Executive Summary
Enterprise infrastructure stopped being maintenance work. Broadcom’s VMware acquisition turned a quiet licensing decision into the largest migration wave in a decade, and AI arrived before most teams finished answering it. The two problems are the same decision made twice.
Broadcom bought VMware in November 2023, killed perpetual licenses, and pushed the installed base toward subscriptions at higher prices. Every VMware customer now weighs staying against moving, and there is no free option. Meanwhile one GPU node costs more than the rest of the cluster, and training jobs run for days without interruption. Power is the constraint, not compute, because data centers pour concrete faster than utilities deliver megawatts. If 95 percent of new AI deployments run on Kubernetes by 2028, the platform choice becomes a capacity choice. Watch the exit terms. Lock-in shows up in egress fees, replatforming cost, and sovereignty questions. Teams that treat this as a vendor swap will repeat the same bet on a different logo.
The conversation about enterprise infrastructure used to be a quiet one. Choose a hypervisor, pick a cloud, renew the licenses, move on. It was maintenance work. Nobody wrote opinion columns about it.
That changed in November 2023. Broadcom bought VMware, killed perpetual licenses, and pushed the whole installed base toward subscriptions and higher prices. The largest infrastructure migration wave in a decade started the same week. It has not stopped.
The migration wave is bigger than the headlines
Every organization that ran on VMware is now answering the same question. Stay and pay the new price, or move and absorb the risk. There is no free option, and that is what makes this decision so hard.
The escape routes are getting good. Open-source virtualization, Kubernetes-first platforms, and managed clouds have all matured since the acquisition. The tools to move are no longer the blocker. The strategy is.
But the migration is only half the story. You cannot rebuild your platform on a new foundation and ignore what is coming next. And what is coming next is AI.
AI just quietly rewired the cost model
You do not need to be a hyperscaler to feel it. A GPU node costs more than the rest of your cluster. Training jobs run for days and cannot be interrupted. The scheduler that treats every pod as disposable was never designed for hardware that bills by the hour and costs a fortune.
Power is the new constraint, not compute. Data centers can pour concrete faster than utilities can bring megawatts to the door. If 95% of new AI deployments run on Kubernetes by 2028, as one analyst report projects, the platform decision becomes an AI capacity decision.
This is where the migration and the AI shift collide. You are choosing a foundation to carry your old workloads and your new AI workloads at once. It is not two decisions. It is one, and the timing is unforgiving.
The teams that handle this well are the ones that stop treating it as a vendor swap. They treat it as a chance to fix the structural problems they have been tolerating for years. The teams that handle it badly make the same bet again, on a different logo, and hope the next acquisition does not come. That is the real risk. Not the price. Not the platform. The repetition.
Where the real decisions live
Most of the analysis you will read stops at the headline. Choose this vendor over that one. It misses what actually decides the outcome.
Which platform limits your exit options later? What does lock-in really cost in egress fees and replatforming? Who owns your data when a sovereign or regulatory question arrives? How do you keep a platform team productive without handing the keys to one vendor?
Those are the questions worth answering before you write the next budget. And they are exactly the ones we dug into for our 2026 State of Enterprise Infrastructure report. We mapped the migration wave, the AI capacity crunch, container and platform maturity, the real price of lock-in, and the sovereignty questions that will not go away. Then we turned it into a practical roadmap for platform teams.
It is a lot to fit in a blog post, which is why it lives in the report. Get it, plus our weekly briefings on the news that actually matters, by grabbing the State of Enterprise Infrastructure below.
The next five years are decided by the platform choices you make now. Not the ones your vendors want you to make. Get the report, read it on your own terms, and go into that decision armed.
The economics behind this migration wave, including the three year cost model and why exits run long, are in the 2026 State of Enterprise Infrastructure report.
Every announced commitment in this space is tracked with its source in our AI data centre power commitments record.
Related reading. Cloud Native Transformation Is an Operating Model Change, Not a Lift and Shift. A Full VMware Exit Costs More Than Three Years of Broadcom’s Bill. The VMware Exit Has a Shortlist. Nutanix Just Took a Front Seat.. The 2026 State of Enterprise Infrastructure.
Get the next one before it is old news
Independent analysis of cloud-native infrastructure, Kubernetes and data centre economics. No vendor spin.