(Editor's Note: This report references slides and discussion presented at a webinar on Aug. 18, 2026, titled "S&P Ratings' Views on Artificial Intelligence & Hyperscalers " . View slides from the event here , and a video replay here . )

This report does not constitute a rating action.

Key Takeaways

• Capital expenditure (capex) by Alphabet, Amazon, Microsoft, Meta, Oracle, and SpaceX will reach $1.3 trillion in 2027, we estimate, a near tripling from two years ago.
• AI ecosystem participants are engaging in circular funding, where the full range of entities involved in AI have dealings with one another; this arrangement is highly constructive as the sector grows, but it could go sharply into reverse if one or more key entity fails.
• While we make analytical adjustments to capture off-balance-sheet financial commitments when assessing ratings on the six biggest hyperscalers, the trajectory of debt-financed AI investments and the still unclear return on these investments adds to the credit risks of AI participants.
• The strong credit ratings on the six hyperscalers reflect our views that they retain highly profitable core businesses with very predictable revenue and cash flow, offering them capacity to take bets on emerging technologies. As capex rises, this balance sheet capacity is rapidly shrinking.

The credit quality of hyperscalers is gradually weakening. Every time we take a deep dive into this sector, we find that capex is rising faster than we anticipated, financings are becoming more complicated and less transparent, and that returns on investment will take years to realize.

The stakes are high. We assume the top six U.S. hyperscalers will spend more than $7 trillion on data centers and AI-related capex from 2025 through to 2030.

The spending has been an increasingly important component of the U.S. economy and markets. Over the past year about half the growth in GDP from the U.S. private sector was for high-tech/AI-centric activity, by our estimate. The top six hyperscalers have raised more than $400 billion of debt and equity in the year to date, largely to fund AI.

Suffice to say, the ability of entities to convert their investment in AI to profits will say much about the health of the U.S. economy and markets.

The scale of investment requires an increasingly diverse and resourceful approach to funding. These span internally generated funds (all six hyperscalers), U.S. and foreign bond issuance (Alphabet Inc. and Amazon.com Inc.), equity issuance (Alphabet, Oracle Corp. and Space Exploration Technologies Corp. [SpaceX]), private credit and private equity (Meta Platforms Inc. through joint ventures, plus others through guarantees).

An emerging web of interconnected financings is supporting demand, and not just raw end-market economics. It also increases the chances that the failure of an unrated entity could dent the credit standing of a highly rated firm.

Hyperscaler Risk No. 1: The Scale Of Investment

Quantity can have a quality all its own. The particular issue that gets our attention is hyperscalers' path to monetization; as entities spend more, the route to offsetting this expenditure with cash inflow gets pushed out.

While we recognize the massive growth potential, we also recognize higher debt leverage in the AI ecosystem adds to risks. If AI monetization lags, credit metrics and ratings may be pressured. Given the rapidly evolving AI industry dynamics, the question isn't just whether AI creates value, it's whether that value arrives fast enough, and in sufficient quantity, to justify the current pace of investment.

The big six hyperscalers don't quantify their returns on investment on AI. This could be because it’s difficult to isolate and quantify --some investments are being integrated into existing products, while others are research and development for large language models and aren't yet monetizable. Or simply, entities just choose not to share that data.

Amazon chief executive Andy Jassy stated on the most recent earnings call that the company is not investing huge sums "on a hunch." He said the payback period for AI data center hardware is less than three years, while their lifespans are about five to six years, and AI contracts typically last at least five years--implying a positive return on investment.

The gap has been growing between the growth rate of AI capex and hyperscalers' revenue gains. We expect hyperscalers' investment rationale to be based on long-term sustainable growth. However, at least in this current phase, AI participants are constantly evaluating where their sustainable returns on investment will come from across business models. It remains hard to predict sustainable growth and profitability with high confidence.

The firms that don't manage this transition and let their credit metrics weaken beyond what’s appropriate at their current ratings may face rating downgrades.

