Infrastructure vendor platform risk: what the VMware pricing shock reveals

When Broadcom closed its $61 billion acquisition of VMware in November 2023, tens of thousands of enterprises discovered what infrastructure vendor platform risk actually costs at scale. A UK university saw its annual VMware support bill climb from £40,000 to £500,000. A company running 20 ESXi hosts received a renewal quote 400% higher than the previous year. Documented price increases across the customer base ran from 150% to 1,500%. The shock was real, but it wasn't unpredictable. Four signals were visible years before the deal closed. Understanding them is how infrastructure teams avoid the same outcome with their next vendor.

What Broadcom actually did

Hock Tan's infrastructure acquisition playbook has been consistent across three deals. At CA Technologies ($18.9 billion, 2018) and Symantec Enterprise Security ($10.7 billion, 2019), the pattern held: identify a business with a dominant installed base and high customer switching costs, cut R&D aggressively, consolidate product lines, and raise prices on the customers who cannot easily leave. VMware, at $61 billion, applied the same model at much larger scale.

The structural changes after the acquisition closed were systematic. Perpetual licensing was eliminated. Every customer moved to subscriptions. VMware's 8,000-plus product SKUs collapsed into four: VMware Cloud Foundation, vSphere Foundation, vSphere Standard (discontinued July 2025), and vSphere Enterprise Plus. Customers who had purchased individual components (only vSAN, only NSX) were required to buy full VMware Cloud Foundation bundles regardless of what they actually used. A 20% late renewal surcharge was added automatically for customers who missed their anniversary date.

By mid-2025, 87% of VMware customers were actively reducing their footprint, according to CloudBolt's industry survey. The average customer renewed only 25% of their previous estate. Broadcom's infrastructure software division still grew revenue, reaching 77% operating margins, up from 67% the prior year. That growth came from extracting more per customer from a shrinking base, not from adding customers. Broadcom publicly acknowledged a strategy of absorbing churn in exchange for margin extraction from the accounts that remained.

Four signals that were visible before the deal closed

The outcome did not require inside knowledge. Hock Tan's prior acquisitions had documented the playbook. Infrastructure architects who had watched CA Technologies and Symantec Enterprise had a working model before VMware was even announced.

Signal 1: Study the acquirer's prior record, not the roadmap

Before an acquisition closes, most affected customers study the announced product roadmap commitments and integration plans. Fewer study what the acquirer did to its previous targets. At CA Technologies post-2018, Broadcom cut 2,000 employees, consolidated mainframe product lines, and raised prices on customers with no exit option. At Symantec Enterprise, Hock Tan described the deal explicitly as an opportunity to apply 'focused cost reduction, premium customer retention, and monetisation of a dominant installed base.' That language appeared in public statements in 2019. VMware customers who read it had four years to adjust their architecture and contracts.

Signal 2: Infrastructure versus software

Not all enterprise software carries the same switching cost. A project management tool or CRM can be replaced in months, with portable data and documented APIs. VMware was structurally different: the compliance approvals, team certifications, backup tooling, networking configuration, and thousands of VMs all depended on vSphere continuing to operate. Products that function as infrastructure — where the migration timeline runs 18-48 months and every adjacent system was built assuming the vendor's continued presence — carry a different category of lock-in risk than application software at the edges.

Identifying whether a product is 'infrastructure' in this sense is straightforward: how long would it actually take to replace it starting tomorrow? Answers under 12 months are software. Answers over 12 months are infrastructure, and infrastructure vendors warrant scrutiny before every renewal, not just when prices move.

Signal 3: R&D investment trajectory

A vendor reducing R&D relative to revenue is harvesting what it has, not building what's next. The financial profile looks attractive: margins expand, cash flow improves, revenue stays flat or grows slightly. But customers on that vendor are watching their platform fall further behind alternatives. The pattern appeared at CA Technologies post-acquisition: product investment declined while prices on legacy mainframe tools increased. For public companies, R&D as a percentage of revenue is in every quarterly filing. For private vendors, watch product release cadence and job posting patterns. A vendor posting primarily for sales and support roles while engineering headcount stagnates is showing harvest-mode signals.

Signal 4: Absence of a quick exit

Price increases sustain when customers cannot leave fast enough. Broadcom could raise prices 800-1,500% because, at enterprise scale, a comparable VMware alternative was not fast or cheap to adopt when the acquisition closed. Nutanix AHV was approaching feature parity but hadn't fully arrived. Microsoft Hyper-V served Windows-centric shops. Open-source KVM and Proxmox required operational maturity most enterprise teams had not built. Migration paths existed, but they were slow and expensive. The absence of a fast, affordable exit is not just a technical fact — it is a direct measure of your negotiating position at every future renewal.

Where enterprises actually went

Three years after the acquisition, the migration picture is clearer, and more complicated than the headlines suggest.

