David Davis

Price out three years of Kubernetes for a 200-node estate and the control-plane fees are a rounding error — EKS charges $0.10 per cluster per hour, about $876 a year per cluster, and GKE matches it. What actually moves the number is people. The teams I see win this decision stopped comparing feature checklists and started comparing headcount: how many platform engineers does each option require you to hire, and what does the vendor’s subscription let you not build?

That framing changes the answer. This brief compares Red Hat OpenShift, Amazon EKS, and Google Kubernetes Engine as they stand in mid-2026 — list-price shape, operating burden, the edge and VM tiebreakers — and gives you thresholds you can defend in a budget meeting.

The verdict

If you are all-in on AWS with a lean platform team and no VM estate, EKS is the cheapest credible answer — likely by millions over three years at scale. If you run multi-cloud or on-prem, carry FedRAMP High or PCI obligations, have edge sites, or are migrating VMs off VMware, OpenShift’s subscription premium buys you a bundled platform and a single support throat that would otherwise cost you a team. GKE sits between them: the strongest managed Kubernetes engineering in the market, and the best default if Google Cloud is already your center of gravity — especially for data and AI workloads. None of these is a wrong answer. Buying the wrong one for your team’s size is.

Side-by-side comparison

The headcount equation

Here is the math that matters. EKS control-plane fees on twenty clusters run roughly $17,500 a year. A fully loaded platform engineer costs $200,000 or more. The Kubernetes bill was never the bill — the team is the bill.

EKS hands you a conformant control plane and expects you to assemble the rest: ingress, observability, policy, registry hygiene, GitOps, cluster lifecycle. AWS-native shops that standardize hard on managed services — and increasingly on EKS Auto Mode, which manages nodes for a percentage surcharge on instance costs — can run large estates with three to five platform engineers. That is where the “millions cheaper over three years” claim is real, and I have watched CFOs approve it on exactly that arithmetic.

OpenShift inverts the trade. The subscription is a genuine premium — per-core pricing on a few hundred cores lands in six figures annually, and list pricing varies enough that you should never sign without negotiating. But Platform Plus bundles cluster management (ACM), security scanning (ACS), a registry (Quay), pipelines, and one support contract across all of it. The rule of thumb I give clients: if declining OpenShift means hiring eight or more platform engineers to assemble and operate the equivalent stack — common once you cross roughly 30 clusters, multiple environments, or serious compliance scope — the subscription is the cheaper option. Below that, it usually is not.

The takeaway for the meeting: price the platform team, not the platform. Then optimize the workload spend separately — our Kubernetes cost optimization guide covers where the real waste hides.

Honest downsides, all three

OpenShift

The premium is real and recurring, and it scales with cores whether or not you use the bundle. Upgrades of the integrated stack demand discipline — you are operating a distribution, not consuming a service, unless you pay more for ROSA or ARO. Teams that buy Platform Plus and use ten percent of it are Red Hat’s best customers and their own worst enemies.

EKS

The assembly tax is paid in engineering time, and the extended-support fee is a quiet trap: fall more than 14 months behind on Kubernetes versions and the cluster fee jumps sixfold to $0.60 an hour. Multiply by dozens of neglected clusters and you have bought a platform engineer’s salary in penalty fees. EKS is also the deepest lock-in of the three — everything you assemble around it is AWS-shaped.

GKE

Autopilot’s per-pod resource billing punishes lazy resource requests — over-provisioned pods are billed as requested, not as used. Google Cloud’s enterprise sales and support motion still trails AWS and Red Hat in regulated accounts, and if your estate is not on Google Cloud, GKE’s excellence is mostly academic.

GKE Autopilot and the ops-cost angle

Autopilot deserves its own section because it attacks the same headcount problem from the opposite direction. Instead of bundling tooling like Red Hat, Google removes the node layer entirely: you pay for pod-level CPU and memory requests, Google patches and scales the nodes, and your team never SSHes into anything. For general-purpose services with well-tuned requests, that converts two or three node-wrangling engineers into zero. GKE’s $74 monthly free-tier credit even covers one small cluster’s management fee outright.