From a credit rating standpoint, we don’t believe companies have to win the AI wars, so long as they don't blow up their balance sheets in this competition and their existing core businesses remain competitive and profitable.

For now, hyperscalers view the risk of underbuilding to be greater than the risk of overbuilding. Aggressive buildout could lead to overcapacity if demand weakens. This risk can be difficult to manage because data center capacity contracts with customers, (e.g. AI labs) are often short-term--typically three to five years--while the useful life of a data center is typically greater than 25 years.

This mismatch between lease terms and asset life can create re-leasing risk for data center owners in transactions where debt doesn’t fully amortize during the lease tenor. A reliance on lease renewals would raise credit risks if market conditions change.

Hyperscalers are already seeking to farm out some of their excess capacity. SpaceX is leasing capacity to Anthropic and Google; and Meta has suggested it could do the same.

Additionally, OpenAI and Anthropic have huge computing needs; both are important customers for the hyperscalers. Both also aim to IPO over the next year or two, and the capital raising will continue from there. The capital intensity required to satisfy their growth ambitions will likely require continuous fundraising in the foreseeable future.

Commoditization is a threat. We are seeing open-weight models, from both the U.S. and China, rapidly closing the performance gap with the most expensive frontier models. For many enterprise customers, a "good enough" model at a much lower cost is often the better business decision. The emergence of open-weight Chinese models is already inducing OpenAI and Anthropic to cut token prices.

We recognize that enterprise adoption will take time, and the current gap between AI capex and enterprise spending does not necessarily indicate that investments in frontier models are poor investments. The monetization timeline for the cloud computing giants--that is, Amazon, Microsoft and Alphabet--in the late 2000s provides a good example.

That said, we acknowledge that the increasing adoption of open-weight models could prompt frontier model providers to lower prices to remain competitive. This could result in higher usage and, consequently, greater inference computing needs, but potentially lower training computing spending.

OpenAI said at the end of July that it was cutting token prices on its benchmark GPT5.6 Luna model by 80%. About one week earlier Anthropic launched Claude Opus 5 as an alternative to its most powerful model, Claude Fable 5, but at thehalf the priceof Fable. Both labs could have cut prices partly in response to the arrival of capable, low-cost Chinese models such as Moonshot's Kimi K3 and DeepSeek's V4 Flash.

We assume the U.S. AI labs are improving profit margins on their AI services. However, they could still be operating at a loss if fully loading noncash compensation and other adjustments required to conform with GAAP reporting. We anticipate competitive strains will intensify, monetization potential could shrink, and returns on investment for these model developers will likely continue to come under pressure.

Hyperscaler Risk No. 2: Counterparty Risk And A Lack Of Transparency

The scale of investment hangs over this segment. Moreover, few understand the extent to which hyperscalers may be overinvesting. Players are engaging in complex financial arrangements, many of them off-balance sheet without sufficient transparency. Our view is that such arrangements are, at least partially, akin to debt and add another layer of risk on top of the significant capex.

These off-balance-sheet transactions by highly rated hyperscalers and chipmakers typically involve providing capital or financial support to speculative-grade or unrated entities, such as the AI labs or "neocloud" entities (specialist firms that rent out data center capacity).

Indeed, the intention typically appears to be to "lend" the investment-grade rating of a hyperscaler or AI chipmaker to a speculative-grade or unrated entity, but without adding debt to their balance sheets. This approach bridges the funding gap until the unrated entities achieve the scale and standing to access capital on their own merits, at a low cost.

Our job is to break down each of the dealings and understand the credit implication for the rated entities. The transactions are complicated; each typically must be disassembled, often through a scrutiny of available information or disclosures and calls to management.

So long as the AI sector continues to grow, such circular financing can be highly constructive. The arrangements provide hundreds of billions of dollars to high-growth entities that would otherwise be largely locked out of conventional funding or would face significantly higher financing costs.

If conditions push an AI lab or prominent neocloud to fail, circular funding arrangements could drop steep liabilities or unpaid debt with the rated entity.

