Microsoft shares fell approximately 5% on Wednesday, July 22, and closed below $400 for the first time in this cycle. The move came ahead of the company’s fiscal Q4 earnings on July 30. Investors are focused on two questions: how much of Microsoft’s roughly $80 billion FY26 AI capex will convert into enterprise revenue, and whether Azure OpenAI monetization is scaling fast enough to justify Microsoft’s premium valuation multiple.

The stock is now down about 28% from the all-time high it set last fall. Wednesday’s session also came with the announcement of an expanded AMD partnership aimed at diversifying AI hardware supply beyond Nvidia — a move that markets read partly as strategic and partly as a sign of the cost pressure Microsoft is under to reduce Nvidia dependency. The $400 level had held as psychological support for months. Losing it opened the door to a technical retest of levels not seen since early 2024.

Microsoft matters to the Software Selloff more than any other name because it is the largest single component of most software indices and ETFs. When Microsoft is down 5% on a day, the entire IGV falls with it. And because Microsoft is also the AI-capex thesis vehicle for most institutional investors, a Microsoft drawdown compounds the sector-wide negative signal.

Salesforce carries the heaviest damage in the Software Selloff. The stock is down roughly 35% year to date as of July 22, making it the worst-performing member of the Dow Jones Industrial Average this year. KeyBanc downgraded the name on July 9 citing AI growth concerns. Salesforce delivered its first revenue miss in 18 years in March. On July 22 the stock fell another 4.02% — one of the biggest single-day laggards on the Dow.

The narrative eating Salesforce is specifically the SaaSpocalypse thesis. Investors are pricing in per-seat CRM revenue as vulnerable to AI agent replacement. Sanofi’s confirmed migration off ServiceNow, Greenleaf’s $300-per-month Salesforce replacement, and Atonom’s 97% SaaS spend cut all reinforce the storyline. Even activist investor Jeff Smith at Starboard Value, who added to his Salesforce position last year, said publicly the multiple has come down and the pressure is on value software specifically, not just Salesforce.

Not even Marc Benioff’s tighter relationship with the Trump administration has arrested the slide. Wedbush’s Dan Ives has been vocal that the market is pricing Salesforce and ServiceNow as if they will lose about 5% of their customers over the next few years — a framing that assumes not only revenue growth stalls but that installed-base retention actually breaks. That is the sharpest version of the SaaSpocalypse thesis, and it is currently in Salesforce’s share price.

Oracle’s Software Selloff story is quieter than Salesforce’s but arguably more severe. Oracle stock is down roughly 55% from the record high it set in late 2025. The company’s OCI cloud business had been the darling of the AI-infrastructure story for much of 2025, with enormous multi-year contract wins from OpenAI and other model labs. Then the story shifted. Investors began questioning whether Oracle’s AI-capex commitments would be recovered inside the reported deal economics.

Oracle’s Q1 2027 fiscal report in September 2025 initially thrilled Wall Street with headline RPO growth. But subsequent quarters have shown gross margins pressured by the AI-infrastructure buildout, and revenue timing on the OpenAI contracts has been pushed further out than the market originally assumed. The result: a stock that ran to record highs on the AI-capex thesis has been taken apart on the AI-capex-return thesis. Same underlying business, opposite market reaction inside 12 months.

Oracle also serves as a warning for how quickly narrative can flip in the Software Selloff. The company is one of the more entrenched enterprise vendors on the market. Its database franchise is famously sticky. But that stickiness has not protected the valuation multiple when investors decided the AI infrastructure spending was outrunning the payoff timeline.

J.P. Morgan Research published analysis this year explicitly labeling the Software Selloff as being driven by “broken logic.” The argument: investors are simultaneously worried about two contradictory AI outcomes. Salesforce and Adobe are being sold because AI will disrupt their revenue models. Microsoft is being sold because AI capex will not generate returns. But both fears cannot be true at the same time. If AI is powerful enough to eat SaaS revenue, then AI capex is by definition generating returns. If AI capex is not generating returns, then AI is not powerful enough to disrupt Salesforce.

The J.P. Morgan team framed the current setup as a factor overshoot rather than a fundamental repricing. The Morningstar equity research desk reached a similar conclusion, arguing software fundamentals still look sound and the drawdowns are running past what earnings and cash flow support. Wells Fargo Investment Institute has echoed the “indiscriminate selling” framing. The consensus among these desks is not that everything is fine — it is that the selloff is over-punishing the fundamental deterioration.

The counter-argument from bears is that the market is not confused, it is just choosing to hedge both risks. Investors who cannot say for sure whether SaaS or AI-capex is the bigger loser are simply reducing exposure to both, which produces exactly the pattern J.P. Morgan calls “broken logic.” Whether this is broken logic or rational hedging depends on what earnings show over the next four weeks.

The Software Selloff has a structural component that is not going to reverse just because Microsoft delivers a good quarter. Per-seat SaaS pricing was built for a world where each seat represented one human user producing incremental value. When one AI agent can do the work of five seats, the pricing model gets pulled apart from underneath the vendor. Salesforce, HubSpot, Zoom, DocuSign — all built businesses on per-seat billing. All are now being asked to price consumption or outcome instead.

GitHub already made the move. Copilot Premium shifted from a flat monthly seat price to token-usage billing on June 1. Zendesk, Workday, and multiple mid-market SaaS vendors have announced pricing structure changes tied to AI feature usage. The direction is consistent across the incumbents: less per-seat, more per-consumption. Vendors that reprice fast enough keep the customer through renewal. Vendors that hold the line on per-seat get outnegotiated.

This repricing wave is the real story behind the 35% drawdown at Salesforce and the 27% drawdown at Adobe. Even if AI does not literally replace their tools, the revenue-per-customer curve is flattening. Wall Street is discounting future ARPU rather than pricing a total revenue collapse — and the discount factor is doing a lot of work in the current multiples.

The path out of the Software Selloff runs through the July-August earnings cycle. Alphabet reported after the close on July 22, with Cloud growth as the single most important number in that print. Microsoft is up July 30 with Azure and Copilot as the read-through metrics. Meta reports on advertising strength and reality-labs capex. Amazon closes the mega-cap cycle with AWS growth as the AI-infrastructure benchmark.

Bullish scenario: cloud growth stabilizes above 25% at all three hyperscalers, enterprise AI monetization shows up in reported segment revenue, and 2027 capex guidance either flattens or is guided lower. In that scenario, J.P. Morgan’s “broken logic” call gets validated, and IGV rebounds meaningfully from the 22% YTD hole. Software regains some multiple, though the SaaSpocalypse thesis remains a structural overhang on per-seat SaaS specifically.

Bearish scenario: cloud growth decelerates, AI monetization stays vague on the earnings calls, and 2027 capex guidance goes higher. In that scenario, the Software Selloff extends into September, and the “25-year worst” framing from Dan Ives gets tested against 2000-2002 dot-com comparisons. Salesforce’s 35% YTD drawdown becomes 45%. Microsoft below $400 becomes Microsoft below $350. The IGV ETF touches 30% YTD losses.