The catch is discipline. Autopilot is priced on what you request, so sloppy requests inflate the bill faster than on node-based billing, and workloads needing exotic node configs still push you back to Standard mode. GKE Enterprise, billed around $0.008 per vCPU per hour, adds fleet management and policy across clusters — worth modeling only above roughly ten clusters. Verdict: Autopilot is the lowest-ops managed Kubernetes you can buy in 2026, and AWS’s EKS Auto Mode is a direct acknowledgment of that.

Edge and VMs: OpenShift’s tiebreakers

Two workload classes break the managed-cloud logic entirely, and both favor Red Hat.

Edge first. Single-node OpenShift runs a full cluster on one bare-metal box, and MicroShift squeezes an OpenShift-compatible runtime into a single service for devices smaller still. This is not slideware — Verizon runs its 5G RAN on single-node OpenShift deployed across thousands of bare-metal edge locations, managed centrally through Advanced Cluster Management. If your roadmap includes hundreds of retail sites, factories, or cell towers, neither EKS nor GKE has a comparably proven answer.

Second, VMs. OpenShift Virtualization runs virtual machines alongside containers on the same cluster, and as of the 4.19-era releases it added live storage migration and reached the major public clouds in preview or GA form. For enterprises pushed off VMware by renewal pricing, it turns the Kubernetes decision and the hypervisor decision into one purchase — we scored that matchup in our VMware-exit comparison. One operational caution: converged VM-plus-container clusters concentrate risk, so treat data protection as a first-class requirement — see our Kubernetes backup teardown before you migrate the first VM.

What to do about it

• Single-cloud AWS, fewer than five platform engineers, no VM estate: pick EKS, evaluate Auto Mode, and put a version-currency SLO in place so extended-support fees never appear on an invoice.
• GCP-first or data/AI-heavy: pick GKE, default new workloads to Autopilot, and enforce resource-request hygiene from day one.
• Multi-cloud, regulated, edge sites, or a VMware exit in flight: shortlist OpenShift, price Platform Plus against the eight-engineer alternative, and negotiate — nobody pays list.
• Whichever you choose: re-run the headcount math annually. The platforms reprice quietly; your team costs do not.

Frequently asked questions

Is OpenShift cheaper than EKS?

On list price, no — EKS’s $0.10 per cluster-hour is far below OpenShift’s per-core subscription. OpenShift becomes cheaper when its bundled tooling and support replace platform engineers you would otherwise hire, typically at larger, regulated, or multi-cloud scale.

What is the difference between OpenShift and Kubernetes?

Kubernetes is the open-source orchestrator; OpenShift is Red Hat’s distribution of it with an integrated registry, CI/CD, security tooling, a hardened OS, and commercial support. EKS and GKE are managed Kubernetes services, while OpenShift is a full platform you can run anywhere.

Is GKE Autopilot cheaper than Standard GKE?

For spiky or modest workloads with well-tuned resource requests, usually yes, because you stop paying for idle node capacity. For dense, steady workloads on committed-use discounts, Standard mode with good bin-packing often wins on raw cost.

Can OpenShift run virtual machines?

Yes. OpenShift Virtualization runs VMs alongside containers on the same clusters, with live migration and, as of recent releases, live storage migration. It is the feature that makes OpenShift a serious VMware alternative rather than only a container platform.

Which Kubernetes platform is best for regulated industries?

OpenShift generally has the strongest compliance posture — FedRAMP High paths via ROSA on AWS GovCloud, mature FIPS support, and a single vendor accountable for the whole stack. EKS and GKE both offer government regions and compliance programs, but you assemble and attest more of the stack yourself.

Enterprise Techie publishes vendor-honest analysis like this daily — get the brief by email, free.

Barclays’ CIO survey put a number on the mood: 83% of enterprise CIOs plan to repatriate at least some workloads. That figure launched a hundred “the cloud is over” headlines. Here is the number the headlines skip — only about 8% of companies move entire workloads back, and Flexera’s State of the Cloud data suggests roughly a fifth of migrated workloads have returned on-prem in some form. Repatriation is real. It is also selective, and the interesting question is not whether to leave the cloud — it’s which workloads clear the bar.

