The VMware Lesson

In a New York court filing, AT&T, a company that negotiates for a living, stated that its VMware renewal under Broadcom's new licensing model amounted to a price increase of roughly 1,050%. Not fifty percent. Not double. Ten and a half times, for software already running, holding up systems that included services used by first responders, which is the detail that got the lawsuit its headlines.

The AT&T number is the famous one, and it isn't the ceiling: European cloud providers organized under CISPE reported member increases between 800% and 1,500%. This post is about why those numbers are the size they are, because the answer isn't greed as a personality trait. The answer is arithmetic performed on an install base that couldn't leave, and that arithmetic is the single most instructive case study available for anyone who signs infrastructure contracts. The lesson isn't "avoid Broadcom". It's that every locked-in vendor relationship you hold is an asset on someone's future balance sheet, priced at exactly your inability to exit.

The timeline of a repricing

The speed is the tell. Broadcom closed its roughly $69 billion VMware acquisition on November 22, 2023, and nineteen days later announced the end of perpetual licenses, the licensing model on which two decades of enterprise datacenters had been built. Products were fused into a handful of subscription bundles, so customers wanting one component now bought suites. The partner program was culled. Renewals began arriving at multiples of prior spend, and when customers balked, the follow-through included cease-and-desist letters to lapsed subscribers and litigation involving names as large as Siemens' US operations, while AT&T's suit alleged bad-faith pressure to convert to subscriptions it didn't want.

No exploratory phase, no gradual drift: a playbook executed in weeks, which tells you it was written before the ink dried. And by the only measure Broadcom answers to, it worked; the strategy has paid off financially, whatever the customer satisfaction surveys say. That's the part to absorb without flinching. This wasn't a miscalculation being punished by the market. It was a calculation being rewarded by it.

Why the numbers are that size

Here's the mechanism, and it's the same one this blog keeps circling from different directions. A hypervisor is close to the deepest possible position in the lock-in taxonomy: it sits under everything, touching every VM, every runbook, every disaster-recovery plan and every ops engineer's muscle memory, so leaving it isn't a migration, it's a re-platforming of the entire estate. Broadcom's pricing team didn't pick 1,000% out of spite. They estimated the cost of that re-platforming, in money, time, and risk, and priced renewals just under it, because any price below your cost of leaving is, in cold economic terms, a price you'll pay.

That's the general formula hiding inside the case study: your renewal price converges on your exit cost. Not on the product's value, not on what you paid last year, not on the relationship. On what leaving would cost you, as estimated by a party with every incentive to estimate it well. Most vendors don't run the formula to its limit because reputation, competition, and growth still discipline them. An acquirer buying a mature install base has exactly none of those constraints; the reputational damage lands on a brand it didn't build, the growth story is the extraction itself, and the customers, by construction, can't leave quickly. The contract terms AT&T thought protected it turned into two years of litigation instead of protection, which is what contracts are worth against a counterparty that has priced your litigation budget too.

The lesson, generalized

So run the VMware scenario against your own stack, because the scenario is portable. Every "strategic vendor" you depend on, virtualization, observability, identity, database, CI, is one acquisition announcement away from new ownership with a spreadsheet, and the consolidation era guarantees a steady supply of announcements. The questions that matter are the ones this case answers: What would a rational new owner charge us, given our real cost of leaving? How long would leaving actually take, tested, not asserted? Which dependencies sit so deep that their exit cost is effectively infinite, and did we price that when we adopted them?

Notice the VMware exodus that did happen: the customers who moved fastest weren't the angriest, they were the ones whose environments were already portable, boring, and mapped. Everyone else discovered that fury is not a migration plan. The time to compress your exit cost is while your vendor still has competitors and you still have alternatives, because the day the acquisition closes, your exit cost stops being an internal metric and becomes someone else's revenue target.

The steelman: Broadcom's defense has real parts

The uncomfortable defense deserves an honest hearing. VMware's perpetual-license pricing arguably underpriced two decades of critical software; subscriptions are the industry norm everywhere else; the product portfolio genuinely needed rationalization, and some percentage of the outrage is customers discovering the market value of something they'd been getting cheap. Markets also responded exactly as textbooks predict: virtualization alternatives, from Proxmox to the cloud providers themselves, have feasted on the discontent, and in five years the ecosystem may be healthier for the shock. Broadcom, on this reading, just marked an asset to market.

Mostly conceded, and it sharpens the lesson rather than blunting it. If the repricing was rational and legal, and it was, then it's repeatable, and "our vendor would never" is not a control. The only variable you govern in that equation is the exit cost, which means portability isn't a nice-to-have on some architectural wishlist; it's the one lever that caps what any future owner of any vendor can charge you. We build for that lever, and this case is why we get to skip the abstract argument for it. The concrete version fits in one question, and AT&T's lawyers can tell you the price of answering it late: if your deepest vendor were acquired this quarter, what number would appear on your next renewal, and what number would it cost you to say no?