Microsoft added nearly $500 billion in market value in a single trading session Thursday, a gain of more than 17%, after reporting fiscal fourth-quarter results that showed Azure revenue surpassing $100 billion for the first time and Microsoft Cloud revenue rising 27% year over year to $59.3 billion. The Azure milestone narrows a long-standing gap with Amazon Web Services, which reported roughly $107 billion in 2024 revenue, and positions Microsoft as the only major rival within striking distance of the cloud market leader.

Even so, blowout results from other companies released around the same time have not been met with comparable enthusiasm. Samsung reported a record $62 billion profit, a 19-fold increase, and still saw its shares sell off. Meta was also punished after another quarter of revenue growth. The Dow fell more than 1,100 points on Wednesday, before Microsoft reported its results after the close.

Three market watchers viewed the week’s moves through three different lenses.

The leverage argument

Steve Sosnick, chief strategist at Interactive Brokers, told Axios earlier in the week that the market narrative had shifted from “all news being good news for AI” to something closer to “let’s look under some rocks and see what the risks are.” After Microsoft’s earnings, he laughed as he told Fortune that the story had changed “just a tad.” But he said the underlying question remained unresolved. “We are in a ‘rip up the script every day’ kind of mode,” he said.

In Sosnick’s view, this week’s price swings are simply too large and too fast to be explained by fundamentals alone. Microsoft’s 17% jump, Micron’s 18% rise on what he called unremarkable news, and IBM’s 25% cut after a profit warning weeks earlier are the kinds of moves that make him uneasy. He acknowledged that it was a cliché to reach for the late-1990s comparison, then did so anyway.

“I hate to say it, but the only time I can recall these sort of swings is the 1999, 2000 period,” he said. That dot-com era also featured “everybody rushing in and out of stocks,” he added, before culminating in a multi-year bear market that erased trillions in technology market value.

What worries him more than the historical comparison is that today’s market has many more tools to magnify swings than it did back then, including weekly options and leveraged ETFs. He stopped short of blaming those products directly — “I’m not going to say they cause the volatility, that’s not fair” — but said there is a “generational” effect running through market behavior, echoing the “financial nihilism” framing some see at work in markets.

That framing lines up with a growing body of research. A Harris Poll found that 46% of Gen Z respondents agree that “no matter how hard I work I will never be able to afford a home I really love,” while 42% of Gen Z investors hold cryptocurrency compared with just 11% who hold a retirement account. The World Economic Forum has flagged financial nihilism, reportedly coined in 2021, as one of the most significant economic trends shaping the next decade.

The specific pressure point that resolved this week, in Sosnick’s view, was the forced unwind of Leopold Aschenbrenner’s hedge fund, Situational Awareness LP. The fund, founded by the 24-year-old former OpenAI researcher, had reportedly returned 439% net through June 30, according to an investor letter cited by the Financial Times, but it was running gross exposure of as much as four times its capital, with its top five positions accounting for more than three-quarters of its disclosed long book.

A decline of roughly 30% or more across core holdings in July — driven by losses in AI infrastructure names such as SK Hynix and a large Nebius stake, along with losses on short bets against software companies including Adobe — was enough to wipe out the fund’s equity cushion. Prime brokers Goldman Sachs, JPMorgan Chase and Bank of America issued margin calls, and the fund ultimately sold its entire public equities book, longs and shorts alike, to a single unnamed buyer before Thursday’s open.

The market’s behavior leading up to that resolution, Sosnick said, reflected a growing awareness that “some leveraged investments were going wrong.” When the Aschenbrenner fund was removed, markets appeared to exhale. “The market is reacting to this like, ‘Okay, we’re done, the leveraged trades are behind us,'” he said. Even so, the size of the moves in both directions still gives him pause. When SK Hynix and Samsung posted positive results and markets barely reacted, he took notice. “When stocks react poorly on good news, it’s telling you there’s something really wrong in the market structure.”

The fundamentals argument

Melissa Otto, who leads research for Visible Alpha at S&P Global, rejected the nihilism framing outright. “I don’t understand where the nihilism comes from,” she said. Her reading of Microsoft’s rally is not psychological but mechanical: for the first time, a hyperscaler has produced a “very quantifiable metric” — Azure’s accelerating growth — that shows the AI business model is actually working, rather than merely promising to work eventually.

