Why Hyper scaler CapEx Is Distorting S&P 500 Earnings

Recently, many investors have started to question the circularity of revenues in the AI universe—and rightfully so, in our view.^1,2 As AI Capital Expenditure (CapEx) spending cascades through the system, it generates multiple layers of revenues and earnings across the supply chain, inflating reported results and giving what we believe is an exaggerated sense of profitability. This earnings multiplier effect arises from differences in the accounting treatment of revenue recognition (typically within a few months) versus depreciation (an expense smoothed out over many years). In our view, this cascade of accounting profits across the entire AI supply chain will likely create a significant downward earnings revision cycle when hyperscalers eventually reduce their CapEx plans (arguably preceded by markets becoming concerned with ever-increasing spending plans like we have started to see with Meta following their Q3 2025 earnings results).³

As long-term investors, we believe that understanding capital cycles is one of the most important aspects of investing. Indeed, eventually the longer-term math matters, similar to what we have witnessed (and written about) for a whole slew of software businesses some years back.⁴ As we try to understand the second-derivative implications of increased hyperscaler CapEx spending, one question is: How could this spending impact the earnings of other companies in the S&P 500? Conceptually, here is the dynamic: when a hyperscaler (Company A) spends $100 on CapEx, it records no expense on its income statement other than an incremental depreciation figure, which is spread over many years. For simplicity, assume this incremental $100 CapEx spend by Company A is fulfilled by Company B. This expenditure becomes incremental revenue for Company B, on which it earns a margin that boosts its earnings. In turn, Company B might have its own CapEx outlay that is then fulfilled by Company C, and so on and so forth.

Distorting Earnings Cascade of Hyperscaler CapEx and Venture Capital (VC) Funding

table

Source: GQG Partners LLC (table). *BofA US Equity & Quant Strategy, FactSet; represents aggregated consensus estimates for 2025 across MSFT, AMZN, GOOGL, ORCL, and META (data). **Pitchbook (data). YTD through Q3 2025, venture funds have raised ~$167B, of which ~65% has flowed through to AI start-ups; run-rating the first three quarters yields ~$223B raised by VC and 65% of that to AI start-ups yields ~$145B. ***Bloomberg as of 5 December 2025 (data). Represents the difference between the sum of expected calendar year 2025 operating income and actual calendar year 2024 operating income across current index constituents. Actual results may differ from projections illustrated above.  You cannot invest directly in an index. PAST PERFORMANCE MAY NOT BE INDICATIVE OF FUTURE RESULTS.

Using a stylized model, and based on the assumptions below, we estimate AI-related CapEx will have contributed over a third of this year’s ~$586B growth in operating income for the S&P 500. As shown in the table, start with the roughly known $540B in AI-related spend this year from hyperscalers and venture capital (~$395B in hyperscaler CapEx plus ~$145B in AI start-up investment that effectively drives compute demand). Let us further assume 70% of this spend converts into tier-one supplier revenues (such suppliers themselves being S&P members across semiconductors and networking equipment), earned at 40% operating margins or ~$151B in dollar terms (in our view a conservative figure as AI accelerators/networking names such as Nvidia, Broadcom, and Arista are expected to boast operating margins well above this).⁵

For the second-order effect, we estimate 50% of that first-order revenue (~$189B) at ~25% margins, or roughly $47B dollars, flows into tier-two businesses involved in power supply, cooling, data center construction, among other things (at this level, more of the spend would leak outside to non-S&P foundry/memory names, but much of the economics are still retained by index members).

Taken together, we believe these first- and second-order effects distort index-level operating income through front-loaded revenue recognition and aggressive estimates on the useful life of AI hardware such as servers and chips. We believe this all creates an artificial accounting profit uplift with significant reversal risks as CapEx growth moderates and businesses realize the useful horizon for AI hardware is closer to two to three years and not the five to six many are baking in today.^6,7

Indeed, it is not hard to imagine these dynamics playing out, considering recent Organisation for Economic Co-operation and Development (OECD) reports that absent such large-scale AI-related spending, real GDP growth in the first half of 2025 would have been negative.⁸

Fragile Foundations

While the cascading impact of hyperscaler CapEx appears to have boosted earnings across the AI supply chain, we believe this dynamic is further amplified by the increasing weight of semiconductors in the S&P 500 Index, introducing the potential for substantially greater cyclicality of profits relative to history. Unprecedentedly, semiconductors now account for about 15% of the index, up sharply from their ~4% median weight over the preceding decade, underscoring their critical role as the first-order beneficiaries of hyperscaler AI infrastructure spending.

