SaaSpocalypse vs Reality: What the Earnings Data Actually Shows

The market has a new favorite fear. Call it the SaaSpocalypse.

Since January, over $1 trillion in market cap has been erased from software stocks. Atlassian down 35%. Salesforce down 28%. Adobe down 30% from its highs. The thesis is simple and terrifying: if AI agents can do the work of five humans, companies don't need five software seats. Per-seat pricing adoption has dropped from 21% to 15% in twelve months. The SaaS business model — the most powerful wealth creation engine in tech for two decades — is supposedly dying.

It's a clean narrative. It's also mostly wrong.

I track a broad cross-section of US public tech companies through their quarterly earnings filings — not stock prices, not analyst ratings, but the actual numbers these companies report every ninety days, cross-referenced against what management said last quarter and what they promised the quarter before that. This earnings season, those filings told a very different story than the one the market is pricing.

What the Filings Actually Show

Across the universe of tech companies I follow, here's what the latest quarterly filings show when I classify each company's growth trajectory based on the actual reported numbers:

• Roughly two-thirds are accelerating or inflecting upward
• About a quarter are stable
• Around 15% are decelerating

Read that again. Nearly two-thirds of these companies are growing faster than they were a year ago. Not stable. Not "hanging in there." Accelerating.

The SaaSpocalypse narrative assumes AI is a wrecking ball aimed at software. The earnings data says AI is a rocket engine strapped to most of the sector — and the few companies it's disrupting were already in trouble for reasons that have nothing to do with seat counts.

The AI Buildout Is Real, Measurable, and Accelerating

Let me walk through what the latest quarter's filings reveal, layer by layer.

The infrastructure layer is exploding. NVIDIA's revenue growth accelerated from 56% to 73% year-over-year across three quarters, adding $10-11 billion per quarter in Data Center revenue alone. Broadcom's AI semiconductor revenue hit 140% growth. TSMC — the bottleneck for every advanced chip on the planet — is running its 3nm lines at full capacity. Vertiv's data center power equipment backlog doubled to $15 billion, with a 2.9x book-to-bill ratio. Celestica's cloud solutions revenue surged 64%, with orders from hyperscalers outstripping capacity.

These aren't projections. They're reported results, verified against the actual press releases.

The hyperscalers are spending at war-footing levels. Amazon, Microsoft, Google, and Meta are collectively deploying roughly $600 billion in capital expenditure this year. AWS reaccelerated to 24% growth — the fastest in thirteen quarters. Google Cloud grew 48% at a $70 billion run rate while expanding operating margins from 11% to 30%. Oracle's cloud infrastructure grew 84%, pulling total revenue growth to 22%.

Memory is in a structural shortage. About 30% of that $600 billion goes to memory chips. Micron reported 37% sequential revenue growth with 68% gross margins, and demand still exceeds supply by 35-50%. AI now consumes more than half of all DRAM production globally, and new fabs won't deliver meaningful supply until late 2027. Every byte of HBM memory destroys roughly four times as much consumer DRAM capacity because of the manufacturing process. The memory companies aren't just doing well — they're in a multi-year super-cycle with no supply relief in sight.

The Software Story Is Bifurcated, Not Broken

Here's where the SaaSpocalypse narrative falls apart completely. If AI agents are going to kill enterprise software, the first casualty should be workflow automation — that's literally what AI agents do. So look at ServiceNow: approximately 20% revenue growth at $13 billion in annual revenue, margins expanding, guidance beaten consistently. Their AI product, Now Assist, has surpassed $600 million in annual contract value with new bookings doubling year-over-year. The company that should be most threatened by AI agents is instead selling AI agents on top of its existing platform.

Datadog accelerated to 29% revenue growth at $3.4 billion annual scale. The most telling detail: the acceleration isn't driven by AI-native startups (the convenient narrative), but by the core enterprise customer base that went from 18% to 23% growth — exactly the customers the SaaSpocalypse thesis says should be cutting software seats. More AI workloads means more infrastructure to monitor.

Snowflake's RPO surged to 42% growth with seven nine-figure deals — up from two a year earlier. More AI means more data to process, store, and query. MongoDB is sustaining roughly 30% growth in Atlas at $2 billion scale, with RPO nearly doubling. Cloudflare accelerated revenue growth to 33.6%, with its developer platform and AI gateway turning a CDN company into core AI infrastructure.

Even Adobe, one of the SaaSpocalypse poster children (stock down 30% from highs), reported subscription revenue accelerating to 13% growth in its latest quarter. Autodesk accelerated from 11% to 14% constant-currency growth as AI-assisted design tools drive adoption rather than replace it.

The cybersecurity sector is accelerating almost across the board — CrowdStrike posted a record $331 million in net new ARR, Zscaler is inflecting upward, Rubrik hit record bookings, and Palo Alto Networks grew revenue 15% while pursuing $35 billion in acquisitions. The logic is straightforward: more AI workloads, more attack surface, more security spending. These aren't companies surviving AI. They're companies being pulled upward by it.

Salesforce is the nuanced case. Organic growth has slowed to roughly 8%, which is a legitimate concern. But 8% growth for a company earning $40 billion in annual revenue is not a company being disrupted — it's a company maturing. Its cRPO still grew 11% on a constant currency basis, and the Agentforce product hit $800 million in ARR within a few quarters. The market's most feared example of AI disruption is a company whose AI product is its fastest-growing business line.

Where the Real Casualties Are

The honest accounting requires acknowledging where growth is genuinely fading. Among the companies that are decelerating, the pattern is clear: the casualties are concentrated in consumer-facing and legacy businesses, not in enterprise software exposed to AI.