Microsoft’s numbers support that view. Azure revenue growth accelerated well beyond expectations in the latest quarter.

Otto’s broader framework for understanding the volatility centers on what she calls an “overhang” — a market thesis stuck in an unusually wide debate, which she said is visible in her firm’s data as widening dispersion in analyst estimates. “When I see estimates narrow, that to me means there’s less debate in the market and you’re going to see less volatility,” she said. “But when I see the opposite… it means the debates are getting much more polarized and much more extreme.”

That overhang, she said, snapped back after Microsoft’s results. The debate around AI spending had been building for roughly six to eight weeks as the market shifted into a “show me the money” posture. Investors have also grown used to seeing large sums of capital returned through buybacks over the years, even as roughly $1.5 trillion in combined AI capex is planned by Microsoft, Meta, Amazon and Alphabet between this year and next — a sum exceeding the annual GDP of countries such as Spain or Mexico.

Otto said Microsoft’s specific advantage rests on enterprise entrenchment rather than model quality, which she expects to commoditize over time regardless of which AI lab wins any given benchmark. “Name me a financial analyst working on Wall Street that doesn’t use Excel,” she said. “Name me an investment banker that doesn’t use PowerPoint… it’s just in the DNA of these industries.”

In her view, that ubiquity gives Microsoft a durable channel to sell Azure and Copilot into enterprises that Amazon Web Services and Google Cloud cannot easily replicate, because Azure is already embedded in existing enterprise software stacks.

The capex skeptic

Derek Horstmeyer, a finance professor at George Mason University, began with the detail that may have been lost in Thursday’s rally: Aschenbrenner’s losing positions likely would have rebounded the next day. “I guess the worst part of it is everything reversed today — if he had survived, he could have made it through.”

To Horstmeyer, that is not coincidence but a feature. The go-for-broke attitude behind Situational Awareness LP — 4x gross leverage, concentrated positions and conviction bets — is the same attitude he sees in retail markets in a different form. “It really is an attitude of go broke or shoot for the moon — use as much leverage as you can.” In his view, the fund and the retail crypto buyer are operating from the same psychology at different scales.

But his deeper concern is not leverage or sentiment. It is the structure of the AI capex race itself. Every hyperscaler is overspending, free cash flow is turning negative across the board, and nobody wants to be left out. He said he has seen this dynamic before. “Every company got into the streaming wars, and it’s money-losing for a lot of them, and they haven’t given up — they all still have their own platforms,” he said, pointing to a pattern in which Disney, Paramount and Comcast have continued operating unprofitable streaming services years after launch rather than concede the category. There will eventually be a winner in AI infrastructure, “but I don’t want to be in the race of finding the one that’s going to win.”

His clearest example of the market’s uneven bets is Apple, largely because it lacks an AI strategy and therefore is not spending hundreds of billions of dollars after missing its chance to become a hyperscaler. “Because they’re not in the AI race, they’re kind of doing okay,” he said. Apple recently hit $5 trillion in market cap and has regained its place above Nvidia as the world’s most valuable company.

Horstmeyer acknowledged the long-term risk that Apple could be genuinely behind, but said the company’s outsider status has, for now, insulated it from the volatility affecting AI-exposed peers. “It’s not a Silicon Valley mindset,” he said.

He also offered a procedural example from his own campus. Horstmeyer runs a student-managed investment fund at George Mason, where major portfolio votes happen every three to six months. The group is four days away from a vote that has split sharply.

“One side wants to get out of the AI trade, and some want to double down and invest in this very niche, fiber-optic company — basically a supplier for a data center,” he said. “Some students are like, ‘this is the absolute future,’ and some don’t see it at all.”

He said there is a noticeable pattern in who lands on which side. “There’s a definite correlation between the kids who like crypto” and a greater willingness to take on AI infrastructure bets, he said, while others, usually more by-the-book accounting students, tend to be more “prudent” and move in the opposite direction. “Human nature,” he said — “go figure.”

This story was originally featured on Fortune.com