This week two things happened on the same day. VMware confirmed it will release an updated vSphere Standard, and Proxmox announced 24/7 enterprise support plus a North American subsidiary. There is no indication that either was a response to the other. Together they say more about the lower end of the virtualization market than almost anything on the Explore main stage.
Start with VMware, because this is the one I did not expect.
Paul Turner, chief product officer for VMware’s Cloud Foundation Division, told The Register at Explore in Las Vegas that the sales incentives which pushed customers toward VCF have been changed. He said, “We corrected this,” and acknowledged that the focus on VCF had at times been too heavy. Ram Velaga, president of Broadcom’s Infrastructure Software Group, described the original decision differently. Concentrating on VCF was meant to make the case that private cloud could be cheaper and easier to operate than public cloud, and shipping a smaller server virtualization product in the middle of that argument risked confusing customers about how committed VMware was to private cloud.
I have been critical of most of what came out of Broadcom after the acquisition closed. The subscription-only switch, the channel cuts, the way small and mid-size customers were quoted VCF or nothing. The same standard applies in the other direction, so I will say plainly that bringing back a current vSphere Standard is the right call. It is late, and it is good news for a large number of customers who were running out of realistic VMware options.
vSphere Standard was never a niche edition
vSphere 8.0 reached general availability on 11 October 2022. In 2025, VMware shipped vSphere 9 as part of VCF 9 with no corresponding release of vSphere Standard. Turner says that was deliberate, while engineering effort went into VCF and security.
Licensing advisors and surviving partners agree that vSphere Standard and Enterprise Plus stayed on the 8 Update 3 branch instead of moving to version 9. They do not agree on whether the editions remained commercially available. Some describe them as frozen but still sellable; others cite an end of sale on 31 July 2025 with no renewals. Broadcom’s own position, given to The Register in December 2025, was that it had announced no change to vSphere Standard availability in EMEA and no end of support, and that the product remained fully available there. That conflict is unresolved.
The same reporting puts vSphere Standard at roughly 60 percent of Broadcom’s VMware licenses. If that is accurate, this was never a small edition at the edge of the portfolio. It was the majority of the licensing base, sitting on a 2022 branch while VCF moved forward.
The commercial path narrowed at the same time
The product freeze did not happen in a vacuum. Perpetual sales stopped on 11 December 2023. Partners were notified on 22 December 2023 that their agreements would terminate on 4 February 2024, ending Partner Connect and its incentive structure and replacing it with the invitation-only Broadcom Advantage Partner Program. At the time, reporting put the invitation threshold near 500,000 USD in annual VMware revenue, which excluded many of the partners serving exactly the customers vSphere Standard was built for.
Cloud providers went through a separate restructuring. Their agreements were terminated on 24 April 2024, the Advantage program for cloud service providers ended on 31 October 2025, and remaining VCSP contracts were not renewed after 26 January 2026. One recent report puts the number of authorized US cloud providers under the new structure at about nineteen, compared with more than a thousand before the acquisition.
Licensing moved too. A 72-core minimum order was announced in spring 2025 and withdrawn roughly two weeks later after the reaction, leaving the 16-core-per-CPU minimum in place. VVF stopped being available in parts of EMEA in December 2025. Broadcom publishes no normal public price list for VMware, but licensing advisors commonly place VCF at around 350 USD per core per year.
Where the 128-core figure comes from
Velaga positioned the revived vSphere Standard at customers running around 128 cores. That looks less like workload sizing than like licensing arithmetic. Four dual-socket hosts at the 16-core-per-CPU minimum equals 128 billable cores, and the physical processors don’t need to contain 128 cores between them. The floor gets you there on its own.
At the figure advisors quote, that estate on VCF runs to roughly 45,000 USD per year. Reasonable for an organization actually using the stack. Much harder to justify for a customer who needs vSphere, HA, vMotion, and conventional shared storage, and has no intention of deploying vSAN, NSX, or the automation layers. That is the gap the revived SKU exists to fill, and for four years there was nothing in it.
The timing is complicated. I do not personally know anyone still running vSphere Standard today, and what I have watched in the market over the past six months is organizations consolidating to three or four hosts specifically to control license cost, only to discover the per-core minimum does not shrink with them. Set that against 60 percent of licenses, and the two do not sit comfortably together. The licensing base is large on paper. Whether those customers are still committed is a different question, and it decides what this announcement is worth. Broadcom is not protecting people who never considered leaving. It is trying to stop decisions already in motion.
What VMware still has not said
None of the important details came from a keynote. It came from interviews. There is no confirmed version number, ship date, price, licensing structure, or route to market, and VMware says more will come when Explore reaches Germany in mid-October.
Route to market matters nearly as much as the product here. Broadcom reshaped the old VMware channel, and the part it cut hardest served smaller customers. A revived SKU that exists on paper but is hard to quote, buy, or renew does not solve the original problem.
The memory tiering remark, and the bill attached to it
Turner mentioned memory tiering as a possible feature of the new vSphere Standard. That one word carries more than it looks.