While this concentration has undoubtedly been a tailwind for index-level earnings during the current CapEx boom, we believe it also creates vulnerabilities. Relative to their long-term revenue trends, global semiconductor sales are far above historical norms (50%+), driven by demand spikes that we think are unlikely to persist indefinitely. As hyperscaler CapEx growth slows, the over-earnings in semiconductors will likely reverse as revenues normalize, exposing the index to heightened earnings growth corrections.

Source: GQG Partners LLC (charts), Alliance Bernstein (data), World Semiconductor Trade Statistics (data). Data as of and through 30 September 2025.

“I would expect that we will see a supply constrained environment for the vast majority of next year is my guess at the moment. But a lot of that has to do with the time of our demand is just too great.” 

Jensen Huang, Nvidia, August 2021

“Hyper cloud and service provider demand continued to be strong…On the supply side, our lead times remain extended and stable. Inventory in our channels and at our customers remains very lean… What we continue to see with the recovery, and I made a point of saying that — and there’s now — we’re now in the midst of a very strong spending recovery in enterprise, particularly, so we’re continuing to see strong demand bookings in the semiconductor side…growth in networking, broadband, server storage lately of some 20% to 30% year-on-year are real, true end demand… going into ’22, firing on all cylinders. And by that, I mean more than just forecasting. We’ve actually seen the backlog. We have the backlog and they keep building up.”

Hock Tan, Broadcom, December 2021

“Our third quarter business was driven by strong demand across all 4 growth platforms… We are entering a period of higher structural growth. The multiyear megatrend of 5G and HPC-related applications are expected to fuel a massive requirement for computation power and propel greater need for energy-efficient computing, which demands the use of leading-edge technologies…we expect TSMC’s capacity remain very tight in 2021 and throughout 2022….is there any possibility of oversupply? Not for TSMC, okay?”

Che-Chia Wei, TSMC, October 2021

In addition to the structural cyclicality inherent in the semiconductor industry, another factor inflating perceived profitability during the AI infrastructure build-out is the exclusion of key costs under Generally Accepted Accounting Principles (GAAP), notably stock-based compensation (SBC/RSUs). It is our view that these metrics further reinforce this over-earnings dynamic; the totals are hardly immaterial, standing at nearly $250B for the most recent fiscal year. While RSUs masquerade as non-cash charges, they effectively represent a portion of total employee compensation expenditure, thus requiring businesses to spend cash on share buybacks to mitigate their dilutive effects.

Hyper scaler spending power (earnings) are themselves at risk

While the over-earnings dynamic affects other companies in the S&P 500, it is important to recognize that hyperscalers themselves—accounting for roughly a fifth of the S&P 500’s weight—appear to be also over-earning, albeit through a different set of mechanisms. As we see it, a portion of their cloud revenues are from the AI start-up ecosystem, which is raising record amounts of capital at unprecedented valuations despite lacking clear long-term moats or enduring revenue potential. In some cases, ex-OpenAI employees are launching their own foundational AI model businesses and hitting valuations of $50B or more before a single dollar of revenue is generated.¹² In other cases, application startups are generating revenue but expending what appears to be an unsustainable percentage of it on costly GPU compute. According to Andreesen Horowitz, many AI startups have already spent 80%+ of capital raised on compute resources.¹³ Including the usage from Large Language Models, almost half of the incremental revenue growth in the hyperscalers’ AI business has been driven by venture capital (VC) funding.