This brief gives you the actual math: the two workload profiles that consistently win on-prem, the thresholds that decide it, and where AWS, Google Cloud, Broadcom/VMware, and Red Hat sit as the traffic starts to move.

What changed

Three forces converged between 2024 and 2026. First, the FinOps easy wins ran out — after rightsizing, reserved capacity, and spot, steady-state cloud bills stopped shrinking, and finance started asking why a workload with flat utilization needs elastic pricing. Second, GPU economics got loud: with on-demand GPU pricing running roughly $2–8 per hour as of mid-2026, an owned inference cluster starts beating rented capacity once sustained utilization crosses the 60–70% range. Third, the exit tax fell. Under pressure from the EU Data Act, both Google Cloud and AWS now waive egress fees for customers moving data off their platforms — which removed the single largest one-time cost of leaving.

The takeaway for the meeting: the barriers to selective repatriation dropped at the same time the savings case matured. That’s why the survey numbers spiked — not because the cloud stopped working.

The numbers, minus the hype

Most repatriation coverage is either hyperscaler damage control or hosting-vendor cheerleading. Here is what the data actually supports, with the hype version alongside it.

Read the table as a diagnostic. If your estate is mostly variable, the honest savings number for you is near zero. If 40% of your compute runs flat around the clock, the 32% figure is worth a business case.

The workloads that actually leave

Two profiles dominate real repatriation projects, and they share one trait: predictability.

Profile 1: steady-state compute and storage

Databases with stable IOPS, internal line-of-business applications, batch processing, and archive-heavy storage. The pattern that qualifies: utilization that varies less than roughly ±20% month over month, a demand forecast you trust for three years, and a 3–5 year amortization window. These workloads pay cloud’s elasticity premium and never use the elasticity. Moved to owned or hosted infrastructure, they are the source of that ~32% average saving — sometimes more when storage egress was a recurring line item, a pattern we quantified in our egress fee benchmark.

Profile 2: sustained-utilization GPU inference

Training bursts belong in the cloud. Inference that hums along at high utilization does not. At $2–8 per GPU-hour on demand, a cluster you keep above 60–70% utilization typically pays for itself well inside its useful life — as of mid-2026 hardware pricing, commonly within two years. The break-even is sensitive to power costs and your ability to actually keep the cluster busy, so run the model before you buy; our on-prem GPU cluster cost guide walks through the full TCO stack, including the parts vendors leave out.

The workloads that stay

Be equally honest about what doesn’t move, because failed repatriations are expensive twice.

• Spiky and seasonal demand. Retail peaks, tax season, campaign traffic — elasticity is what you’re paying for, and it’s worth it.
• Anything welded to higher-order managed services. If the workload leans on BigQuery, DynamoDB, or a managed streaming stack, the replatforming labor usually eats the infrastructure savings. Count the engineering hours before you count the server savings.
• Globally distributed, latency-sensitive front ends. Replicating a hyperscaler’s edge footprint on your own is not a savings project.
• Anything owned by a team with no ops bench. Repatriation converts an invoice into an operational responsibility. If you don’t have the staff, you don’t have the option.

The verdict a VP can repeat: repatriate the base load, rent the spike.

How AWS and Google Cloud are playing defense

AWS’s response has been pragmatic rather than panicked. It waives egress for customers moving off the platform — conditional, and coordinated through support rather than automatic — and it is leaning on the two levers that keep steady-state workloads in place: deeper commitment discounts and hybrid hardware. Outposts is effectively AWS’s “repatriate without leaving” play, and its European Sovereign Cloud targets the compliance-driven slice of repatriation demand. AWS’s real strength is gravity — the breadth of managed services makes full exits genuinely painful. Its weakness is the flip side: on raw, predictable compute, list pricing loses to owned hardware, and AWS knows it.