Memory tiering appeared as a technology preview in vSphere 8.0 Update 3 and became production-ready with VCF 9.0, built on vSphere 9.0, with further work in 9.1. vSphere Standard remains on 8.0 Update 3. If the production implementation ships in vSphere Standard, that points strongly toward vSphere Standard moving onto the 9.x code base. That is an inference rather than proof, since Broadcom could backport it to the 8.x branch, though that would be the less obvious engineering path. If vSphere Standard does move to 9.x, this is more than a refreshed commercial bundle. It means the edition is back on its own code line, with its own roadmap.
The feature also has conditions. Local PCIe NVMe is architectural rather than advisory, with no SAN, NFS, SATA, or SAS substitute, and tiered capacity currently capped at 4 TB per participating host. Broadcom’s guidance points to Class D or better endurance, Class F or G performance, or enterprise mixed-use NVMe rated at 3 DWPD or above, validated through the vSAN section of the Broadcom Compatibility Guide. In practice, that means spare local PCIe capacity and drives from the expensive end of the catalog.
2026 is a bad year to go shopping for either memory or enterprise flash. TrendForce reported conventional DRAM contract prices rising 90 to 95 percent quarter-over-quarter in Q1 2026 and another 58 to 63 percent in Q2. NAND rose 55 to 60 percent in Q1 and 70 to 75 percent in Q2, driven heavily by enterprise SSD demand from large-scale AI deployments, with enterprise SSD contract prices specifically up around 80 percent in Q1. Q3 increases have moderated but still climb from a far higher base. Gartner has described memory increases of 50 to 200 percent across the first half of 2026, with some server costs rising more than 125 percent, and vendor quotes sometimes valid for only a week.
Memory tiering is sold as an economic trade: buy cheaper NVMe capacity instead of more expensive DRAM. That trade still exists because NVMe remains far cheaper per gigabyte, and plenty of designs still make sense. But the gap has narrowed, and the drive-class tiering that wants is the one absorbing AI demand. Anyone sizing it today needs current quotes rather than a 2024 assumption. For the customer consolidating to three or four hosts specifically to control VMware cost, more memory without more RAM can arrive with a large flash bill attached. A feature that reads as an operating saving on a slide can still land as capital expenditure in the quote.
The clock is already running
vSphere 8 end of general support is widely cited as 11 October 2027, with technical guidance extending into October 2029. Treat that as a working assumption, not a fixed date. It follows the usual five-year-from-GA pattern; Broadcom has not presented that pattern as an unchangeable policy, and it has already extended other VMware support dates. Verify the lifecycle that applies to your own contract.
If meaningful Standard detail arrives in mid-October 2026, customers have roughly twelve months to decide. For a three- or four-host environment, twelve months is also enough time to evaluate another platform, test it, validate backup and DR, train the administrators, and finish the migration. Every month without pricing and licensing details is a month spent proving the alternative works.
What Proxmox actually announced
The Proxmox announcement was a different kind of thing entirely. Nothing in it changes the hypervisor, introduces a version, or alters the technical roadmap. Engineering and the roadmap stay in Vienna.
The announcement was 24/7 enterprise support from 19 October 2026 covering Proxmox VE, Proxmox Backup Server, and Proxmox Datacenter Manager, including offline updates and key activation for air-gapped environments. Alongside it, Proxmox North America Inc. opens in Kingston, Ontario, under CEO Bill Hughes, handling sales, contracts, and procurement in US and Canadian dollars, and business-hours technical support across North American time zones. Existing North American contracts move to the new entity at renewal during 2027.
Most of the coverage has flattened the tiers. Premium subscribers get 24/7 from launch, with unlimited tickets, two-hour prioritized response, and SLA-backed escalation. Standard-tier subscribers get 24/7 access during an onboarding period in Q4 2026. Basic keeps its existing SLA, with Proxmox saying only that round-trip times should improve as global teams pick up cases outside Austrian working hours. So “Proxmox now has 24/7 support” is accurate for Premium on 19 October and conditional for everyone else.
Proxmox states 2.3 million active Proxmox VE servers, more than 60,000 enterprises with active support plans, and more than 3,000 integration partners. Those are company figures, and I have not seen them independently verified.
The product moved too
The support news arrives after a year of real product work. Proxmox VE 9.2 shipped on 21 May 2026 with a Dynamic Load Balancer, so the cluster resource scheduler can now run dynamically, using node and guest utilization to migrate HA-managed workloads and reduce imbalance while staying inside the existing HA rules. The same release added WireGuard and BGP as SDN fabric protocols, route maps and prefix lists for BGP and EVPN filtering, GUI management for custom CPU models, and HA arm and disarm controls, on Debian 13.5 with kernel 7.0, QEMU 11.0, LXC 7.0 and ZFS 2.4.
Proxmox Datacenter Manager reached 1.0 on 4 December 2025 and 1.1 on 28 May 2026, giving centralized management across multiple PVE clusters and Backup Server instances. Backup Server itself reached 4.2 in April 2026. The ecosystem moved as well, with StorPool integration in April, NAKIVO integration in March, and a Kasm partnership around VDI. Twelve months ago, there was no stable multi-cluster management product and no released dynamic workload balancer. Both are now shipping products.
