What the accelerating collapse of SaaS giants tells us about the next era of computing — and who gets left holding the bag
A note before we begin: In 2021, the software industry was the Titanic steaming full speed toward New York. Not a care in the world. Record valuations. Record revenues. Record arrogance. Then ChatGPT arrived in late 2022, quietly, almost politely. Three years later, the iceberg is dead ahead — and the crew is only now beginning to understand what’s underneath. What’s happening is one of those rare, structural shifts that rewires entire economies. Like the iPhone in 2007. Like the internet in 1995. If you use software at work, this affects you. If you run a business, this could save you a fortune. If you invest, it might save you from losing one. Read on.
In the Beginning, There Was a Floppy Disk
Before the cloud, before subscriptions, before “SaaS” was even a word, software came in a box.
You bought Microsoft Office for a few hundred dollars, installed it from a stack of floppy disks, and it was yours. Forever. No annual renewals. No per-user fees. No consultants billing you $200 an hour to configure a form.
Then the internet grew up. A handful of clever businesspeople realized they could move software off your hard drive and onto their servers — and charge you every month for the privilege. Forever.
It was, by any objective measure, the greatest business model ever invented.
Why sell something once when you can sell it twelve times a year? Salesforce went public in 2004 and spent two decades becoming a $300 billion behemoth. ServiceNow. Workday. HubSpot. Atlassian. An entire generation of empires built on the same locked-in logic: make switching painful, make customization limited, make implementation expensive, and watch the revenue compound.
For a generation, it worked magnificently.
So it goes.
The Quiet Shift Nobody Saw Coming (Except the Markets)
Atlassian — makers of Jira and Confluence — saw its share price drop roughly 50% in the opening weeks of 2026. Co-founders Mike Cannon-Brookes and Scott Farquhar watched approximately $7.2 billion of their combined net worth evaporate. In weeks.
They may not be worried. But they should be.
Salesforce dropped 30% since January. ServiceNow cratered from an all-time high of $239 to $109. Adobe has been obliterated — down 51%, erasing every pandemic-era gain, back to April 2020 levels. HubSpot is down 54% year-to-date. Across the ten biggest names in enterprise SaaS, over $450 billion in market cap has been wiped since January 1.
On February 4th alone — a single trading day — over $300 billion evaporated from the software sector. Microsoft shed $360 billion in one session. The iShares Expanded Tech-Software ETF plunged 27% from its highs.
Wall Street has a name for it now: the SaaSberg.
The markets aren’t panicking without reason. They’re pricing in something structural. And unlike most panics, this one has a clear, real cause.
AI is coming for the subscription model.
The iPhone Moment Nobody Announced
When Steve Jobs introduced the iPhone in 2007, Nokia’s executives reportedly laughed. Better specs. Better batteries. Decades of infrastructure. What could a computer company know about phones?
Within five years, Nokia was a footnote.
What’s happening in software rhymes with that story in uncomfortable ways. The difference is there’s no single product announcement this time. No stage. No black turtleneck. The disruption arrived quietly — ChatGPT in November 2022, then Claude, then Gemini, then a thousand tools built on top of them. In just over three years, an entire industry’s business model has been called into question. The disruption is distributed — happening across thousands of companies, through tools most executives still haven’t fully experimented with.
The core shift: AI models are now capable enough that companies can build their own custom software. Not “hire a developer,” which was always theoretically possible and always prohibitively expensive. Actually build it themselves. Operations managers. Business owners. People who have never written a line of code.
Platforms like Lovable are enabling exactly this. Its CEO Anton Osika posted recently that every change to a software system will soon go through AI-powered vulnerability analysis before hitting production. Security, implementation, configuration — the three most lucrative professional services revenue streams in the legacy SaaS world — are on the chopping block.
This is not aspirational. This is where software is heading.
The Racket That’s Finally Being Exposed
Here’s the uncomfortable truth nobody in the industry wanted to say out loud: a significant portion of the value these companies charged for was friction.
Not features. Friction.
Implementing Salesforce at a mid-sized company could run hundreds of thousands in consulting fees. Implementation partners charged $200 to $300 an hour — for months — to configure fields and set up workflows that should have been included in the product. Workday implementations at enterprise scale sometimes ran into the millions. You paid for access. Then you paid again to make access useful. Then you paid annually, forever, for the privilege of not losing everything you’d built.
The lock-in was the product.
Companies paid because the alternative — building something custom — was impossibly expensive. AI just destroyed that calculus.
A company can now build a system tailored to their exact workflows, without a consultant, without a $150k implementation, without a subscription that resets every time they want to add a field. They own it. It evolves with them.
That’s not a marginal improvement. That’s a fundamental restructuring of who holds power in the software relationship.
Not All Survivors Are Created Equal
But here’s where the story gets more interesting.