Google Cloud moved first on exit fees, announcing free egress for departing customers back in January 2024 — but read the terms: the waiver applies when you close the account and move everything within a 60-day window, which is stricter than the press release implied. Google’s retention story is stickiness of the productive kind: BigQuery and Vertex AI are hard to replicate on-prem, and GKE Enterprise gives it a credible hybrid answer for Kubernetes estates. Where Google is thinner is the on-prem hardware story — there is no Outposts equivalent with real traction, so its repatriation-prone customers tend to leave rather than hybridize.

The landing zone problem: Broadcom/VMware vs Red Hat

Here’s the twist most repatriation math ignores: the default landing zone got expensive at exactly the wrong moment. Broadcom’s overhaul of VMware — subscription-only licensing, the product catalog consolidated down to a handful of bundles around VMware Cloud Foundation — has produced widely reported renewal increases in the 2–5x range for estates pushed from bare vSphere into the full VCF stack, with some customers reporting far worse. The April 2025 announcement of a 72-core license minimum was reversed back to 16 cores after customer pushback, but the episode did lasting damage to renewal psychology. If you repatriate onto VMware without negotiating hard, you can hand your cloud savings straight to Broadcom.

That said, dismiss VCF at your peril. It remains the most complete private-cloud stack on the market, your operations team already knows it, and the ecosystem — backup, DR, monitoring — is unmatched in depth. It fits large estates that consolidate onto dense hosts (fewer phantom cores under the minimums) and treat the renewal as a procurement event, not a formality.

Red Hat is the beneficiary of the moment. OpenShift Virtualization runs VMs on KVM alongside containers on one platform, the migration toolkit for virtualization automates vSphere imports at scale, and OpenShift Virtualization Engine is the VM-only edition aimed squarely at the “I just want a hypervisor without the drama” buyer. Its strengths: one platform for VMs and containers, and no hypervisor renewal shock. Its honest weaknesses: the operational model is Kubernetes-first — vSphere admins face a real learning curve — and the third-party tooling ecosystem around backup and DR, while growing fast, is younger than VMware’s. Fit: organizations already invested in OpenShift or containers, and mid-size estates willing to retrain.

The move: get both quotes. Even if you intend to stay on VMware, a credible OpenShift Virtualization quote is the best negotiating instrument Broadcom has ever given its customers.

What to do about it

• Inventory by utilization curve, not application age. Flat curves are repatriation candidates; volatile curves are not. This one query reframes the whole conversation.
• Price the full exit. Egress waivers help, but the real costs are replatforming labor and landing-zone licensing. A business case without a VMware or Red Hat quote attached is not a business case.
• Set thresholds and hold them. Utilization variance under ±20%, a three-year forecast you believe, GPU utilization above 60–70% — below the bar, stay in the cloud without guilt.
• Rebalance, don’t exit. The 8% doing full exits are mostly special cases. The defensible posture for 2026 is cloud for the spike, owned infrastructure for the base.
• Re-run the math annually. Hardware prices, GPU rates, and hyperscaler discounts all moved in the last eighteen months. A 2024 analysis is already stale.

Frequently asked questions

What is cloud repatriation?

Cloud repatriation is moving workloads from public cloud providers back to on-premises data centers, colocation, or private cloud. In practice it is almost always selective — specific workloads move while the rest of the estate stays in the cloud.

Are companies really leaving the cloud in 2026?

Mostly no — they are rebalancing. Barclays found 83% of CIOs plan to repatriate some workloads, but only around 8% of companies move entire workloads off the cloud. The dominant pattern is moving steady-state workloads back while keeping variable ones in the cloud.

Which workloads should move back on-premises?

Workloads with flat, predictable utilization over a 3–5 year horizon — stable databases, internal apps, batch, and archives — plus GPU inference running above roughly 60–70% sustained utilization. Spiky demand and workloads built on managed services should generally stay.

Does AWS charge egress fees when you leave?

As of mid-2026, AWS waives data transfer fees for customers migrating off the platform, following Google Cloud’s January 2024 move. The waivers are conditional — AWS coordinates them through support, and Google’s requires a full account closure within a 60-day window — so plan the exit before you start it.