Our analysis goes beyond reported debt. We examine long-term contractual commitments, lease obligations, take-or-pay agreements, use of special purpose vehicles, guarantees, and residual value support. Such dealings may create debt-like risk even when they do not appear as traditional debt on the balance sheet.

The message for credit investors is that these arrangements are potential liabilities that must be reflected in our assessment of creditworthiness, quantitatively or qualitatively.

These are main types of circular financings used in the AI sector:

Residual-value guarantees: These are obligations in which a hyperscaler provides a guarantee to an unrated entity that data center infrastructure will meet a valuation threshold over a period, typically coinciding with the maturity of a debt financing. If the asset falls below that level, the hyperscaler pledges to make up lost value. This helps the data center operator borrow against the assets (see graphic below). We apply a debt-like adjustment if the guaranteed value exceeds the S&P stressed valuation of the underlying assets.

Chip financing: Unrated entities often receive financial support to buy AI chips, typically in the form of residual value guarantee provided by investment-grade chipmakers. We calculate the chipmaker's exposure under such guarantees by comparing the amortized loan value to the carrying value of the AI chips. Our depreciation methodology will likely be more conservative than chipmakers'.

Lease guarantees for third-party leases: We include these in our debt measure if the ultimate beneficiary of the guarantee is not investment grade, and don't include them if it is investment grade.

Power purchase agreements: This is a contract between a power producer and a data center owner or operator to buy electricity at agreed terms over a fixed period. Under certain circumstances, we may make a debt adjustment for the fixed and minimum component of these agreements by accounting for likely data center capacity, utilization rates, energy prices, etc., or based on more specific information from management.

Lease liabilities: We include lease liabilities in our debt metrics, recognizing their substantial long-term financial implications.

Backstop guarantees: We treat arrangements such as guarantees for unpurchased cloud capacity similarly to lease guarantees; we include these in our debt-like adjustments.

Equity investments: We view equity investments between supplier and customer as another aspect of the circular financing that could amplify risks.

Interconnected financing structures could amplify volatility if demand weakens unexpectedly. The failure of one entity would have implications for entities that have lent it money, have guaranteed the value of its assets, or are just expecting payment for goods delivered.

The broader question is whether circular financing is creating leverage collectively that is individually manageable but could become highly correlated should demand fall sharply.

If demand disappoints, chip values fall, or data center utilization declines, do the losses sit with the operator, the sponsor, or the guarantor? Also, how would the support mechanisms and their related mitigants effectively materialize?

Our analysis considers the sheer interconnectedness of this segment. Players are invested in each other, guaranteeing their borrowing, and buying each other's goods and services--they will also increasingly be competing with each other.

NVIDIA Corp., in just one example, is building AI models and data centers. Hyperscalers and OpenAI are also designing their own chips to reduce their reliance on merchant AI chip vendors such as NVIDIA and Advanced Micro Devices Inc. Competitive pressure resulting in failed investments could create ripple effects throughout the increasingly interconnected AI ecosystem.

The scale of overlap and interconnectedness is vast. In a downturn, even the best capitalized and most profitable firms may incur substantial pain.

Why We Rate The Hyperscalers As Investment Grade

The above outlines substantial risks. We are describing entities taking trillion-dollar bets on fast-moving technology with an unproven business model. Why, then, are all these entities investment grade?

The essential answer is that they retain highly profitable core businesses with highly predictable revenue and cash flow. Alphabet for example generated about $167 billion in operating cash flow in 2025. The entity has a lot of capacity to take large bets on emerging technology.

Our ratings on all these firms account for their growing AI capex. Their credit metrics are weakening, the pace of which is highly dependent on the gap between the rate of their capex growth and overall revenue and profitability growth.

The outlier perhaps are Oracle and SpaceX, both rated in the 'BBB' rating category, predominantly a result of the relative size of their core businesses to their AI ambitions. Both would have more to lose from a downgrade given their financing requirements and higher borrowing costs.

Demand for AI, chips and compute capacity appears to stay strong through 2027. While we make predictions about the health of AI business models and forecast capex beyond 2027, we recognize the rapidly evolving AI landscape and our forecast could change materially.