Further to this point, selling to an unprofitable customer base introduces significant risks, as this revenue source is vulnerable to external shocks and funding cycles. Anthropic and OpenAI together account for ~$30B in run-rate revenues and are likely spending an even larger amount on compute, making them dependent on raising additional funds.^14,15 One example is OpenAI’s recent commitment to purchase an additional $250B of compute from Microsoft over the next few years.¹⁶ OpenAI, which is now 27% owned by Microsoft, remains heavily reliant on external financing. In 2025 alone, OpenAI has raised over $40B, up from ~$7B in 2024, in addition to receiving revolving credit lines and striking infrastructure partnerships with other players to sustain its operations.

Investor sentiment among limited partners (LPs) remains deeply negative according to Pitchbook research, with net cash flows to LPs running a cumulative deficit of -$169B since 2022. Compounding this, VC funds with vintages dating as far back as 2018 have delivered effectively no meaningful distributions to capital paid in. The resulting hesitation among LPs to commit additional capital has also led to a meaningful increase in the time it takes to close new funds. By the end of Q3 2025, venture firms had secured just $45B in new commitments for the year, the lowest since 2017. If such VC funding weakness were to further slow and spillover more to the AI startup side, this could indirectly impact hyperscalers’ revenue growth trajectory and further exacerbate risks tied to the broader data center build-out.^18,19,20

The realities of GPUs going from scarcity to glut

The broader implication in our view is a fundamental misallocation of capital: trillions of dollars in planned investment over the coming years are at risk of generating subpar economic returns. This is a zero-sum game that diverts capital from other productive sectors. As we have written recently, we view this as the “Dotcom Bubble on Steroids,” where the market’s distortion lies not in the P/E ratios themselves (which are undoubtedly elevated by any historical standard), but in the inflated earnings in the denominator itself.

Ultimately, we believe the market is pricing a perpetual boom in an environment that is already flashing clear signs of saturation. As hardware scarcity turns to glut and the circular flow of hyperscaler deals and VC slows, the earnings multiplier that has propped up the S&P 500 faces a potential violent unwind. The current valuation framework appears to assume that hyperscaler CapEx will continue to flow indefinitely, enriching all those that are part of the supply chain—but our data suggests the faucet is already tightening. We believe Investors who are ignoring these second-derivative risks are not merely betting on AI; they are betting against the fundamental laws of supply and demand. In our view, when the CapEx merry-go-round stops, the over-earnings will vanish, leaving behind a market that is not only overvalued, but fundamentally mispriced.

END NOTES

¹Singh Dang, PhD, Sanjit. “AI Infrastructure Is Fueling A Circular Economy”. Forbes. 26 October 2025.

²Forgash, Emily and Ghosh, Agnee. “OpenAI, Nvidia Fuel $1 Trillion AI Market With Web of Circular Deals”. Bloomberg. 7 October 2025.

³Becker, Sam. “Meta’s AI spending spree has even the most bullish stock analysts wondering: How much capex is too much?” Fast Company. 18 November 2025.

⁴“Is Software the New Shale?” GQG Partners LLC. 27 December 2022.

⁵Data fetched from the Bloomberg Terminal expects calendar year 2025 operating margins for Nvidia, Broadcom, and Arista Networks to come in at 61%, 43% and 43%, respectively as of 5 December 2025.

⁶Leswing, Kif. “The question everyone in AI is asking: How long before a GPU depreciates?”. CNBC. 14 November 2025.

⁷GQG Research. Management groups we have spoken with—who have good visibility and experience with the technology—believe the lion’s share of the chips’ economic useful life will end up being closer to 18 months.

⁸United States: OECD Economic Outlook, Volume 2025 Issue 2 | OECD

⁹Jensen Huang, Nvidia Q2 2022 earnings call. August 2021.

¹⁰Hock Tan. Broadcom Q4 2021 earnings call. December 2021.

¹¹Che-Chia Wei. TSMC Q3 2021 earnings call. October 2021.

¹²“Mira Murati’s Thinking Machines seeks $50 billion valuation in funding talks, Bloomberg News reports”. Reuters. 13 November 2025

¹³Appenzeller, Guido, et al. “Navigating the High Cost of AI Compute”. Andreessen Horowitz. 27 April 2023.