Enterprise Techie publishes vendor-honest analysis like this daily — get the brief by email, free.

Move 10 TB out to the internet this month and the bandwidth line alone runs roughly $913 on AWS, $882 on Azure, and $1,137 on Google Cloud’s Premium Tier. The same transfer out of Cloudflare R2 or Backblaze B2 costs approximately zero. That spread — nearly four figures a month on a single line item — is why egress pricing keeps showing up in every multicloud architecture review I sit in on, and why the 2024 “free exit” announcements from Google and AWS confused so many budget owners.

This brief gives you the current per-GB numbers in one table, explains what the exit-fee waivers actually cover (much less than the headlines implied), and closes with thresholds for when egress should change your architecture rather than just your invoice.

The 2026 egress benchmark

These are internet egress rates as of mid-2026, standard list pricing, US regions. Committed-use discounts, CDN offload, and private interconnects all change the math — but this is the baseline every negotiation starts from.

Two footnotes that matter more than the table. First, these are internet egress rates only — cross-availability-zone traffic still bills at about $0.01/GB each way on all three hyperscalers, and NAT gateway processing adds roughly $0.045/GB on AWS. Chatty microservices generate egress-class charges without a single byte leaving the cloud. Second, list pricing varies by region and moves; treat anything beyond two decimal places as negotiable at enterprise scale.

What changed

In early 2024, Google Cloud announced free data transfer for customers leaving its platform, and AWS followed within weeks. Microsoft made a similar commitment for Azure. The driver was regulatory: the EU Data Act, which began applying in September 2025 and requires switching charges to be phased down, with a full ban on profit-bearing switching fees taking effect in January 2027.

Here is the nuance the pricing calculators miss: the waivers cover exit, not operations. AWS’s program, for example, requires that you be moving essentially all your data off the platform, requests go through support, and the credits carry a limited migration window — around 60 days. Google’s works similarly. If you are running a steady-state multicloud architecture — analytics in BigQuery, serving from AWS, backups to a third provider — every gigabyte of that ongoing traffic still bills at the full rates in the table above. The waivers are a one-way door for leavers, not a discount for hybrid operators. That distinction is what actually drives architecture decisions, and it’s why the repatriation math we walked through in our cloud repatriation analysis now excludes exit bandwidth as a blocker: the toll to leave is gone, but the toll to straddle remains.

AWS: predictable tiers, expensive middle

AWS charges $0.09/GB for the first 10 TB of monthly internet egress after a 100 GB free allowance, stepping down through published tiers to $0.05/GB above 150 TB. The strength of the AWS model is predictability and negotiability — the tiers are documented, CloudFront offload is priced aggressively for content-heavy workloads, and at commitment levels egress is a standard line in every Enterprise Discount Program conversation. AWS also deserves credit for making its exit waiver unconditional on destination: you can move to any provider, not just on-premises.

The weakness is the middle of the curve. A workload pushing 20–50 TB/month is too small for meaningful EDP leverage and too large to ignore a $1,700–$4,300 monthly bandwidth line. That band — common for SaaS companies serving media or model artifacts — is exactly where the zero-egress alternatives poach AWS storage workloads. AWS fits teams that value ecosystem depth and can either negotiate scale discounts or push static egress to CloudFront.

Microsoft Azure: cheapest hyperscaler under 50 TB

Azure’s $0.087/GB first tier makes Microsoft the cheapest of the big three below roughly 50 TB/month; above that its tiers converge with AWS at $0.07/GB and then $0.05/GB past 150 TB. For the Microsoft-standardized enterprise the practical advantage is bigger than the headline three-tenths of a cent, because so much traffic that would be billable egress elsewhere stays inside the Microsoft estate — ExpressRoute to on-premises, Microsoft 365 integration, Azure Front Door offload.

