By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel generally agrees that hyperscalers' increasing capital intensity in AI infrastructure poses a near-term liquidity risk, with potential dilution and delayed payback cycles. However, they differ on whether this is a cyclical stress or a structural re-rating of the tech sector's risk profile.
Risk: Sequential equity raises from multiple hyperscalers in 2025 could front-run any revenue inflection, locking in dilution even if ROIC later stabilizes above 15%.
Opportunity: Potential long-run AI upside, assuming secular demand and efficient AI inference and software-enabled services.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
If the facts change, then I have to change. I learned that from reading the legendary British economist John Maynard Keynes, and it has never steered me wrong. I was a huge bull on this market because of the terrific things happening in artificial intelligence. And because we had a new Federal Reserve chair in Kevin Warsh, who wanted to cut interest rates to support growth, given the frozen housing market and an underclass struggling with higher gas prices, a drop in health care coverage, and the gratuitous cut in food stamps. I made this judgment, as I always do, by going to the source: the top twenty public U.S. retailers, who know far more than the government does about spending habits. Those include the auto and the auto parts companies, as well as ancillary home care businesses. The idea that we don't need rate cuts is fanciful. But then we got the monthly employment report on Friday, which showed that job growth unexpectedly surged in May, with nonfarm payrolls jumping a seasonally adjusted 172,000, far above the Dow Jones consensus estimate of 80,000. Of course, that figure never distinguishes between those who are employed and doing well and those who are employed and trying to hold it together. A job means self-sufficiency in this country, and there's no counterargument to that line of thinking, despite my attempt to make one when I interviewed Kevin Hassett, President Donald Trump's chief economic adviser. But it wasn't worth the effort, even though the nature of the poor working class is obvious to anyone except those who only have wealthy friends and acquaintances. Knowing people of all levels of the economic spectrum is like knowing the cost of milk and gasoline. It helps you to do your job well. The stronger-than-expected report wiped out the chance of rate hikes, or even a single one, this year, and that had been a major prop to my bull case. I could have looked past the macro if not for the dramatic change in the data center buildout, where costs have risen sharply on everything from labor and construction materials to power and site development. We went from thinking there would be a payback in the near future to having no sense of when it would occur. As recently as a month ago, I felt that Amazon's confidence about a return on invested capital was so definitive that there was no reason to worry. Now I worry that Amazon may need to make an equity offering because the payoff from AI is more elusive than definite. I also thought that we would only have to deal with equity offerings from a few private companies: OpenAI, Anthropic, and SpaceX. But after Alphabet announced plans to raise $80 billion through stock sales to fund its AI efforts, it looks like the biggest tech companies, including Amazon, Microsoft , and Meta , may need to raise huge sums for AI by selling stock. There is no way this market can handle that much equity and stay at these levels. So we find ourselves going from the prospect of a rate cut in 2026 and a smooth glide path to hyperscaler profitability, to a possible rate increase this year and an endless series of gigantic fund raises. Now, let's take a break for a moment and let me don my hedge fund hat. I don't like to do so for many reasons. One is if I were going to do it, I might as well be a hedge fund manager, and I did that and retired because, well, go read "Confessions of a Street Addict," and you will know why. Another is that I can't do what a hedge fund manager does. When I saw the Alphabet offering succeed, I should have realized that we would then face a deluge of deals without the money to do them. I have said over and over to you that when we have so many deals, we can't think straight, and it's time to hit the road. Hit the road in hedge fund lingo means selling everything you were long and going short those same stocks, while pivoting to stocks unrelated to the data center, à la Johnson & Johnson and Procter & Gamble . It also means that you blow out of stocks like AI stocks like Arm , Qnity , and even Corning , as well as Marvell and Nvidia . You would do that because you would realize that you are going to get multiple stories about how the data center is hated all over the country, too hard to build, too expensive, and a never-ending source of more equity. The hyperescalers, which were the source of the greatest stock story ever told, are now at the epicenter of the bear case. I can't do any of that now. I can't trade around that much, and I am always talking about my positions. That's why the Morning Meeting (at 10:20 a.m. ET) is so important: you hear what I would do, not what I can't do. The rules protect everyone. I am in a position to influence prices, and I don't take advantage of that personally. So what do we make of a case where we believe we could get an offering from Microsoft or Amazon at any moment? We have to first posit what happens when the SpaceX deal gets