A breakdown of the credit standing of the top six hyperscalers

Microsoft's heavy investment in data centers and AI is driving leverage higher, rising from 0.2x in fiscal year 2026 to 0.9x by June 2028. Despite positive free cash flow, shareholder returns push discretionary cash flow modestly negative, reducing financial flexibility at its current rating. While leverage remains low and credit quality strong, execution risks and capital demands could pressure ratings if returns fall short. Microsoft's diversified revenue, strong market position, and prudent financial management help mitigate these risks. Microsoft is the only one of the six hyperscalers that has positive free cash flow throughout our forecast period but modestly negative discretionary cash flow after likely shareholder returns. Their plans to monetize all or a portion of their sizable investment stake in OpenAI post-IPO and investment lock-up period will be an important factor in our rating assessment.

Alphabet's credit risk related to data center and AI investments is mitigated by robust liquidity, further supported by its stakes in Anthropic and SpaceX (not yet reflected in our analysis). Strong demand for computing--driven by cloud services, search, and AI--supports revenue growth, with global data demand likely to rise over 20% annually. While heavy capex continues in the near term, it should level off by 2028, easing pressure on cash flow and leverage. Alphabet's strong market position, diversified revenue, and potential liquidity from strategic investments will help maintain its solid credit profile during this investment cycle.

Amazon's credit risk is heightened by rising leverage, projected at 1.3x this year and 1.6x next year--exceeding our 1.5x downgrade trigger. While revenue growth and EBITDA margins remain reasonable despite elevated capex, the outlook for capital spending beyond 2026 is uncertain. The company's guidance of $220 billion in capex for 2026 contrasts with consensus estimates exceeding $300 billion for 2027, creating uncertainty about cash flow and leverage. Sustained high investment in data centers and AI could further pressure credit metrics, making capex discipline and cash flow generation critical for maintaining our rating on Amazon. Besides their operating performance, their plans to monetize all or a portion of their sizable investment stake in Anthropic or OpenAI post-IPO and investment lock-up period will be an important factor in our rating assessment.

Meta's credit risk is rising as leverage approaches our 1.0x downgrade threshold, with limited financial flexibility at the 'AA-' rating. We forecast two years of negative cash flow, largely due to aggressive investments in data centers and AI. The rise in leverage is driven by significant use of joint ventures and third-party leased capacity; by 2029, our estimate of lease obligations will match Meta's outstanding debt. While we do not expect leverage to exceed the threshold, negative cash flow and high lease commitments may weigh on our rating on the entity.

SpaceX is increasing capex in our updated forecast, reflecting accelerated investments and short-term data center lease deals with Google and Anthropic. While these short-term leases boost near-term revenue, they lack the stability of triple-net lease structures preferred for credit quality. As a result, we expect debt to EBITDA to rise to about 1.4x by 2028, still below our 2.0x downgrade threshold. Despite being free cash flow negative, SpaceX is rated higher than Oracle due to its lower leverage (<2.0x S&P-adjusted), consistent with its 2x-3x gross leverage financial policy.

Oracle's credit risk has increased amid heavy investment in data centers and AI, reflected in our recent downgrade of the entity. The company's aggressive capex and acquisition strategy will likely drive a larger free cash flow deficit in fiscal 2027 and adjusted leverage near our downgrade trigger over the next two fiscal years. Unlike peers, Oracle's higher debt levels and less diversified revenue streams constrain its financial flexibility. Although high remaining purchase obligations portend strong revenue growth, rising leverage and ongoing negative free cash flow will continue to pressure its ratings. Nevertheless, we do expect Oracle's credit strains to moderate in fiscal 2029 as strong top line growth begins offsets stabilizing capex. We reflect this assumption in our stable outlook.