Where Azure is weaker is transparency. Its bandwidth pricing page spreads charges across zones and services in a way that makes forecasting harder than on AWS, and inter-region rates vary by geography pairings that surprise teams expanding into Asia-Pacific. Azure fits organizations already committed to Microsoft licensing that can route most traffic over private paths; it is not obviously the value play for a greenfield, internet-facing workload.

Google Cloud: priciest Premium Tier, with an escape hatch

Google Cloud’s default Premium Tier — traffic rides Google’s backbone to the edge — is the most expensive first tier of the three at $0.12/GB for the first 1 TB, stepping down through tiers to roughly $0.05/GB at very high volumes, and its free allowance is negligible. Google was first to waive exit fees, a genuinely pro-customer move that AWS and Microsoft then matched, and its Standard Tier (public-internet routing at around $0.085/GB) is an escape hatch the other two don’t offer as cleanly.

The catch is that Standard Tier trades performance for price — no global backbone, regional-only routing — so latency-sensitive products generally stay on Premium and pay for it. Google Cloud fits data-platform-centric shops where the gravity is BigQuery and Vertex AI and serving egress is modest; it fits poorly as a high-volume content origin unless you deliberately architect around Media CDN or Standard Tier.

The zero-egress challengers

Cloudflare R2 charges nothing for egress, ever — you pay for storage and per-operation fees. Backblaze B2 gives free egress up to three times your average monthly stored data, then about $0.01/GB. Neither is a hyperscaler replacement: you don’t get VPCs, managed databases, or an IAM estate. What they are is a pressure valve for the specific workload class where egress dominates cost — media origins, model weights and artifact distribution, backup restore paths, public datasets.

The rule of thumb I give teams: when annual egress spend on an object-storage workload exceeds the storage spend itself, moving that bucket to R2 or B2 usually pays for the migration inside two quarters. S3-compatible APIs keep the switching cost low, which is precisely why AWS, Microsoft, and Google have not matched the model — zero egress is the challengers’ entire wedge.

What to do about it

Benchmarks are only useful with thresholds attached. Here are the diagnostics I use:

• Egress under 5% of cloud spend: ignore it. Optimize compute and licensing first.
• Egress at 5–15% of spend: tactical fixes — CDN offload for anything static, compress or dedupe transfer paths, kill cross-AZ chatter. Kubernetes clusters are chronic offenders here; topology-aware routing is one of the highest-leverage items in our Kubernetes cost optimization guide.
• Egress above 15% of spend, storage-driven: re-architect. Move the heavy buckets to zero-egress object storage and keep compute where it is.
• Considering a full exit: the waivers mean bandwidth is no longer the blocker — model the migration on labor and refactoring cost instead, and file the waiver request before you start moving data.

The takeaway for the meeting: egress is now a design decision, not a tax. Leaving is free; staying multicloud is not.

Frequently asked questions

Which cloud provider has the cheapest egress fees?

Among hyperscalers, Azure is cheapest below 50 TB/month at $0.087/GB; AWS and Azure converge above that. Cloudflare R2 and Backblaze B2 undercut all three with zero or near-zero egress, at the cost of a much narrower service catalog.

Are AWS egress fees really free now?

Only if you are leaving. AWS waives data transfer charges for customers migrating all their data off the platform, via a support request with a limited migration window. Ordinary operational egress still bills at standard rates starting at $0.09/GB.

Does the EU Data Act ban egress fees?

It bans charges for switching providers, phased in from September 2025 with a full prohibition on profit-bearing switching fees in January 2027. It does not ban operational egress pricing for workloads that stay put.

How do I reduce cloud egress costs without changing providers?

In order of effort: put a CDN in front of static and media traffic, eliminate cross-AZ and NAT gateway paths for high-volume flows, compress data in transit, and move egress-heavy object storage to a zero-egress provider while keeping compute in place.

Enterprise Techie publishes vendor-honest analysis like this daily — get the brief by email, free.

Broadcom renewal quotes are commonly resetting annual VMware spend 100–250% before anyone negotiates — and the old catalog of 8,000-plus SKUs has collapsed into a short list of VCF-led bundles. That combination changes the nature of the renewal meeting. It is no longer a procurement exercise; it is a build-vs-leave decision, and every quarter through 2027 another wave of enterprises hits it as their bridge agreements expire.