Proxmox is still not vSphere
I have not run Proxmox in production, so what follows is a design-level assessment from published behavior.
Proxmox is not vSphere, and anyone claiming the two are interchangeable is flattening real architectural differences. The gaps are narrower and more specific than they were, which is a more useful conversation than saying Proxmox lacks enterprise features. The Dynamic Load Balancer works on HA-managed guests using CPU and memory utilization. vSphere DRS goes deeper, and Storage DRS is a separate capability the load balancer does not replace. Datacenter Manager is a young 1.x product with no native HA for the management service, though the architecture is loosely coupled: if it fails, the underlying PVE clusters and their guests keep running, and what disappears is the centralized management plane. Proxmox lists an active-standby design as future work. There is no native equivalent to the vSphere Metro Storage Cluster architecture and its failure semantics; fleet-scale lifecycle and firmware management is far behind vLCM, and the hardware compatibility and certification ecosystem is much thinner. None of that is fatal for two or three clusters and a few hundred virtual machines. Above that line, it starts to matter a great deal.
That is why support news matters more than feature news. Direct vendor support around the clock, backed by an organization that can escalate to the engineers who build the platform, plus a local legal entity for contracts and procurement, removes a category of objection that has nothing to do with the hypervisor. Enterprise decisions rarely fail because the software cannot run the workload. They fail because procurement asks who signs the contract, finance asks what currency the invoice arrives in, legal asks which entity carries liability, and operations asks who answers at three in the morning. Proxmox spent years proving people want to run the platform and is now building the commercial organization around that demand. That is the right order. Plenty of vendors do the opposite and wrap an enterprise support structure around software nobody particularly wants to operate.
The argument Turner made, and why I do not buy it
Turner said the new vSphere Standard will be better and more resilient than Proxmox. That is a fair claim for VMware to make about a product he has seen and I have not. He also pointed to Proxmox only now introducing 24/7 support as evidence that its maturity trails VMware and its partners, and that argument is much weaker.
Support hours measure commercial and organizational maturity. They do not measure engineering maturity. Proxmox VE had its first public release on 15 April 2008 and its first stable version that October. Eighteen years of production use says more about the platform than the year the company decided to staff the night shift. The arrival of 24/7 support reflects a changing customer base and a change in ambition. The code did not become mature in October 2026. Using support hours as the maturity marker reverses the causality, and it invites an uncomfortable reply from the exact customers VMware is now trying to win back.
Conclusion
Two things are true at once. Broadcom got this segment wrong for nearly three years, and it is now correcting it. I have been critical of the licensing changes, the channel cuts, and the VCF-or-nothing quoting, so now I think they have made the right decision. A supported, current, reasonably priced edition for the three or four host customers should exist. It should never have stopped moving forward.
Proxmox is not vSphere; anyone who works with both knows this. The feature distance is real, and it will not close this year. But 24/7 direct vendor support and a local legal entity remove two objections that kept Proxmox out of enterprise evaluations regardless of what the platform could actually do. Proxmox built the platform first and the commercial structure second, which is the harder order and the right one.
The question now is timing. Details on the new vSphere Standard arrive in Germany in mid-October. vSphere 8 general support is expected to end in October 2027. That leaves roughly twelve months between learning what the product is and needing a supported path forward, and for a small environment, twelve months is also enough to evaluate, test, and migrate to something else. An upgrade only beats a migration for people who haven’t finished one, and every month without a price, a licensing model, and a ship date is another month the alternative gets tested in a real environment.
Both companies moved on the same weakness from opposite directions. Proxmox had the product and is closing the commercial gap. VMware had the technology and has now acknowledged, on the record, that it deliberately chose not to refresh vSphere Standard while its commercial model pushed customers toward VCF.
The question I cannot answer yet is whether the new vSphere Standard sits on the 9.x code base. Turner’s memory-tiering remark points that way, because the feature does not exist as a production capability below VCF 9.0, while vSphere Standard remains capped at 8.0 Update 3. Nobody has confirmed it. That single answer separates a genuinely current product with a roadmap from a longer life for the existing branch, and it is the first thing I will be looking for when VMware gets to Germany.
Final note
In the same week Broadcom said it had corrected its approach to smaller customers, it terminated the person who held that community together. Corey Romero posted on LinkedIn that after 19 years working with the VMware community, most of it as vExpert Global Evangelist Program Manager, “I was terminated by Broadcom yesterday.” He is looking for a new role in tech and asking for recommendations while he does.
I have been a vExpert for 11 consecutive years, and for all of them Corey was the person on the other end. He ran the applications, the directory, the Slack, the briefings, the Explore meetups, and the partner perk negotiations. He built the first VMware WordPress instance and kept maintaining the platform long after his title stopped saying engineer. Eric Nielsen built the program with him and still runs community work at Broadcom, and that credit is real. For the members, Corey was the face of it.
What made it work was a view of what the program was for. He ran vExpert as something built for its participants rather than a marketing channel aimed at them, which is why the community held together through an acquisition that cost it a great deal.
Thank you, Corey!