¹⁴Bort, Julie. “Sam Altman says OpenAI has $20B ARR and about $1.4 trillion in data center commitments.” Tech Crunch. 06 Novembre 2025.

¹⁵Hu, Krystal and Seetharaman, Deepa. “Exclusive: Anthropic aims to nearly triple annualized revenue in 2026, sources say.” Reuters. 15 October 2025.

¹⁶“The Next Chapter of the Microsoft-OpenAI Partnership.” OpenAI. 28 October 2025.

¹⁷James, Liam. “Deutsche Bank Explores Hedging Data Center Risks Amid AI Boom”. Markets.com. 7 November 2025.

¹⁸Tarhuni, Nizar, et al. “Venture Monitor: The definitive review of the US venture capital ecosystem”. Pitchbook Monitor, NVCA. 30 September 2025.

¹⁹2026 US Venture Capital Outlook, Pitchbook, 2025 December 1.

²⁰Walker, Peter, et al. “Q2 2025 VC Fund Performance”. Carta. 23 September 2025.

DEFINITIONS

Blackwell chips are NVIDIA’s next-generation GPUs designed for large-scale AI, data centers, and high-performance computing.

CapEx (capital expenditure) refers to the funds a company uses to acquire, upgrade, or maintain long-term physical or intangible assets. Unlike operating expenses, which cover daily business costs, CapEx is an investment in assets expected to generate economic benefits for more than one year.

Earnings Per Share (EPS) is a measure of a company’s profitability, calculated by dividing quarterly or annual income (minus dividends) by the number of outstanding stock shares.

H200 GPUs are NVIDIA’s data center GPU, the successor to the H100 GPU.

Limited partners are investors in a limited partnership who provide capital but have no say in the business’s management and liability that is limited to the amount of their investment. They are often referred to as “silent partners” and are common in venture capital funds and private equity, where the partnership structure is used to raise capital from passive investors.

Net IRR (Internal Rate of Return) is the actual, after-fee rate of return an investor receives on an investment, accounting for all costs, management fees, and carried interest.

Operating margin is a company’s profit for every dollar of sales after deducting production costs like wages and raw materials but before accounting for interest and taxes. A higher ratio in an operating margin is generally better, as it indicates the company is efficient in its operations and is good at turning sales into profits.

The Price-to-Earnings (P/E) ratio is a stock valuation metric that divides a company’s share price by its earnings per share (EPS), indicating what investors are willing to pay for each dollar of a company’s earnings. A higher P/E suggests investors expect higher future earnings, while a lower P/E may indicate a stock is undervalued or has lower growth potential.

SBC stands for Stock-Based Compensation, and RSUs are Restricted Stock Units, a common type of SBC. These are forms of non-cash payment that companies use to reward and retain employees by granting them an ownership stake in the company.

IMPORTANT INFORMATION

This document may be distributed by GQG Partners LLC and its affiliates (collectively “GQG”).

The information provided in this document does not constitute investment advice and no investment decision should be made based on it. Neither the information contained in this document or in any accompanying oral presentation is a recommendation to follow any strategy or allocation. In addition, neither is it a recommendation, offer or solicitation to (i) sell or buy any security, (ii) purchase shares in any investment fund that GQG may sponsor, offer or manage, (iii) establish any separately managed account, or (iv) implement any investment advice. It should not be assumed that any investments made or recommended by GQG in the future will be profitable or will equal the performance of any securities discussed herein. Before making any investment decision, you should seek expert, professional advice, including tax advice, and obtain information regarding the legal, fiscal, regulatory and foreign currency requirements for any investment according to the law of your home country, place of residence or current abode.

This document reflects the views of GQG as at a particular time. GQG’s views may change without notice. Any forward-looking statements or forecasts are based on assumptions and actual results may vary.

GQG provides this information for informational purposes only. GQG has gathered the information in good faith from sources it believes to be reliable, including its own resources and third parties. However, GQG does not represent or warrant that any information, including, without limitation, any past performance results and any third-party information provided, is accurate, reliable, or complete, and it should not be relied upon as such. GQG has not independently verified any information used or presented that is derived from third parties, which is subject to change. Information on holdings, allocations, and other characteristics is for illustrative purposes only and may not be representative of current or future investments or allocations.