This brief compares the exit paths that actually come up in those meetings — Red Hat OpenShift Virtualization, Nutanix AHV, Microsoft Hyper-V, and Proxmox — with the break-even math and the negotiating leverage each one buys you, even if you never migrate a single VM.

What changed

Broadcom finished what it started in 2024. Perpetual licensing is gone, the per-CPU metric became per-core with minimum core counts, and the product line contracted to a handful of subscription bundles anchored on VMware Cloud Foundation. vSphere 9 ships inside VCF 9 — there is no standalone Standard edition path to the current release. Customers who took one-year bridge deals in 2024–25 are now getting their first full-freight quotes, and the reported resets range from roughly double to several multiples of prior spend depending on how discounted the old agreement was.

The strategic read: Broadcom is optimizing for its largest few thousand accounts and is comfortable shedding the long tail. If you are in that long tail, the renewal quote is the eviction notice. If you are a strategic account, the quote is an anchor for a negotiation Broadcom fully expects. Either way, walking in without a costed alternative is the one unforced error.

The verdict up front

There is no drop-in VMware replacement. Pick by destination, not by nostalgia. If your three-year direction is Kubernetes and containerized workloads, OpenShift Virtualization is the strongest strategic bet — you consolidate VMs and containers on one control plane and stop paying for two platforms. If you want the closest operational feel to vSphere with a mature HCI stack, Nutanix AHV is the safest like-for-like move. If you are a heavy Microsoft shop with Windows Server Datacenter licensing already on the books, Hyper-V is close to free at the margin. Proxmox wins on cost for smaller estates and edge sites, with the trade-off of a thinner enterprise support bench. Break-even on migration cost versus the Broadcom delta typically lands 9–14 months out — which means a funded migration plan pays for itself even as a pure negotiating instrument.

Side-by-side: the four exit paths

OpenShift Virtualization: a platform shift, not a hypervisor swap

Red Hat’s positioning is unambiguous. RHV — its old standalone hypervisor — was sunset, and OpenShift Virtualization is the designated successor. Built on the upstream KubeVirt project, it runs VMs as first-class Kubernetes objects next to containers, managed by the same operators, RBAC, and GitOps pipelines. The Migration Toolkit for Virtualization connects to vCenter, discovers inventory, and executes batched cutovers, and Red Hat has been aggressively courting Broadcom refugees with migration assessments and funded pilots.

Be clear-eyed about what you are buying. This is an architectural shift, not a hypervisor swap. vMotion, DRS, and vSAN have functional equivalents — live migration, descheduler-driven rebalancing, ODF or third-party CSI storage — but they are configured differently, and some are less turnkey than the VMware features your team has run for fifteen years. Teams that already operate OpenShift absorb this quickly. Teams with zero Kubernetes muscle should budget six months of enablement before the first production wave, or the migration stalls in pilot. The upside is real: one platform, one skills investment, and a direct on-ramp to the same infrastructure conversation covered in our private AI platform comparison, where OpenShift’s container-native footing pays a second dividend.

The pragmatic paths: AHV, Hyper-V, Proxmox

Nutanix AHV is the path of least operational disruption. The hypervisor is bundled into the Nutanix Cloud Platform licensing, Move handles agentless migrations from ESXi well, and the day-2 experience — one console, integrated DR, one throat to choke — is the closest thing to what vSphere admins already know. The caveats: it is strongest on Nutanix’s HCI model, so estates committed to external SAN arrays need to validate the fit, and you are consciously trading one commercial vendor dependency for another. Nutanix knows why you are shopping and prices accordingly, but multi-year list pricing has been the more predictable of the two — as of mid-2026, that predictability is itself a selling point.

