2024 in Review: The Year Cloud Lock-In Started Losing

If a single thread runs through the cloud-infrastructure stories of 2024, it is the erosion of lock-in. The forces are not coordinated, and they are far from finished — lock-in is structural and deeply entrenched, and one year does not undo it. But across regulation, pricing, licensing, and open-source governance, 2024 was the year that the mechanisms holding workloads and data in place started, visibly and in several directions at once, to lose ground. This is a review of how that happened, threaded through the events we covered as they unfolded.

The Egress Wall Cracked

The most concrete barrier to leaving a cloud has always been the cost of taking your data with you. Egress fees — the charges for moving data out of a provider’s network — turned the act of switching into a bill large enough to deter it on its own. In 2024 that wall cracked.

Early in the year, Google Cloud announced it would waive exit-migration egress fees for customers leaving the platform, citing the EU Data Act explicitly. The question we asked at the time was whether the other hyperscalers would follow — and over the course of the spring, they did, announcing comparable exit-egress changes of their own. The scope of these moves is narrow: they address the cost of fully leaving a provider, not the much larger volume of ordinary day-to-day transfer, and they leave untouched the deeper lock-in that lives in proprietary APIs and managed services. But the symbolism is real. The major providers conceded, under pressure, that the cost of departure should not itself be a barrier to departure. That is a concession that would have been hard to imagine a couple of years ago.

Regulation Set the Direction

The egress moves did not happen in a vacuum. They happened because the EU Data Act entered into force in January, carrying switching provisions designed precisely to remove the barriers — commercial and technical — that make changing cloud providers difficult. The substantive obligations sit on a transition runway and are not yet fully applicable, but the Act’s mere existence reshaped provider behavior ahead of its enforcement, as the egress-fee announcements demonstrated. Regulation acted as a lever on a structural problem that market competition alone had not dislodged: customers had complained about egress fees for years without effect, and it took a regulatory direction of travel to move the incumbents. The Data Act’s portability and interoperability provisions — the parts that address technical, not just commercial, lock-in — are still ahead, which is part of why this story is unfinished.

The VMware Exodus

While regulation pried at the public cloud, a parallel upheaval ran through private infrastructure. Following Broadcom’s acquisition of VMware, 2024 opened with VMware moving to a subscription-only licensing model, ending the perpetual licenses that the bulk of its enormous installed base relied on. The result was the largest-scale re-evaluation of private-cloud infrastructure in years. Organizations that had treated VMware as permanent, unexamined bedrock found themselves pricing alternatives — Proxmox, OpenStack, OpenNebula, Nutanix, Hyper-V — many for the first time.

The significance for the lock-in story is twofold. First, it was a sharp, involuntary lesson in the cost of deep single-vendor dependence on a proprietary platform. Second, it drove real adoption pipelines toward open-source private-cloud alternatives that had seen modest uptake in the preceding years. A captive audience, suddenly motivated, went looking for infrastructure that no single commercial owner could reprice unilaterally — and much of what it found was open.

The Fork Playbook Matured

The same governance anxiety that drove VMware refugees toward open platforms played out in open source itself, where 2024 confirmed that the relicense-and-fork dynamic of the prior year was now an established pattern. In the spring, Redis moved to source-available licensing and was forked as Valkey under the Linux Foundation within days — following almost exactly the template set by the Terraform-to-OpenTofu fork. The sequence is now recognizable as a playbook: a commercially stewarded project relicenses, the community objects on governance grounds, a fork appears quickly, and it lands under neutral foundation governance with backing from the very commercial users the relicense targeted.

What this matured into in 2024 is a clear lesson about where durable trust in infrastructure software lives: not in a single company’s stewardship, however benign, but in neutral governance that no one actor can override. The community has learned that the answer to relicensing risk is foundation governance, and it now reaches for that answer quickly and confidently.

The Common Thread

Read together, these stories describe a single underlying shift. Egress-fee removals attacked the cost of leaving. The Data Act attacked the legal and commercial barriers to leaving. The VMware exodus was a demonstration of what deep proprietary dependence costs when terms change. The fork playbook established neutral governance as the structural guarantee against unilateral control. Different mechanisms, different domains — all pushing in the same direction, against the assumption that workloads and data, once placed, stay placed.

It would be a mistake to overstate this. Lock-in remains the default condition of the cloud. The deepest form of it — applications built around proprietary managed services that have no portable equivalent — was barely touched by anything that happened this year, because no egress policy and no license change addresses an application that is architecturally welded to one provider’s services. The egress changes were narrow. The Data Act’s strongest provisions are still on the runway. The open-source forks address governance risk but not the broader portability problem. What changed in 2024 was the direction, not the destination.

A Distributed-Systems Reflection

The year’s events vindicate a principle the distributed-systems tradition has long held: the resilience and freedom of a system depend on the structure of its dependencies, not merely on their quality. A dependency controlled by a single party — whether a hyperscaler’s proprietary service, a commercial software vendor’s licensing, or a single-company open-source project — carries a structural risk that no amount of present good behavior removes, because the party retains the unilateral ability to change the terms. The countervailing patterns that gained ground in 2024 — neutral governance, portability mandates, open alternatives — are all, at bottom, attempts to distribute control so that no single actor can hold a workload hostage. That is an old idea finding new purchase.

Looking Ahead — Cautiously

Predicting the next year is a mug’s game, so this is offered tentatively rather than confidently. The plausible direction is continuation: the Data Act’s stronger provisions moving toward applicability, the VMware re-evaluation continuing to feed open-platform adoption, and the fork playbook standing ready for the next relicense. The hardest and most interesting question is whether anything will start to erode the deepest lock-in — the proprietary-managed-service coupling that this year’s developments left almost entirely intact. That would require either standardization that makes managed services portable or a shift in how organizations architect against them, and there is no strong reason to expect it imminently. We would rather flag it as the open question than pretend to forecast its answer.

What 2024 established is that lock-in is contestable — that regulation, market pressure, and community governance can push against it, in concert, with visible effect. That is a meaningful change in the weather, even if the climate has not yet turned.