A high Wall Road analyst has sounded an alarm over the U.S. fairness bull market, warning that its outstanding run is constructed on a precariously slender basis: a surge in spending on, and optimistic assumptions about, infrastructure for synthetic intelligence (AI). This spending has fueled a growth within the shares of many of the so-called Magnificent 7 and some dozen associated companies, which have now come to account for roughly 75% of the S&P 500’s returns for the reason that rally of the previous few years started.
The commentary on September 29 by Morgan Stanley Wealth Administration’s chief funding officer, Lisa Shalett, frames the present market growth as a “one-note narrative” nearly totally depending on large capital expenditures in generative AI, elevating questions on its sturdiness as financial and aggressive dangers begin to mount. Shalett’s critique got here squarely in the course of some individuals within the AI area — and plenty of monetary commentators round Wall Road —fretting at market exuberance and starting to speak overtly a few bubble.
In an interview with Fortune, Shalett stated she was “very involved” about this theme in markets, saying her workplace had broadened from a perception that the market would solely bid up seven or 10 shares to roughly 40. “On the finish of the day … this isn’t going to be fairly” if and when the generative AI capital expenditure story falters, she stated.
Shalett stated she’s nervous a few “Cisco second” like when the dotcom bubble burst in 2000, referring to the corporate that was briefly essentially the most helpful firm on this planet earlier than an 80% inventory plunge. [By “Cisco moment” did she mean a whole bunch of circular financing coming back to bite the company? If so, that would be worth adding/briefly explaining.] When requested how shut we’re to such a second, Shalett stated in all probability not within the subsequent 9 months, however very presumably within the subsequent 24. Whenever you take a look at the precise spending and the quantity of capital coming into the area, “we’re rather a lot nearer to the seventh inning than the primary or second inning,” she stated.
‘Beginning to do what all final dangerous actors do’
Shalett’s feedback centered on a number of current multibillion-dollar offers to scale up data-center infrastructure. As notable substacker and former Atlantic author Derek Thompson just lately famous in a submit titled “That is how the AI bubble will pop,” a lot cash is being spent to assist AI’s energy-consumption wants that it’s the equal of a brand new Apollo area mission each 10 months. (Tech firms are spending roughly $400 billion this yr alone on data-center infrastructure, whereas the Apollo program allotted about $300 billion in in the present day’s {dollars} to get to the moon from the Nineteen Sixties to the ’70s.)
What’s greater than a bit of regarding to Shalett is that one firm alone, Nvidia—essentially the most helpful firm within the historical past of the world, with an over $4.5 trillion market cap—is on the heart of a big variety of these offers. In September alone, Nvidia invested $100 billion in OpenAI in an enormous deal, simply days after pledging $5 billion to Intel (the Intel settlement was tied to chips, not data-center infrastructure, per se).
Fortune‘s Jeremy Kahn reported in late September on important issues about “round” financing, or Nvidia’s money primarily being recycled all through the AI business. Shalett sees this as a significant concern and a significant signal that the enterprise cycle is headed towards some sort of endgame. “The man on the epicenter, Nvidia, is mainly beginning to do what all final dangerous actors do within the ultimate inning, which is extending financing, they’re shopping for their traders.”
Shalett expanded on her issues by saying that firms round Nvidia “are beginning to change into interwoven.” She famous that OpenAI is partially owned by Microsoft, however now Nvidia has additionally made an funding within the startup, whereas Oracle and AMD every have their very own buying agreements with OpenAI. However OpenAI additionally has a data-center cope with tech large Oracle, with the “dangerous information,” Shalett notes, that this deal is “completely debt-financed.” OpenAI additionally struck a deal in October with chip-maker AMD that permits OpenAI to purchase as much as 10% of AMD. “Primarily, Nvidia’s major competitor goes to be partially owned by OpenAI, which is partially owned by Nvidia. So, Nvidia can ‘personal’ a chunk of its largest competitor. It’s completely round and will increase systemic threat.”
When reached for remark, a spokesperson for Nvidia stated, “We don’t require any of the businesses we make investments in to make use of Nvidia know-how.”
Nvidia CEO Jensen Huang mentioned the OpenAI funding in an look on the Bg2 podcast with Brad Gerstner and Clark Tang on September 25, calling it an “alternative to speculate” and a part of a partnership geared towards serving to OpenAI construct their very own AI infrastructure. When requested in regards to the allegation of round financing normally and the Cisco precedent particularly, Huang talked about how OpenAI will fund the deal, arguing that it must be funded by OpenAI’s future revenues, or “offtake,” which he identified are “rising exponentially,” and by its future capital, whether or not it’s raised by a sale of fairness or debt. That may relies on traders’ confidence in OpenAI, he stated, and past that, it’s “their firm, it’s not my enterprise. And naturally, now we have to remain very near them to ensure that we construct in assist of their continued progress.”