Microsoft Hyper-V is the option most enterprises underrate. If you already license Windows Server Datacenter, the hypervisor rights are effectively paid for, and Azure Arc has quietly become a credible multi-site management layer. Windows-heavy estates can cut the VMware line item to near zero. The honest counterweight: Microsoft’s energy goes where Azure goes, standalone SCVMM tooling feels dated next to vCenter, and Linux-heavy or KVM-ecosystem shops will feel the friction.

Proxmox VE has graduated from lab curiosity to legitimate line item, with built-in ESXi import tooling and subscription support that costs a rounding error next to a VCF quote. For sub-500-VM estates, edge sites, and dev/test tiers it is often the rational answer. At true enterprise scale the ecosystem thins — certified backup integrations, GPU virtualization depth, and 24/7 severity-one support all require more diligence than the incumbents demand.

The honest downsides — both sides

Staying has a defensible case. VCF is genuinely good engineering — the integrated stack, mature HA/DRS behavior, and the deepest third-party ecosystem in the industry are not marketing fiction. If you negotiate the delta down to 30–50% and lock multi-year caps, staying can beat a disruptive migration on three-year cost, especially for estates over a few thousand VMs where migration labor dominates.

• Leaving costs more than the license math suggests: runbooks, monitoring integrations, backup re-platforming, and staff retraining routinely add 20–40% on top of naive migration estimates.
• Staying costs more than the quote suggests: Broadcom has shown it will re-price at every renewal, and each cycle you stay, your negotiating leverage decays as alternatives-readiness goes stale.
• Partial exits are underrated: moving the 40–60% of workloads that are easy — dev/test, stateless tiers, Windows fleets to Hyper-V — shrinks the renewal while sidestepping the hard 10%.

What to do about it

Start 12 months before renewal, not three. Build a real, costed migration plan for one credible alternative — vendor quotes, migration-services estimate, pilot cluster actually running. A funded alternative, even one you never execute, is reliably worth another 10–20% off the Broadcom quote plus multi-year price caps, because Broadcom’s account teams discount against demonstrated flight risk, not against complaints. Then run the break-even: if the annual delta between your old spend and the new quote pays for the migration in 9–14 months — the common range we see — leaving is the financially conservative choice, not the adventurous one. Above 24 months, negotiate hard and stay. In between, do the partial exit. The broader market context — including who is repatriating what — is in our cloud repatriation analysis and the full VMware alternatives field guide.

The line for your steering committee: every VMware renewal is now a build-vs-leave decision, and the cheapest option in the room is the one with a funded Plan B attached.

Frequently asked questions

Is OpenShift Virtualization a direct replacement for VMware vSphere?

No. It runs VMs capably via KubeVirt, but it is a Kubernetes-native operating model, not a vSphere clone. Features like DRS and vSAN have equivalents that work differently and demand Kubernetes skills. Treat it as a platform migration with a VM-hosting outcome, and it succeeds; treat it as a hypervisor swap, and it disappoints.

How much have VMware prices increased under Broadcom?

It varies widely by prior discounting and bundle mapping, but 100–250% resets on renewal are common before negotiation, and outliers run far higher. The shift to per-core subscription bundles means lightly-loaded hosts and previously discounted enterprise agreements see the sharpest jumps.

What happened to Red Hat Virtualization (RHV)?

Red Hat sunset RHV and named OpenShift Virtualization as its successor. Existing RHV customers were given migration paths onto OpenShift, which is why Red Hat’s VMware-exit pitch centers on one converged platform rather than a standalone hypervisor.

Is Proxmox ready for enterprise production workloads?

For small-to-mid estates, edge sites, and non-critical tiers, yes — thousands of organizations run it in production. For large regulated enterprises, the gaps are ecosystem depth and support scale rather than core hypervisor quality, so scope it to the segments of the estate where those gaps do not bite.

How long does a VMware migration actually take?

Plan on 6–18 months end to end for a mid-size estate: one quarter for pilot and tooling validation, then wave-based cutovers. The long pole is rarely moving disks — it is re-platforming backup, DR, and monitoring integrations, and retraining the operations team on the new stack.

Enterprise Techie publishes vendor-honest analysis like this daily — get the brief by email, free.