The information contained in this document is unaudited. It is published for the assistance of recipients, but is not to be relied upon as authoritative and is not to be substituted for the exercise of one’s own judgment. GQG is not required to update the information contained in these materials, unless otherwise required by applicable law.

The contents of this document are confidential and intended solely for the recipient. No portion of this document and/or its attachments may be reproduced, quoted or distributed without the prior written consent of GQG.

GQG is registered as an investment adviser with the U.S. Securities and Exchange Commission. Please see GQG’s Form ADV Part 2, which is available upon request, for more information.

Any account or fund advised by GQG involves significant risks and is appropriate only for those persons who can bear the economic risk of the complete loss of their investment. There is no assurance that any account or fund will achieve its investment objectives. Accounts and funds are subject to price volatility and the value of a portfolio will change as the prices of investments go up or down. Before investing in a strategy, you should consider the risks of the strategy as well as whether the strategy is appropriate based upon your investment objectives and risk tolerance.

There may be additional risks associated with international and emerging markets investing involving foreign, economic, political, monetary, and/or legal factors. International investing is not for everyone. You can lose money by investing in securities.

Past performance may not be indicative of future results. Performance may vary substantially from year to year or even from month to month. The value of investments can go down as well as up. Future performance may be lower or higher than the performance presented, and may include the possibility of loss of principal. It should not be assumed that recommendations made in the future will be profitable or will equal the performance of securities listed herein.

GQG Partners LLC is a wholly owned subsidiary of GQG Partners Inc., a Delaware corporation that is listed on the Australian Securities Exchange (ASX: GQG). GQG Partners LLC and its affiliates provide certain services to each other.

INFORMATION ABOUT INDEXES

The S&P 500 Index is a float-adjusted market cap weighted equity index of stocks of 500 leading companies in the United States.

The Philadelphia Semiconductor (SOX) is a capitalization-weighted index comprising the 30 largest U.S.-traded companies primarily involved in the design, distribution, manufacture, and sale of semiconductors.

The Silicon Data H100 Rental Price Index is a daily financial benchmark that tracks the average hourly cost of renting an NVIDIA H100 graphics processing unit (GPU) across various global cloud and bare-metal providers.

You cannot invest directly in indices, which do not take into account trading commissions and costs.

NOTICE TO RECIPIENTS IN AUSTRALIA & NEW ZEALAND

The information in this web page is issued and approved by GQG Partners LLC (“GQG”), a limited liability company and authorised representative of GQG Partners (Australia) Pty Ltd, ACN 626 132 572, AFSL number 515673. This information and our services may be provided to wholesale and retail clients (as defined in section 761G of the Corporations Act 2001 (Cth)) domiciled in Australia. This web page contains general information only, does not contain any personal advice and does not take into account any prospective investor’s objectives, financial situation or needs. In New Zealand, any offer of a Fund is limited to ‘wholesale investors’ within the meaning of clause 3(2) of Schedule 1 of the Financial Markets Conduct Act 2013. This information is not intended to be distributed or passed on, directly or indirectly, to any other person.

NOTICE TO RECIPIENTS IN CANADA (Alberta, British Columbia, Manitoba, New Brunswick, Nova Scotia, Ontario, Québec, Saskatchewan (the “Canadian Jurisdictions”))

GQG Partners LLC relies on the (i) international adviser exemption pursuant to section 8.26 of NI 31-103 in each of the Canadian Jurisdictions, and (ii) non-resident investment fund manager exemption pursuant to section 4 of MI 32-102 in Ontario and Québec and is not registered as an adviser or investment fund manager in the Canadian Jurisdictions.