Shalett stated that she and her workforce had been “beginning to watch” for indicators of a bubble popping, highlighting the deal introduced roughly every week earlier than OpenAI struck its $100 billion data-center cope with Nvidia, when it struck one other with Oracle price $300 billion. Analysts at KeyBanc Capital Markets estimated that Oracle must borrow $100 billion of that quantity—$25 billion a yr for the subsequent 4 years.
“Each morning the opening display screen on my Bloomberg is what’s occurring with CDS spreads on Oracle debt,” Shalett stated, referring to credit score default swaps, the monetary instrument that was obscure earlier than the Nice Monetary Disaster, however notorious for the function it performed in a world market meltdown. CDSs primarily function insurance coverage to traders in case of insolvency by a market entity. “If individuals begin getting nervous about Oracle’s means to pay,” Shalett stated, “that’s gonna be an early indication to us that individuals are getting nervous.” She added that each one the indications to her converse of the tip of a cycle and historical past is plagued by cautionary tales from such instances.
Oracle didn’t reply to requests for remark.
90% progress for the reason that final bear market
For the reason that October 2022 bear market backside and the launch of ChatGPT, in accordance with Shalett’s calculations, the S&P 500 has soared 90%, however most of those good points have come from a small group of shares. The so-called “Magnificent Seven”—together with high-profile names like Nvidia and Microsoft—plus one other 34 AI data-center ecosystem firms, are accountable for, as cited by Shalett and individually by JP Morgan Asset Administration’s Michael Cembalest, about three-quarters of general market returns, 80% of earnings progress, and a staggering 90% of capital spending progress within the index. Comparatively, the opposite 493 names within the S&P 500 are up simply 25%—displaying simply how concentrated the rally has change into.
The so-called “hyperscaler” firms alone are actually spending near $400 billion yearly on capex supporting AI infrastructure, Morgan Stanley Wealth Administration calculated. The financial affect of AI capex is now immense, contributing an estimated 100 foundation factors—absolutely one share level—to second-quarter GDP progress, in accordance with Morgan Stanley’s analysis. This tempo outstrips the speed of underlying shopper spending progress by tenfold, underscoring its centrality to each market efficiency and broader financial information.
“Folks conflate AI adoption, which is within the first inning, with the capex infrastructure buildout, which has been going full-out since 2022,” Shalett instructed Fortune. She cited issues in regards to the prominence of personal fairness and debt capital coming into play, as that “tends to provide bubbles, as a result of it might be unspoken-for capability.” In different phrases, individuals have cash to burn they usually’re throwing it at issues that will not repay.
Shalett waved away macro theories in regards to the labor market or the Federal Reserve. “We predict that’s lacking the forest for the timber as a result of the forest is totally rooted on this one story” about AI infrastructure. Morgan Stanley’s bull-case mid-2026 worth goal for the S&P 500 is an eye-popping 7,200, however Shalett highlights that even essentially the most optimistic outlook admits that threat premiums, credit score spreads, and market volatility don’t appear to totally account for the vulnerabilities lurking beneath the AI-fueled advance.
Shalett’s evaluation means that AI capex maturity is approaching and a few potential slowdowns are already seen. As an illustration, hyperscalers have already seen free-cash-flow progress flip destructive, an indication that funding might have outpaced underlying know-how returns. Strategas, an unbiased analysis agency, estimates that hyperscaler free money circulate is about to shrink by greater than 16% over the subsequent 12 months, placing stress on lofty valuations and forcing traders to demand extra self-discipline in how these funds are deployed.
Shalett was requested about information facilities’ disproportionate impression on GDP all through 2025, which media blogger Rusty Foster of At the moment in Tabs described as: “Our economic system would possibly simply be three AI information facilities in a trench coat.” The Morgan Stanley exec stated “That’s what makes this cycle so fragile,” including that sooner or later, “we’re not gonna be constructing any information facilities for some time.” After that, it’s only a query of whether or not you crash: “Do you may have a light 1991-92-style recession or does it actually change into dangerous?”
