This article is based on reporting originally published at it.toolbox.com (February 2020). Some figures and product details reflect the state of the market at that time.
A Pricing Shift That Caught Enterprise IT Off Guard
As chip manufacturers continue to pack ever-higher core counts into modern CPUs, VMware announced a fundamental change to how it prices its vSphere hypervisor product line. The move — from a per-socket licensing model to a per-core model — landed as a significant cost shock for many enterprise IT teams, and it reignited a debate that has simmered for years: is the economics of on-premises virtualization still competitive with the public cloud?
The new fee structure took effect in April 2020, applying to VMware’s hypervisor products that allow multiple operating systems to run on a single physical server. For organizations with large fleets of high-core-count servers, the change translated directly into materially higher licensing bills.
How the New Licensing Model Works
Under the legacy model, VMware charged per CPU socket — essentially, per physical processor slot on the motherboard. A dual-socket server required two licenses regardless of how many cores each CPU contained.
Under the new core-based model:
- One license covers up to 32 cores per CPU.
- A CPU with more than 32 cores requires a second license for that socket.
- These licensing costs are in addition to any fees for software managing physical and virtual machine operations.
VMware positioned the change as bringing its pricing in line with how operating system vendors and database software companies — such as Oracle and Microsoft — have long charged for enterprise software: by core count rather than by socket. Base pricing for the enterprise edition of vSphere was reported at around $3,500 per month, with optional features and support packages potentially increasing that figure by an order of magnitude.
Why Core Counts Matter Now More Than Ever
The timing of VMware’s change is directly tied to a hardware trend that shows no signs of reversing: CPU vendors are aggressively increasing per-socket core counts, squeezing more parallel compute threads onto a single die.
At the time of this pricing change, several high-core-count platforms were either shipping or in active development:
- Intel Ice Lake and Cooper Lake: Targeting 38 and 48 cores per socket, respectively, in the server segment.
- AMD EPYC Rome: Already shipping with up to 64 cores per socket, making dual-license requirements almost certain for any EPYC-based host.
- Ampere Computing: A startup developing an 80-core Arm-based processor aimed at cloud infrastructure operators.
- Arm Cortex server chips: Arm itself was pushing into the data center with 32-core Cortex-based designs.
- Huawei Kunpeng: A 64-core Arm-licensed chip launched for data center workloads.
- AWS Graviton2: Amazon Web Services’ custom Arm ISA implementation, used internally across its hyperscale fleet.
The practical implication: any enterprise running or planning to run AMD EPYC or upcoming Intel platforms would automatically cross the 32-core threshold and incur double licensing costs per socket. For organizations with hundreds or thousands of host servers, that arithmetic compounds quickly.
VMware’s Market Position and Revenue Context
VMware, then a Dell EMC subsidiary listed on Nasdaq, held an overwhelmingly dominant position in the enterprise data center hypervisor market. Industry analysts estimated that VMware hypervisors were running on approximately three-quarters of all data center servers at the time. License revenue represented roughly 40% of the company’s reported $9 billion annual revenue in fiscal year 2019.
The transition to core-based licensing was therefore not simply a pricing policy adjustment — it was a significant revenue strategy. As chip vendors compete to deliver more cores per socket, VMware’s new model is designed to capture a share of that performance growth in the form of additional license fees, rather than allowing customers to benefit from hardware improvements within a fixed per-socket cost envelope.
Customer Discontent and Transition Incentives
VMware appeared aware that the change would generate friction. To soften the transition, the company offered a grace period: customers who purchased 32-core licenses before the end of April 2020 could receive additional licenses to cover their existing server estate at no extra charge — a one-time concession for organizations that had already committed their 2020 IT budgets under the old model.
Despite this, criticism from the enterprise IT community was swift and pointed. Key complaints included:
- Perceived favoritism toward Intel: Critics noted that most Intel server CPUs at the time fell below the 32-core threshold, meaning Intel-based customers faced a smaller pricing impact than those running AMD EPYC hardware. This raised questions about whether the licensing structure indirectly steered customers toward a particular chipmaker.
- Justification language: Some IT professionals challenged the framing of the change as an “industry alignment,” arguing it obscured a straightforward price increase for a large portion of the customer base.
- Budgeting disruption: Enterprise procurement cycles operate on annual or multi-year timelines. An unplanned licensing cost increase mid-cycle forced difficult conversations between IT and finance teams across many organizations.
The Cloud Migration Calculus
The pricing change arrives at an awkward moment for the case for on-premises virtualization. Public cloud providers — AWS, Microsoft Azure, Google Cloud — have long argued that the total cost of ownership for on-premises infrastructure, when license fees, hardware refresh cycles, power, cooling, and staffing are factored in, is comparable to or higher than cloud consumption pricing for many workloads.
VMware’s per-core fee increase hands cloud vendors a concrete, quantifiable talking point. A cloud account executive can now present a specific scenario: a customer running dual-socket AMD EPYC servers who previously paid for two vSphere licenses will now pay for four — and can compare that directly against the cost of equivalent compute capacity consumed elastically from a public cloud provider.
Dell EMC also operates its own cloud hosting services covering compute, storage, and networking. The pricing pressure VMware creates for third-party managed service providers and colocation vendors that run VMware in their own data centers could reshape competitive dynamics in that market as well, potentially funneling enterprise workloads toward Dell’s own cloud offerings.
Key Takeaways for DevOps and Infrastructure Teams
- Audit your host topology now. Identify every physical host running VMware hypervisors and document the exact core count per socket. Any CPU exceeding 32 cores triggers an additional license requirement.
- Model the cost delta. Calculate the difference between your current licensing spend and projected spend under core-based pricing, accounting for planned hardware refreshes over the next three to five years.
- Include cloud in the TCO comparison. Use the updated VMware costs as a forcing function to run a genuine total cost of ownership comparison against equivalent workloads hosted on public cloud or a managed VMware-on-cloud offering.
- Watch the chipmaker roadmap. Core counts are not plateauing. Planning infrastructure purchases without accounting for the licensing cost of future high-core-count CPUs is a budgeting risk.
- Engage your VMware account team. Large enterprise accounts may have negotiating leverage, particularly if the pricing change creates a credible migration threat.
The broader question the industry will continue to work through is whether VMware’s model encourages the digital transformation its parent company Dell EMC promotes — or whether it inadvertently accelerates the migration of workloads to public cloud platforms that do not carry the same per-core hypervisor licensing overhead.