Gartner projected in September 2025 that 35% of VMware workloads (roughly 140,000 customers) would migrate by 2028. Among IT leaders surveyed in 2026, 74% are actively exploring or piloting alternatives. But 'exploring' and 'migrated' are different states. Large-scale migrations involving 2,000 or more VMs realistically take 18-48 months. Most enterprises are mid-transition, running reduced VMware footprints while moving workloads incrementally. The organisations moving fastest are those that had already been containerising workloads or had maintained multi-hypervisor skills before the acquisition closed.

Five layers where lock-in accumulates

Understanding why VMware migrations take years is useful for thinking about vendor lock-in in general. The technical layer (actually moving the VMs) is not the hardest part.

Technical layer

VM formats, hypervisor-specific management APIs, proprietary networking (NSX), proprietary storage (vSAN). Converting workloads to a new format is tractable with time and tooling. This layer is the most visible and usually the most overestimated hurdle.

Operational layer

Your team holds VMware certifications. Your runbooks reference vSphere operations. Your on-call procedures were written around VMware's management console. Rebuilding the operational knowledge base takes longer than migrating the workloads, and cannot be parallelised as efficiently as the technical work.

Compliance layer

SOC 2 evidence, ISO 27001 control documentation, and internal security policies were built against a specific VMware configuration. Migrating to a different hypervisor requires re-attesting those controls — a process that runs six to twelve months on its own schedule, independent of the technical migration.

Contractual layer

Perpetual licences gave customers a slower-burning exit option: stop renewing support, stay on the current version, migrate at your own pace. Once perpetual licensing was eliminated and subscription became mandatory, each renewal cycle tightened the dependency. Subscription models make staying cheap in the short term and expensive to exit over time.

Ecosystem layer

Backup vendors, monitoring tools, and partner integrations all built against VMware APIs over two decades. Migrating the hypervisor requires auditing and replacing that toolchain alongside the core platform. No current alternative carries VMware's partner ecosystem. Nutanix has part of it. Proxmox has significantly less. The ecosystem layer is the one that catches teams by surprise at the point of migration, not during the initial assessment.

Reading infrastructure vendor platform risk in your current stack

The VMware situation will not repeat in identical form. The pattern — dominant installed base, high switching cost, financially-driven acquisition, R&D decline, price extraction — does appear in enterprise software more broadly. A basic vendor risk audit takes a few days of architecture time and costs less than a single year's unexpectedly-tripled renewal.

Start with the acquirer question, applied prospectively and retroactively. For every critical infrastructure vendor, find who owns it and what that owner has done to previous acquisitions. The pattern shows up in SEC filings, press releases, and analyst commentary on earlier deals. If the acquirer has a documented track record of acquisition-mode price extraction, that is your baseline expectation for your vendor — not the roadmap commitments made at closing.

Build a switching cost inventory. Three columns cover the essentials: vendor, realistic migration timeline if you started this quarter, and rough migration cost order of magnitude. Any vendor with a timeline over 12 months and migration cost above one year's contract value belongs in the high-risk category. That list needs contractual portability provisions before the next renewal.

Negotiate portability into contracts, not just price protection. Annual increase caps are useful. Data export commitments, API stability guarantees, and explicit exit support clauses matter equally. A vendor willing to document their migration-assist process has lower effective lock-in than one who won't, regardless of what the pricing schedule says.

Monitor R&D signals quarterly. For public companies, R&D as a percentage of revenue is in every earnings report. Declining ratios over two to three consecutive quarters warrant a conversation. For private vendors, watch product release cadence and engineering job posting patterns over 12-month windows. Stagnating engineering headcount while sales and support hiring grows is a harvest-mode signal worth logging.

Build portability into the architecture before you need it. Container workloads are more portable across hypervisors than bare-metal deployments. Open standards for identity (OAuth 2.0, SCIM), storage, and networking reduce dependency on vendor-specific implementations. The enterprises with the smoothest VMware exit paths had already containerised some workloads and avoided deep NSX dependencies — architectural decisions made years earlier, not contract clauses signed last quarter.

What the migration wave leaves behind

Broadcom's infrastructure software division achieved 77% operating margins in 2025 by raising prices on customers who were sufficiently locked in that 800-1,500% increases were still cheaper than the exit. The model worked financially for Broadcom's shareholders and expensively for the enterprises on the other side.

Migrations will complete over the next two to three years. Some enterprises will land on Nutanix, some on Proxmox, some on hyperscaler-managed VMware services with a different form of vendor dependency. The consensus among infrastructure architects coming out the other side is not complex: single-vendor critical infrastructure dependency is a structural risk that looks invisible until the pricing review arrives. The signals that predict it are measurable before that happens. Running the analysis costs substantially less than not running it.