A extra bullish case
Financial institution of America Analysis weighed in on the semiconductors sector in a Friday be aware, writing that vendor financing within the area, particularly Nvidia’s $100 billion dedication to OpenAI, has been “elevating eyebrows.” However, the workforce, led by senior analyst Vivek Arya, argued that the deal is structured by efficiency and aggressive want, quite than pure speculative frenzy.
In an interview with Fortune, Arya defined why he wasn’t nervous regardless of the “optics” being fairly clearly dangerous. “It’s very straightforward to say, ‘Oh, Nvidia is giving [OpenAI] cash and they’re shopping for chips with that cash” and so forth, however he argued the headlines are deceptive about how a lot cash is definitely being spent and the $100 billion sticker worth on the OpenAI deal “scared everybody.” Noting that the deal has a number of tranches that may play out over a number of years to return, he stated it’s not like Nvidia is “simply handing a $100 billion verify to OpenAI [and saying] you recognize, go have enjoyable.”
“Nvidia didn’t fund all of it,” Arya stated of the broader generative AI capex growth. Citing public filings, Arya argued that Nvidia’s whole funding within the AI ecosystem is actually lower than $8 billion or so during the last 12 months, not such a big determine in spite of everything. And he’s nonetheless bullish on Nvidia and OpenAI, he added, as a result of he sees them because the winners of this specific story. “We predict they’ll be among the many 4 or 5 ecosystems that come up. It’s not like Nvidia goes and investing in each a type of ecosystems, proper? They’re solely investing in a type of 5, which is, after all, essentially the most disruptive,” that being OpenAI.
When requested about his personal fears of a bubble, Arya really sounded a calmer however strikingly comparable tune to Shalett. “I’m extraordinarily comfy with what is going to occur within the subsequent 12 months,” Arya stated, “And I’ve excessive sense of optimism about what is going to occur within the subsequent 5 years. However can there be durations of digestion in between? Yeah.” Explaining that that is the character of any infrastructure cycle, “it’s not all the time up and to the fitting.” In different phrases, after the subsequent 9 months in Shalett’s opinion and the subsequent yr in Arya’s, the data-center buildout endgame may very well be in play. “When these information facilities are constructed,” Arya stated, “they don’t seem to be constructed for in the present day’s demand. They’re constructed with some anticipation of demand that may develop within the subsequent, you recognize, 12 to 18 months. So, are they going to be 100% utilized on a regular basis? No.”
Rising worries a few bubble
A number of the greatest names in tech and Wall Road supplied had been hedging arduous about the potential for a bubble on Friday. Goldman Sachs CEO David Solomon and Jeff Bezos, each talking at a tech convention in Turin, Italy, stated they had been seeing the identical patterns as Shalett. Solomon stated the large quantities of spending weren’t essentially completely different from different booms and busts. “There might be quite a lot of capital that was deployed that didn’t ship returns,” he stated. That’s no completely different from how funding works. “We simply don’t know the way that may play out.”
Bezos characterised it as “sort of an industrial bubble,” arguing that the infrastructure would repay for a few years to return.
OpenAI CEO Sam Altman, who obtained markets jittery in late August when he talked about the B-word, was requested once more to touch upon the topic whereas touring (what else?) an enormous new information heart in Texas. “Between the ten years we’ve already been working and the various a long time forward of us, there might be booms and busts,” Altman stated. “Folks will overinvest and lose cash, and underinvest and lose quite a lot of income.”
For his half, Cisco CEO John Chambers, one of many faces of the dotcom bubble, instructed the Related Press on October 3 that he sees “quite a lot of great optimism” about AI that’s much like the “irrational exuberance on a very massive scale” that marked the web age. It signifies a bubble to him, however solely “a future bubble for sure firms. Is there going to be prepare wreck? Sure, for those who aren’t capable of translate the know-how right into a sustainable aggressive benefit, how are you going to generate income in spite of everything the cash you poured into it?”
When requested whether or not the dimensions of this potential bubble represents uncharted waters for the economic system, particularly contemplating the one-note nature of the lengthy bull market, Shalett stated Wall Streeters are all the time evaluating threat. However placing on her “American citizen hat,” she warned in regards to the media consolidation that sees Oracle’s founder Larry Ellison additionally now enjoying a significant function in TikTok (as a part of a shopping for consortium of Trump-friendly billionaires) and Paramount in Hollywood and CBS Information in New York (via his son, David Ellison, the media firm’s new proprietor). Shalett stated she’s nervous about “groupthink” filtering into the functioning of markets. “That isn’t one thing that the majority of us have skilled in our lifetimes,” she stated. “You cease factoring in threat premiums into markets, there is no such thing as a bear case to something.”