet no longer shows the full picture. When the leases begin, the rent comes due regardless of whether AI revenue has caught up.
The accounting gap runs deeper than leases alone. Companies are using residual value guarantees - promises to cover the difference if a data center's market value falls below an agreed threshold when a lease ends or is cancelled - to give landlords the security they need to build. Because the payout is deemed "not probable," no liability is recorded. Meta, for example, disclosed leases commencing in 2029 with an initial commitment of about $12.3 billion alongside a residual value guarantee with a threshold of $28 billion, with no liability booked. An investigation by Nikkei estimated that off-balance-sheet obligations for five of the firms have surged roughly eightfold in four years. These structures are legal, but they leave the rating agencies - and investors - estimating risk that the financial statements do not show.
The funding shift is visible in the bond market. Neuberger Berman notes that hyperscaler, data-center and semiconductor financing reached approximately $165 billion before the midpoint of 2026 - more than the roughly $138 billion issued across all of 2025. The five largest hyperscalers sold about $121 billion of U.S. corporate bonds in 2025, more than four times their annual average over the preceding five years, and technology has climbed to approximately 10% of the main U.S. investment-grade corporate bond index, sitting alongside financials as one of its largest sectors. Before this cycle, the five largest hyperscalers represented close to 20% of global equity market capitalization but only around 3.5% of public investment-grade debt. That gap is closing fast, and with it the sector's historical exemption from serious credit scrutiny.
The Circular AI Ecosystem
Moody's flagged a second, subtler risk: structural circularity inside the boom. Hyperscalers have invested billions in pre-IPO AI laboratories such as OpenAI and Anthropic. Those labs, in turn, sign multibillion-dollar contracts to buy cloud capacity from the same companies. Part of the backlog that hyperscalers report - and that investors treat as evidence of durable demand - comes from customers whose ability to pay depends on the same AI boom financing.
This is not a peripheral concern. Oracle's $638 billion backlog, which S&P cited when it cut the rating, is nearly half attributable to OpenAI. If AI demand slows or a key customer's ability to pay shifts, the contracts that justify today's spending become stranded assets. The overlapping relationships mean the industry's largest companies are increasingly dependent on the same customers, the same financing arrangements and the same assumptions about future demand. Concentration of that kind is exactly what credit analysts are trained to discount, not to take at face value.
The risk is concentrated at the margin. Moody's was careful to separate the group: Microsoft, Alphabet, Amazon and Meta retain among the strongest corporate balance sheets in the world and face no imminent downgrade threat. The immediate pressure sits with Oracle, rated Baa2 with a negative outlook, and CoreWeave, rated Ba3 in the high-yield market and reliant on complex private-debt structures to finance its GPU fleets. The rating agencies are not calling for a broad downgrade of big technology. They are drawing a line between the cash-rich incumbents and the highly leveraged builders, and warning that the whole sector's financial profile has structurally changed.
The Counter-Case: Early, Not Right
The strongest argument against the bearish read is that the rating agencies have been early before. Demand for AI computing remains robust, cloud businesses continue to grow, and hyperscalers have signed hundreds of billions of dollars in long-term customer contracts that should provide predictable revenue. AllianceBernstein's 2026 credit outlook notes that the investment-grade universe has seen more upgrades than downgrades since the third quarter of 2022, and that it is "too early to deem AI a smashing success or a bubble about to burst." Neuberger Berman points out that the highest-rated names retain the balance-sheet strength to absorb the spending, even though aggregate free cash flow for the five largest fell about 24% in 2025 versus 2024.
The counter-argument has force. A rating warning is not a downgrade, and a downgrade is not a default. The four largest hyperscalers can fund this cycle from operating cash flow and cheap debt in a way that a highly leveraged specialist cannot. The bond market has so far shown a considerable capacity to digest the supply, and strong demand at spreads wider than these issuers have commanded historically has provided attractive entry points for investors. Credit quality within the cohort is generally high: outside Oracle and SpaceX, which both carry BBB ratings, the remainder of the major hyperscalers sit in the single-A to double-A range.