There’s a meaningful difference between companies that bolted AI onto existing platforms and companies that were born from AI as their foundation. Think of it like the EV moment. Ford makes electric vehicles now — but Ford also makes combustion engines. It has legacy plants, dealer networks, and institutional muscle memory pulling it in two directions. Tesla never made anything else. There’s no old guard fighting for budget.
In software, the Teslas of this era are Palantir, Lemonade, Upstart, SymphonyAI, and Canva. Companies where AI isn’t a feature — it’s the skeleton everything is built around.
Palantir, trading at $309 billion in early February 2026, has built an ontological data fusion layer that maps an organization’s entire decision-making structure. Governments and militaries have spent years integrating it. You can’t replicate that with a prompt. Lemonade’s behavioral ML models get sharper with every claim, every policy, every interaction — compounding in ways a human underwriter cannot match at scale. Upstart maps over 1,600 variables to assess credit risk versus FICO’s twelve, approving roughly twice as many loans at half the default rate. The model is the moat.
These companies have data flywheels that compound over years. Legacy companies that decide today to go AI-first can’t catch that. The moat isn’t just technical. It’s temporal.
The Companies Trying to Turn the Ship
Some legacy players are fighting back — and deserve credit for it.
Salesforce launched Agentforce, partnered with Anthropic, and is now rolling out flat-fee enterprise licensing that abandons the seat-based model entirely — taking losses on early contracts just to stop customers from leaving. That’s not a confident company playing offense. That’s a company that smells smoke.
ServiceNow is pushing AI Agent Studio aggressively. HubSpot has shipped over 80 AI features across its platform. Adobe’s Firefly generative AI is embedded throughout its creative suite.
But retrofitting AI onto platforms built for a different era carries a structural problem no engineering budget fully resolves. You’re adding intelligence to a system designed around human workflows, seat-based pricing, and integration lock-in. You’re making the car faster. The car was designed for roads that may not exist in five years.
Shopify is the instructive exception. Its core product is genuinely enhanced by AI rather than threatened by it — the stock decline has been modest by comparison. Compare that to Zendesk, whose core product is ticketing. If AI agents resolve 90% of customer service tickets autonomously, why do you need Zendesk at all? That’s the question every legacy vendor’s customers are starting to ask.
And Those That Should Be Losing Sleep
DocuSign has the clearest risk profile. E-signature was always thin — a legal wrapper around a PDF. AI agents now draft contracts, negotiate terms, collect signatures, and file in one workflow. Its market cap has fallen from $12 billion to roughly $8 billion since January. That trajectory doesn’t reverse.
Salesforce’s Agentforce is a real differentiator — but it’s racing against a model where entire CRM categories get rebuilt from scratch using Claude or ChatGPT in a few weeks. Down 30% since January, its forward P/E compressed from above 30x to roughly 22x. The market has marked down its future pricing power in real time.
ServiceNow at $109, down from $239. IT workflows, HR automation, change management — high-value tasks, but highly replicable ones for well-configured AI agents. The -44% tells you what institutional investors think of the survival odds.
The Democratization Nobody Asked For (But Everybody Needed)
AI is to software what the internet was to publishing. Before the internet, distributing a newspaper required presses, trucks, newsrooms, and capital. After: anyone with a laptop could reach a global audience. Power didn’t disappear — it redistributed.
Before AI, building custom software required development teams, months of time, and real capital. After AI, it requires a good specification and a few iterations with a model. The power is redistributing again.
For the regional retailer, the healthcare clinic, the logistics company — for businesses that spent decades paying premium prices for standardized tools that never quite fit — this is genuinely liberating. The vendor monopoly was never inevitable. It was a function of barriers. Those barriers are collapsing.
Companies like Wix are betting on this moment, embedding AI deeply at no extra cost to customers — prioritizing adoption over margin. For a company their size, that’s feasible. For Microsoft, doing the same across Office 365, Teams, and Azure threatens the margins that keep investors happy. So Microsoft hedges, partnering with Anthropic — betting on the infrastructure layer of the next era the way Amazon built AWS for the last one.
The companies most likely to look unrecognizable by 2030 are those whose core value proposition was simply: we did the hard work of building this so you didn’t have to. That value proposition just expired.
A Question Worth Sitting With
The SaaS era created real value. It made sophisticated software accessible to companies that couldn’t afford development teams. It standardized tools that needed standardizing.
But it also created a comfortable monopoly. The vendor knew your data better than you did. Leaving cost more than staying. The consultants were expensive because the platforms made them expensive.
AI is disrupting that monopoly. And the disruption, for once, seems likely to benefit the customer rather than just rearrange who captures the rent.
So the question isn’t really whether the SaaS empire is falling. The evidence suggests it is.
The question is what gets built in the space it leaves behind — and whether the companies that inherit the next era learn anything from the ones being displaced.
History suggests they won’t.
But that’s a column for 2030.
Notes Beyond the Map 
Leave a comment