priced next Friday. There are many people who don't have any link to E-Trade or Robinhood (or any other firm with the stock) who will buy some shares at the market price. They will be the ones who determine the opening. The stock should be worth $4 trillion unless SpaceX founder and CEO Elon Musk says that the institutions can take profits. Right now, because I have no idea where the supply comes from, the opening should be like the spring of 1999. That kind of opening actually creates some liquidity that might be helpful for the next five deals. Or the company just sops up every available dollar and the edifice crumbles, which is a very realistic possibility. The novelty of the deal does not help its case, because nobody seems to know what will happen. I don't care about all these tranches; I just care about the first batch of shares coming to market. I can't expect it to be good. I just want it to be on a day when Trump doesn't create some manufactured turmoil. The real issue here is how much money is allocated to tech and how much to non-tech stocks. Growth investors used to the Rule of 40 — a software metric that posits a company's revenue growth rate and profit margin should equal at least 40% — are fleeing tech for health-care and consumer companies, among others, that still have strong organic growth. So you have to reposition yourself. We've tried. We have more to do. I like all of our health care, especially Cardinal, which had no flies to begin with. I am mindful that we started our Intel position too early with our initial buy last week . But if the market absorbs all of this supply, Intel will be the next Micron . Let's talk about that possibility. At some point, an Nvidia client will have to step up and say how lucrative this all is. I believe that will be Anthropic. When that deal is done, the worst may be over. If I were a hedge fund manager, I would know to rebuild the whole shooting match. But I am not. So it is too difficult, but I will do my best to let you know. So what will my posture be tomorrow? Simple: a chastened bull, not an aggressive bear. Chastened because I kept saying that when we get too many deals, we have to skedaddle. But a bull because the long-term impact of the Fourth Industrial Revolution isn't so long-term; it's happening now. It's just that Microsoft doesn't want to be marginalized. Amazon wants Amazon Web Services to remain the leading cloud platform. Meta might want to have a robust Meta Web Services. And Alphabet wants to preserve its frontrunner status. They all have to keep spending heavily or be left behind. But in the near term, there are a lot of mouths to feed and not enough to eat. The market could get overwhelmed. That's where we find ourselves. It's a suboptimal place to be. (See here for a full list of the stocks in Jim Cramer's Charitable Trust.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust's portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
Four leading AI models discuss this article
"Near-term equity dilutions and rising data-center cash burn risk capping multiple expansion for hyperscalers, even as secular AI demand persists."
The piece correctly flags near-term funding pressures on hyperscalers and the risk that big equity raises could dilute shareholders. But it may overstate the inevitability of a bear case: AI demand remains secular, and cloud leaders still enjoy pricing power and potential margin upside as workloads shift to efficient AI inference and software-enabled services. The key risk is persistent, elevated data-center cash burn and potential dilution delaying multiple expansion, not a collapse in AI demand itself. If capex and energy costs remain stubbornly high, the timing of profitability could slip, keeping valuations range-bound despite long-run AI upside.
The strongest counter would be that AI adoption accelerates faster than capex can displace, and hyperscalers monetize AI gains through software, services, and edge strategies, unlocking cash flow sooner than feared; dilution may be offset by higher margins and share buybacks.
"The shift from organic cash-funded AI investment to dilutive equity financing signals that the near-term ROI on data center infrastructure is significantly lower than current market valuations suggest."
The article correctly identifies a critical inflection point: the transition from 'AI potential' to 'AI capital intensity.' Hyperscalers are shifting from experimental spending to massive, opaque infrastructure buildouts, creating a supply-demand imbalance in equity markets. When companies like Alphabet and potentially Amazon resort to equity financing for capex, they dilute shareholder value just as the cost of capital remains 'higher for longer' due to sticky employment data. The 'Rule of 40' migration is already underway, as institutional capital rotates toward defensive, cash-generative sectors like healthcare (e.g., Cardinal Health). The risk isn't just a valuation correction; it's a liquidity crunch caused by a deluge of tech-sector supply.
The thesis ignores that hyperscaler capex is largely self-funded through massive free cash flow, and equity offerings may be tactical balance sheet optimization rather than distress signals. If AI efficiency gains begin to manifest in enterprise software margins by Q4, these 'excessive' costs will be reclassified as high-ROI investments, triggering a rapid valuation re-rating.
"The article conflates capex inflation with equity dilution risk, but doesn't establish whether the underlying AI investments will generate sufficient returns to justify either."