But the counter-thesis answers the wrong question. The issue is not whether Microsoft or Alphabet survive the next two years. It is whether the sector's return on invested capital can recover once the buildout slows, and whether the leases signed today will look cheap or expensive in 2029. The rating agencies are not predicting defaults. They are saying the credit profile has changed permanently, and that investors should price it that way. The direction of travel is what matters: capital intensity is rising, free cash flow is falling, and off-balance-sheet obligations are growing faster than reported debt. History is not reassuring here - the last time a group of highly rated issuers transformed themselves from low-capital to asset-heavy operators on the back of a technological revolution, the telecom sector of the early 2000s, the credit cycle ended in downgrades and defaults on a scale the agencies had not anticipated.
Second-Order Effects: Where the Stress Travels
The immediate consequence of a rating warning is a higher cost of debt, but the second-order effects travel further. First, the bond market itself is being reshaped. Technology has become one of the largest sectors in the investment-grade index, which means index-tracking funds and mandated buyers must absorb whatever the hyperscalers issue. That technical support is real, but it is also mechanical - it does not distinguish between a well-funded incumbent and a leveraged builder. If downgrades begin, forced selling by mandate-bound investors could amplify moves in both directions.
Second, the stress travels to the financiers of the buildout. Private-credit funds have lent against GPU fleets and data-center projects on the assumption that hyperscaler demand is durable. If lease commitments become stranded, those loans sit on assets with no tenant. The same is true for data-center developers who have raised equity, bank loans and securitized bonds against pre-leased capacity. A slowdown in AI demand would not just hit the hyperscalers; it would hit the entire capital stack that was built to serve them.
Third, there is a feedback loop between the credit market and the equity market. As the cost of debt rises and equity issuance becomes the funding source of last resort, the returns that justify today's valuations become harder to reach. Alphabet's $85 billion equity raise - dilutive in a company that defined the buyback era - is the clearest sign that the funding model is shifting. The rating agencies are not causing this shift. They are documenting it, and in doing so they are removing the benefit of the doubt that the market has extended to the sector for years.
What Comes Next
The beneficiaries and the exposed are now clearly separated. The four mega-cap hyperscalers with fortress balance sheets are likely to absorb the scrutiny without rating action, though their cost of debt may edge higher as investors demand a premium for asset-heavy risk. Oracle and CoreWeave carry the near-term downside. More broadly, the warning ripples into the suppliers and financiers of the buildout - chipmakers, data-center developers and the private-credit funds that have lent against GPU fleets.
In the short term, expect more rating-agency research notes and possibly negative outlook changes for the marginal names. Over the medium term, the decisive variable is monetization: whether AI cloud revenue grows fast enough to cover the rent on $1.2 trillion of lease commitments. Over the long term, the sector has crossed from asset-light to asset-heavy, and credit analysis will treat it accordingly.
The base case is that the big four hold their ratings, the marginal names face further pressure, and credit spreads widen modestly. The upside case is that AI revenue accelerates faster than capital expenditure, free cash flow turns positive again, and the warning is remembered as an overreaction. The downside case is that AI demand disappoints, lease commitments become stranded, and downgrades spread beyond the marginal names.
One signal would falsify the structural-deterioration thesis: if the four largest hyperscalers report positive free cash flow for four consecutive quarters while capital expenditure stays above $150 billion each per quarter, the buildout is being absorbed without lasting damage to credit quality. Until then, Moody's has set the new standard plainly: "Investors will increasingly focus on these companies' ability to realize an adequate return on investment." The era of funding first and asking questions later is over.
Data as of July 24, 2026. Moody's note released the week of July 22, 2026; S&P action on Oracle dated July 9, 2026; Alphabet Q2 2026 earnings reported late July 2026.
Explore more exclusive insights at nextfin.ai.