Snap is in user decline in North America (-6%). Pinterest's revenue growth has fallen from 17% to 9% on a constant-currency basis. Yelp's advertising revenue turned negative. These are companies with structural challenges that predate the AI narrative entirely.

The consumer hardware story is also getting harder. Qualcomm's handset business faces a 23% sequential decline driven partly by memory supply constraints — the same dynamic benefiting Micron and Samsung is punishing smartphone makers.

And yes, some software companies are decelerating: GitLab (growth slowdown from ratable model), Okta (persistent NRR erosion at 106%), Freshworks (13% growth as it prioritizes profitability). But even in these cases, the deceleration predates the SaaSpocalypse by multiple quarters and has company-specific explanations that have nothing to do with AI replacing seats.

The Accountability Ledger

Since this blog has been dormant since November, I owe readers an honest accounting of what I wrote then versus what happened.

UiPath — In September, I wrote that the company was "at a crossroads" and the case was "challenging until a data-backed turnaround emerges." Six months later, ARR growth remains stuck at 11% for three consecutive quarters. NRR has eroded to 106% on a constant-currency basis. FY2027 guidance implies approximately 9% revenue growth — a slight deceleration from FY2026. The enterprise deal momentum is genuine ($1M+ customers surged to 357, deals over $1M up 50%), but it hasn't translated into accelerating topline growth. The September assessment holds.

Coursera — In October, I argued Coursera was "a generative AI opportunity mispriced as a threat". The latest quarter shows 10% revenue growth for three consecutive quarters — stabilization, but not acceleration. The Consumer segment grew 10-13%, while Enterprise decelerated to 6%. The turnaround story is partially confirmed (the bleeding stopped) but the acceleration thesis I was optimistic about hasn't materialized. This is an area where the initial assessment was too generous.

Duolingo — In August, I wrote about strong bookings growth while noting the early signs of slowing engagement with a bullish lean. The engagement deceleration I flagged as a risk turned out to be the dominant signal — DAU growth fell from 51% to an expected 20%, and bookings guidance for FY2026 stepped down to 10-12%. The market has punished Duolingo harshly, pricing in a narrative that AI makes language learning obsolete. I think that's an overreaction. The deceleration is real and needs to be respected in the near term, but Duolingo has 116 million monthly active users, a proven gamification engine, and is already expanding beyond languages into math, music, and chess. The longer-term opportunity is a genuine AI-powered tutoring platform — the kind of agentic application that could expand the company's addressable market well beyond language learning. The growth rate call was wrong. The "Duolingo is finished" narrative the market is running with may also be wrong.

Three assessments, zero perfect accuracy — but the quarterly data caught all three trajectory shifts within a quarter or two. That's the value proposition: not being right the first time, but having a structured way to detect when reality diverges from the thesis before it costs you too much.

The Agentic Evolution in Practice

Let me explain how I can make claims about what this many companies reported in a single quarter, because that's the third dimension of this story.

Every quarter, when a company I track files its earnings report, the filing is automatically retrieved — press release from EDGAR, transcript from the earnings call. AI models analyze each filing against the company's previous quarters: what metrics changed, what management promised versus what they delivered, where the growth trajectory shifted. The output is a structured forensic assessment — growth trajectory, confidence level, reasoning, and specific predictions for the next quarter.

This isn't a sentiment tool or an AI stock-picker. It's a forensic system that reads the actual filings the way a thorough analyst would — except it can process a full earnings season in the first week while maintaining consistency across every single company.

I built this system over the past year using Claude Code — Anthropic's agentic coding tool — working alongside the AI to architect the database, write the analysis methodology, and design the research workflow. The irony is thick: the same AI technology that the market fears is destroying software businesses is what makes it possible for an individual investor to maintain deep forensic coverage across the entire tech sector. A year ago, this would have required a team of junior analysts. Now it requires one person and a system that costs less per month than a single Bloomberg terminal.

The per-token economics of AI are shifting fast. The agentic evolution — AI models that can execute multi-step tasks autonomously, not just answer questions — means the gap between institutional research capacity and what's available to individual investors is closing. Not theoretically. I'm doing it. Every claim in this piece is backed by specific numbers from the actual quarterly filings, cross-referenced and verified.

What This Means for Investors

The SaaSpocalypse narrative has exactly one thing right: the way software is built, sold, and priced is changing. Per-seat pricing is under genuine pressure. AI agents will reduce headcount in some organizations, and that will affect some SaaS companies' user counts.

But the narrative has the causality backwards. AI isn't destroying tech — it's restructuring it. The $600 billion in infrastructure spending, the memory shortage, the cloud reacceleration, the cybersecurity expansion, the data platform surge — this is the largest capital expenditure cycle in technology history, and it's pulling the majority of software companies upward, not pushing them down.

The bifurcation is real. Companies with strong platform positions, consumption-based models, or products that become more critical in an AI-driven world (security, observability, data infrastructure) are accelerating. Companies with commodity features, weak moats, or consumer-facing models where engagement is discretionary are under pressure. But that's not "AI kills software." That's "AI accelerates the strong and exposes the weak" — exactly what every major technology cycle has done.

Nearly two-thirds of the tech companies I track are accelerating. The filings are public. The math is not ambiguous.

The question isn't whether AI is transforming the tech sector. It is. The question is whether investors are positioned for what the data actually shows — or for the story that sounds scariest on a Bloomberg terminal.

I know which one I'm betting on.