This is a rambling personal narrative masquerading as market analysis. The core claim—that massive AI capex raises will overwhelm equity markets—conflates two separate issues: (1) whether hyperscalers can profitably deploy capital, and (2) whether equity issuance will crash valuations. Alphabet's $80B raise succeeded without visible market stress. The author conflates job strength (bullish for demand) with rate cut odds (now lower), then pivots to capex concerns without quantifying them. Missing: total AI capex as % of S&P 500 market cap, historical precedent for tech capex cycles, and whether higher rates actually reduce hyperscaler ROI or just extend payback periods.
If AI capex truly generates 20%+ returns (as hyperscalers claim), equity raises at current valuations are rational capital allocation, not distress—and the market has historically rewarded companies that raise cheap equity to fund high-ROI projects. The author's 'too many deals' intuition may be emotional rather than structural.
"Multiple large equity raises by hyperscalers to fund AI will create near-term supply overhang that the market cannot absorb without de-rating."
Cramer's shift highlights a credible near-term liquidity risk: hyperscalers' surging data-center capex (labor, power, construction) is forcing equity raises after Alphabet's $80B plan, with MSFT, AMZN, and META likely to follow. This coincides with stronger May jobs data removing rate-cut support. The market may struggle to absorb multiple large offerings while growth investors rotate into health care and staples. However, the piece underplays balance-sheet strength and recurring cash flow at these firms. It also conflates private-company raises (OpenAI) with public ones without quantifying net share-count impact or timing.
These companies generate enough free cash flow and can tap debt markets at low rates, so equity raises may be smaller or slower than feared, allowing AI revenue to offset dilution before multiples compress.
"One large equity raise doesn't prove resilience; ROI lags and higher debt costs could still compress valuations if multiple hyperscalers issue equity before AI capitalization translates into durable cash flow."
Claude's jab that Alphabet's $80B raise proved no market stress ignores the signal: one successful issue doesn't prove structural resilience if multiple issuings follow and ROI lags. The risk is a delayed payback cycle in AI capex, not demand collapse. If rates stay higher for longer or debt costs rise, dilution could still pressure multiples, even with strong FCF. Focus on ROIC/ROCE of AI deployments, not issuance size alone.
"Hyperscalers are transitioning into capital-intensive utilities, which necessitates a structural compression of their valuation multiples."
Claude is right to call out the lack of quantitative rigor, but both Gemini and ChatGPT miss the primary risk: the 'utility' transition. As hyperscalers pivot to massive, long-tail energy and compute infrastructure, they are becoming capital-intensive utilities, not high-margin software firms. If ROIC (Return on Invested Capital) trends toward utility-like 8-10% levels rather than software-like 25%+, current P/E multiples are fundamentally unsustainable. We are witnessing a structural re-rating of the tech sector's risk profile, not just a liquidity squeeze.
"Hyperscaler capex intensity may stabilize at sustainable utility levels without collapsing ROIC, but only if buildout peaks soon—the inflection timing is everything."
Gemini's utility re-rating thesis is the sharpest risk here, but needs stress-testing: Alphabet's 2024 capex is ~$60B on ~$307B revenue (19.5%). Utilities run 15-20% capex/revenue sustainably. If hyperscalers stabilize there post-buildout, ROIC needn't collapse to 8-10%—software-like returns are possible if AI workloads scale faster than infrastructure. The real question: does capex peak in 2025-26, or does it remain elevated? That timing determines whether this is cyclical stress or structural re-rating.
"Equity raises will likely precede AI revenue scale, creating dilution before any ROIC recovery materializes."
Claude correctly flags the capex peak question but misses how sequential equity raises from MSFT, AMZN and META in 2025 could front-run any revenue inflection by 18-24 months, locking in dilution even if ROIC later stabilizes above 15%. Gemini's utility re-rating assumes a full business pivot that core segments won't experience. The sequencing risk between issuance windows and AI monetization remains untested.
The panel generally agrees that hyperscalers' increasing capital intensity in AI infrastructure poses a near-term liquidity risk, with potential dilution and delayed payback cycles. However, they differ on whether this is a cyclical stress or a structural re-rating of the tech sector's risk profile.
Potential long-run AI upside, assuming secular demand and efficient AI inference and software-enabled services.
Sequential equity raises from multiple hyperscalers in 2025 could front-run any revenue inflection, locking in dilution even if ROIC later stabilizes above 15%.