The SaaSpocalypse is already here

Three major enterprise software companies showed their hand this week. Datadog built its own AI model because it's scared. VMware is bleeding customers after Broadcom's pricing overhaul. SAP is €2 billion behind its own cloud migration targets.

These aren't separate stories. They're the same one told three times.

What's happening

Datadog published research this week about a model called Toto-Open-Base. It's a 151-million-parameter time-series model trained on over two trillion data points from their own operations. The reason they built it: customers are starting to use AI to build their own monitoring tools instead of paying Datadog's subscription fees.

Their Chief Product Officer named the threat "SaaSpocalypse." That's not my word. That's an executive at a publicly traded software company using it internally to describe what she sees coming.

When a company builds a defensive AI moat around its own product, it's because the walls are already shaking.

VMware is a different kind of problem, but points the same direction. Broadcom restructured licensing into all-or-nothing bundles after the acquisition. Customers pushed back hard. A Virtified survey of 450 VMware customers across 14 countries found that 50% plan to reduce their VMware usage by 2028. Nutanix, Microsoft, and Red Hat are picking up the business. This one isn't about AI disruption. It's customers running from bad pricing.

SAP's situation is the third flavor: the expensive migration losing the argument. SAP missed its own cloud migration targets by €2 billion. In 2022, they projected legacy on-premises revenue would fall to €8.5 billion by 2025. It came in at €10.5 billion. Customers aren't moving. So SAP stopped pushing them and pivoted to AI upselling instead. As one SAP practice lead put it: "Modernization has come to an end in terms of rip-it-out-and-start-again."

Three different situations. But the same thing underneath: businesses are resisting expensive, locked-in software at exactly the moment when they have more options than they used to.

The shift nobody's naming

For years, vendors held the power. You paid the subscription or you didn't have the capability. That was the deal.

Now a small operations team with access to a modern LLM and some engineering bandwidth can build internal tools that would have taken a dedicated product company years to ship. I want to be careful not to oversell this. Building a rough replacement is genuinely feasible for a lot of operational tools. Maintaining it reliably, keeping it secure, getting new staff up to speed on it, debugging it at 2am when something breaks — that's harder. The math only works if the tool is genuinely simple and your team actually wants to own it long-term. Most tools don't meet that bar.

But the economics have shifted enough that vendors are scared. Datadog's CPO told reporters: "I do not worry about a race to develop models, but applying them." Read that carefully. The goal isn't to build the best model. It's to get embedded deeply enough that leaving becomes more trouble than it's worth. That's a retention strategy dressed up as a product vision.

The evidence right now is mostly on the vendor side. The businesses actually replacing SaaS tools with AI-assisted alternatives aren't publishing case studies. But when the vendors start building defensive moats, that's usually a sign the pressure is real.

What to ask about every tool in your stack

If you're running a company, you probably have somewhere between 10 and 30 SaaS tools active. Some are critical. Some you pay for out of habit. Some you're locked into because switching feels like a bigger project than it's worth.

The question worth asking: is this vendor using AI to make the product genuinely better, or using AI to make it harder to leave?

Broadcom's VMware is the obvious answer to "harder to leave." Tightened bundles, opaque licensing, aggressive audits on customers who signal they want to downsize. That's not product investment. It's friction by design.

Worth being honest about the other side too: building your own alternative isn't free. LLM API costs, engineering time, ongoing maintenance — it adds up. The math needs to be done per tool, not assumed.

A vendor worth keeping is one where AI is making the product do things you couldn't replicate cheaply even if you tried.

Where to start

Don't try to audit the whole stack at once. Start with the two or three tools where you've complained about pricing in the last six months. Those are the ones worth examining first.

For each of those, ask: what does this actually do for us? Could an AI-assisted internal tool cover 80% of it at lower cost? Is this vendor making it easier or harder to eventually leave?

You don't need to cancel anything right now. But knowing the answers changes how you negotiate renewals and where you put engineering time.

The SaaSpocalypse isn't arriving in three years. Vendors are already building their strategy around it. Worth thinking about yours.