A violent, AI-fueled selloff erased more than $400 billion from software names this week before a Friday bounce offered thin relief. The Nasdaq still closed down more than 2% for the week. Salesforce and ServiceNow dropped over 9% as investors repriced how AI agents could displace seat-based software. “You’re getting to a point where this probably seems overdone,” Invesco’s Brian Levitt told Yahoo Finance. The question into next week: is this a sentiment purge or the start of a new valuation regime?
The trigger was not a single earnings miss but a narrative reset. Uptake of Anthropic’s Claude Code and a drumbeat of AI agent demos stoked fears that enterprise buyers will swap bundled, intelligent assistants for a patchwork of point solutions. Alphabet’s latest earnings call amplified it with “productivity” comments that read as code for doing more with fewer seats. In three sessions, software was hit in a way that looked indiscriminate: quality, growth, and profitability were sold alike as investors modeled faster cannibalization of workflows that used to require separate licenses. By Friday, dip buyers surfaced, but the term sheet had changed. The market is now treating AI as a near-term revenue risk to software, not just a long-dated margin tailwind.
Underneath the price action, multiples cracked. Sector P/E ratios have slid from roughly 35x in late 2025 to near 20x today, the lowest since 2014, even after a modest end-of-week bounce. That retrenchment lines up with a rough 25% compression in valuation metrics over the past week as investors re-rate the durability of seat-based growth. And yet, even at 20x, large-cap software still trades at a steep premium — more than double — to the equal-weight S&P 500. That disconnect is the battleground. Bulls argue the selloff front-loads fear without evidence of broad churn or collapsing net retention; bears say the sector has further to fall as AI automation caps pricing power and flattens unit expansion. One inconvenient truth for both sides: the dispersion inside “software” is widening, and averages hide it.
The Big Tech capex story is the second shoe. Amazon, Alphabet, Meta, and Microsoft are steering AI spend that is set to top $650 billion, and they intend to monetize it fast. If hyperscalers bundle powerful agents across productivity suites, cloud platforms, and developer tools, third-party vendors will feel the squeeze on both price and relevance. This is not theoretical. Microsoft is pushing Copilot deeper into Office and Azure, Google is hardwiring AI into Workspace, and Amazon is aiming Bedrock at enterprise buildouts. The platform gravity risk is real: when incumbents can cross-subsidize AI features and enforce distribution through existing seats, independent vendors must prove they offer unique data, workflow depth, or compliance that the platforms cannot. Without that proof, valuation erosion becomes structural, not cyclical.
Not every ticker wears the same scarlet letter. “Larger software companies that are positioned to adapt are going to be fine,” said Mike O’Rourke at JonesTrading. Investors are already sorting for scale, data moats, and mission-critical workloads. Names with telemetry reach across clouds, robust marketplaces, and clear AI attach paths stand out as candidates for a rebound once panic selling fades. Street targets imply notable upside for a handful of franchises tied to observability, data clouds, and enterprise platforms that can embed AI agents without nuking seat economics. The market is signaling a preference for vendors that can sell outcomes and consumption, not just seats. Conversely, mid-tier applications with overlapping features and tenuous switching costs are the most exposed to AI consolidation and procurement scrutiny.
SaaS is pivoting under pressure. The per-seat era is giving way to hybrid models — usage, outcomes, or agent-based pricing that ties revenue to measurable productivity. That transition is healthy long-term but messy in quarters. CFOs are resisting blanket seat expansions and asking vendors to prove incremental ROI against AI alternatives. Expect more land-small, expand-on-usage motions and granular metering. That shift could dent near-term revenue recognition and muddy annual contract value metrics, even if dollar-based net expansion stabilizes later. It also changes how investors value the space. Clean, high-visibility subscription streams trade one multiple; metered, variable SaaS-plus-consumption blends trade another. The faster management teams quantify AI-driven uplift without cannibalizing core SKUs, the quicker multiples can bottom.
Positioning made the slide worse. Software entered the year crowded, and when AI agents spooked the tape, capitulation met systematic de-risking. Tech is negative year-to-date while investors rotated into staples and energy, where relative valuations look cheaper and earnings revisions steadier. Some of the week’s damage had the feel of a behavioral flush. Several high-quality franchises now trade at multi-year relative lows despite intact balance sheets and strong gross margins. Corporate buybacks and insider buying, if they materialize at scale, could help set a floor. But the flows will not reverse on vibes alone. Funds will want evidence — churn not spiking, seat reductions offset by AI attach, stable cash conversion — before re-risking into a group whose premium, even after compression, still assumes durable growth.
Strip out the noise and the core questions are testable. Are AI copilots and agents demonstrably replacing licenses in customer cohorts today, or just capping net new seats? Are customers consolidating vendors faster than expected as platforms bundle AI, and if so, where? Are pilots converting to paid usage at levels that backfill slower seat growth? Early reads from select enterprises suggest caution, not collapse. Budgets are being re-allocated, not eliminated. The risk is pacing: if the transition from seats to usage drags a few quarters, the street will mark numbers down again. But that is different from an existential reset. For investors willing to parse product telemetry, pipeline commentary, and attach-rate disclosures, the dispersion trade — long adaptive platforms, short redundant apps — is becoming clearer.
Guidance and language matter from here. Listen for how Salesforce, ServiceNow, and peers frame AI agent attach, pricing experiments, and churn in enterprise cohorts. Watch hyperscaler commentary on AI free-to-paid conversion and whether they tighten bundling to pull features out of the third-party ecosystem. Track signals on per-seat contractions versus workload expansions and whether vendors are willing to sacrifice near-term billings to lock in usage-pricing footprints. Any credible visibility on stabilizing net revenue retention, improving AI monetization, or stepped-up buybacks can flip the tape from fear to repair. Until then, the market will keep demanding a higher proof bar and a lower multiple. The selloff may be “overdone” in the short run — but in software’s new AI regime, only operators who adapt quickly will get their premiums back.
