Four companies beat earnings estimates on the same August morning. Two got rewarded. Two got sold anyway. That split, not the headline numbers, is the story developers should actually be watching.
For the better part of a year, the working theory on Wall Street was brutal in its simplicity: AI coding agents write code, code is what software companies sell, therefore AI agents were going to eat software companies alive. The Morningstar US Software Application Index, home to stocks such as Salesforce, ServiceNow, and Adobe, fell roughly 27% from October 2025 through mid-July 2026 while the broader market rose nearly 15% over the same stretch. By some measures the damage was worse still — the S&P 500 Software & Services index fell more than 33% from its October 2025 peak to April 2026 as the market wrestled with AI disruption. Traders had a name for it: the SaaSpocalypse. On one especially ugly session in April, Cloudflare plunged 12%, Snowflake dropped 9%, ServiceNow fell 7%, and Salesforce slid 4% in a single trading session that erased billions in market capitalization. Weeks earlier, Microsoft alone shed approximately $360 billion in market capitalization in one session, an outcome that demonstrated the disruption thesis was being applied not only to mid-cap point solutions but to the largest enterprise platform on earth.
Then the actual numbers started arriving, and the theory ran into a problem: the infrastructure that AI-generated code depends on turned out to be doing better than ever, not worse.
The reversal, name by name
Salesforce and CrowdStrike delivered the clearest rebuttal. Salesforce surged 21.6% and CrowdStrike jumped 19.4% after both crushed estimates and raised guidance, with Agentforce annual recurring revenue hitting $1.5 billion — up 240% year-over-year — effectively silencing fears that AI would gut the seat-based model. CrowdStrike's own numbers backed the point directly: Q2 revenue of $1.47 billion topped $1.44 billion estimates, and full-year revenue guidance was raised to $5.99 billion to $6.01 billion. The read-through, per that coverage, was blunt — cybersecurity spending isn't being displaced by AI, it's being accelerated by it. The relief spread past the two headline names: European software stocks caught the bid, with SAP's ADR rising 4.4%, Nemetschek up 3.5% and TeamViewer up 3.0% following the U.S. peers higher, while Adobe gained 5.9%, helped by an expanded Stagwell partnership.
The developer-tools layer told a messier, more interesting version of the same story. A wave of eight companies reported within a single window in early August — HubSpot, Datadog, Cloudflare, JFrog, Atlassian, Twilio, Duolingo, and Figma — and every one beat on the headline, yet four rose and four got sold hard anyway, a split one analyst argued wasn't noise but a genuine fork in how investors are pricing AI exposure. Atlassian and Twilio became the pivot point: the narrative began shifting as Atlassian and Twilio posted historic earnings pops, reviving confidence in pockets of the sector that many had written off. Figma's case was the opposite and the most telling — it had staged a roughly 67% rally from a low into its August 5 earnings, and one analyst described the quarter as great, with revenue continuing to accelerate, yet the stock sold off 13% because it had run aggressively into the print. Datadog and JFrog sit closer to the plumbing than the canvas — they get paid when more code ships, more services run, and more artifacts need tracking, which is exactly what an economy full of AI coding agents produces in volume.
Nuveen's chief investment officer, Saira Malik, made the underlying case for why the panic overshot. She argued that investors spent much of 2026 worrying generative AI would gut software company headcounts and revenue growth, a fear that hit the sector broadly and indiscriminately, but the numbers never backed up the panic. "When you separate the signal from the noise, it didn't show up in software companies fundamentals," she said. Margins and earnings held up too, and the wave of job cuts many expected AI to trigger across the sector has not materialized to the degree feared.
Why the plumbing wins
The irony is that the tools developers actually use to write code faster are the same tools generating the demand software-infrastructure companies now report. Roughly 84% of developers now use or plan to use AI coding tools, and GitHub Copilot leads with about 20 million users while Cursor reached $2 billion in annual recurring revenue by February 2026. Every one of those agents needs somewhere to store artifacts, somewhere to log errors, somewhere to run tests, and somewhere to catch the vulnerability an autonomous agent introduced at 2 a.m. That is Datadog's ledger, JFrog's registry, Cloudflare's edge, Atlassian's ticket queue. AI didn't remove the plumbing. It multiplied the water flowing through it.
None of this settles the argument, though. The next test lands in October — the same month bankers are eyeing for Anthropic's public debut. Anthropic investors expect the company to pursue an IPO in October 2026 at a valuation of $2 trillion or more, which would make it the largest IPO in history, surpassing SpaceX's record-setting June 2026 debut, and the speculation itself is already being read as a catalyst expected to draw fresh capital into the space and generate meaningful downstream demand for software products tied to AI development. A pricing that lands anywhere near that number will either confirm that the infrastructure bulls read the cycle correctly, or hand the bears their receipts back. Software didn't die. It's waiting to see what it's worth.