This web page has been prepared solely for information purposes and is not an offering memorandum or any other kind of an offer to buy or sell or a solicitation of an offer to buy or sell any security, instrument or investment product or to participate in any particular trading strategy. It is not intended and should not be taken as any form of advertising, recommendation, investment advice or invitation to trade. This information is confidential and for the use of the intended recipient only. The distribution of this web page in Canada is restricted to recipients who are qualified “permitted clients” for purposes of National Instrument 31-103 Registration Requirements, Exemptions and Ongoing Registrant Obligations. This information may not be reproduced, redistributed or copied in whole or in part for any purpose without the prior written consent of GQG. Upon receipt of this information, each Canadian recipient hereby confirms having expressly requested that all information evidencing or relating in any way to the information described herein be drawn up in the English language only.  Par la réception de ce document, le détenteur au Canada de celui-ci confirme par les présentes avoir expressément exigé que tous les documents faisant foi ou se rapportant de quelque manière que ce soit aux informations présentées dans ce document soient rédigés en anglais seulement.

NOTICE TO RECIPIENTS IN SOUTH AFRICA

GQG PARTNERS LLC is an authorised financial services provider in the Republic of South Africa and regulated by the Financial Sector Conduct Authority (FSCA) with FSP number: 48881. Investors should take cognisance of the fact that there are always risks involved when buying or selling any financial product. Past performance of a financial product is not necessarily indicative of future performance. The value of financial products can increase as well as decrease over time, depending on the value of the underlying securities and prevailing market conditions. The investment value of a financial product is not guaranteed, and any illustrations, forecasts or hypothetical data are provided for illustrative purposes only. This web page does not constitute financial advice, a solicitation, invitation or investment recommendation. Prior to selecting a financial product or investment, it is recommended that South African based investors seek and obtain specialised financial, legal and tax advice.

NOTICE TO RECIPIENTS IN THE UNITED KINGDOM

GQG Partners LLC is not an authorised person for the purposes of the Financial Services and Markets Act 2000 of the United Kingdom (“FSMA”) and the distribution of this web page in the United Kingdom is restricted by law. Accordingly, this web page is provided only for and is directed only at persons in the United Kingdom reasonably believed to be of a kind to whom such promotions may be communicated by a person who is not an authorised person under FSMA pursuant to the FSMA (Financial Promotion) Order 2005 (the “FPO”). Such persons include: (a) persons having professional experience in matters relating to investments; and (b) high net worth bodies corporate, partnerships, unincorporated associations, trusts, etc. falling within Article 49 of the FPO. The services provided by GQG Partners LLC and the investment opportunities described in this web page are available only to such persons, and persons of any other description may not rely on the information in it. All, or most, of the rules made under the FSMA for the protection of retail clients will not apply, and compensation under the United Kingdom Financial Services Compensation Scheme will not be available.

GQG Partners (UK) Ltd. is a company registered in England and Wales, registered number 1175684. GQG Partners (UK) Ltd. (FRN: 844184) is an appointed representative of Sapia Partners LLP (FRN: 550103), which is authorised and regulated by the Financial Conduct Authority (“FCA”).

NOTICE TO RECIPIENTS IN ADGM

GQG Partners Ltd, a company limited by shares, registered in Abu Dhabi Global Markets (“ADGM”), having its address at Unit No. 1 and 2, Floor 14, Al Maryah Tower, Abu Dhabi Global Market Square, Abu Dhabi, Al Maryah Island, United Arab Emirates. GQG Partners Ltd is licensed by the ADGM’s Financial Services Regulatory Authority (FSRA) (license number 240015). GQG Partners Limited is licensed by the ADGM’s Financial Services Regulatory Authority (FSRA) to conduct the regulated activities of Managing a Collective Investment Fund, Advising on Investments or Credit, Arranging Deals in Investments, Managing Assets, Shari’a-compliant Regulated Activities. This web page is intended for distribution only to persons of a type specified in the FSRA’s Rules (i.e., “Professional Clients”) and must not be delivered to or relied on by any other type of person. It is for the exclusive use of the persons to whom it is addressed and in connection with the subject matter contained therein. The FSRA, or any other regulatory authority, has no responsibility for reviewing or verifying this web page or any other information in connection with it. Accordingly, the FSRA, or any other regulatory authority, neither approved this information or any other associated information nor taken any steps to verify the information set out in this web page and has no responsibility for it.

© 2025 GQG Partners LLC. All rights reserved. This document reflects the views of GQG as of December 2025.