A selection of key slides from the hyperscaler slide deck

Writing: Jasper Moiseiwitsch; Digital design: Victoria Schumacher

No content (including ratings, credit-related analyses and data, valuations, model, software, or other application or output therefrom) or any part thereof (Content) may be modified, reverse engineered, reproduced, or distributed in any form by any means, or stored in a database or retrieval system, without the prior written permission of Standard & Poor’s Financial Services LLC or its affiliates (collectively, S&P). The Content shall not be used for any unlawful or unauthorized purposes. S&P and any third-party providers, as well as their directors, officers, shareholders, employees, or agents (collectively S&P Parties) do not guarantee the accuracy, completeness, timeliness, or availability of the Content. S&P Parties are not responsible for any errors or omissions (negligent or otherwise), regardless of the cause, for the results obtained from the use of the Content, or for the security or maintenance of any data input by the user. The Content is provided on an “as is” basis. S&P PARTIES DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE OR USE, FREEDOM FROM BUGS, SOFTWARE ERRORS OR DEFECTS, THAT THE CONTENT’S FUNCTIONING WILL BE UNINTERRUPTED, OR THAT THE CONTENT WILL OPERATE WITH ANY SOFTWARE OR HARDWARE CONFIGURATION. In no event shall S&P Parties be liable to any party for any direct, indirect, incidental, exemplary, compensatory, punitive, special or consequential damages, costs, expenses, legal fees, or losses (including, without limitation, lost income or lost profits and opportunity costs or losses caused by negligence) in connection with any use of the Content even if advised of the possibility of such damages.

Some of the Content may have been created with the assistance of an artificial intelligence (AI) tool. Published Content created or processed using AI is composed, reviewed, edited, and approved by S&P personnel.

Credit-related and other analyses, including ratings, and statements in the Content are statements of opinion as of the date they are expressed and not statements of fact. S&P’s opinions, analyses, and rating acknowledgement decisions (described below) are not recommendations to purchase, hold, or sell any securities or to make any investment decisions, and do not address the suitability of any security. S&P assumes no obligation to update the Content following publication in any form or format. The Content should not be relied on and is not a substitute for the skill, judgment, and experience of the user, its management, employees, advisors, and/or clients when making investment and other business decisions. S&P does not act as a fiduciary or an investment advisor except where registered as such. While S&P has obtained information from sources it believes to be reliable, S&P does not perform an audit and undertakes no duty of due diligence or independent verification of any information it receives. Rating-related publications may be published for a variety of reasons that are not necessarily dependent on action by rating committees, including, but not limited to, the publication of a periodic update on a credit rating and related analyses.

To the extent that regulatory authorities allow a rating agency to acknowledge in one jurisdiction a rating issued in another jurisdiction for certain regulatory purposes, S&P reserves the right to assign, withdraw, or suspend such acknowledgement at any time and in its sole discretion. S&P Parties disclaim any duty whatsoever arising out of the assignment, withdrawal, or suspension of an acknowledgement as well as any liability for any damage alleged to have been suffered on account thereof.

S&P keeps certain activities of its business units separate from each other in order to preserve the independence and objectivity of their respective activities. As a result, certain business units of S&P may have information that is not available to other S&P business units. S&P has established policies and procedures to maintain the confidentiality of certain nonpublic information received in connection with each analytical process.

S&P may receive compensation for its ratings and certain analyses, normally from issuers or underwriters of securities or from obligors. S&P reserves the right to disseminate its opinions and analyses. S&P's public ratings and analyses are made available on its Web sites, www.spglobal.com/ratings (free of charge), and www.ratingsdirect.com (subscription), and may be distributed through other means, including via S&P publications and third-party redistributors. Additional information about our ratings fees is available at www.spglobal.com/usratingsfees.

Any Passwords/user IDs issued by S&P to users are single user-dedicated and may ONLY be used by the individual to whom they have been assigned. No sharing of passwords/user IDs and no simultaneous access via the same password/user ID is permitted. To reprint, translate, or use the data or information other than as provided herein, contact S&P Global Ratings, Client Services, 55 Water Street, New York, NY 10041; (1) 212-438-7280 or by e-mail to: research_request@spglobal.com.