Some briefings are a polite walk through a slide deck. Others leave you with the feeling that a company just moved from “interesting” to “worth a serious look.” VergeIO’s session at the 70th IT Press Tour was the second kind.
When a company hires a new CEO, the press release usually says all the right things. When that CEO spent much of his career at VMware, and the company’s board now includes a former VMware CTO, well, you sit up straight and you pay closer attention.
That was the backdrop in Palo Alto last week, when VergeIO put its new chief executive, Parag Patel, in front of the tour. Chief Marketing Officer George Crump and technical solutions strategist Dave Vincent joined him. Between them, they covered a new dawn of leadership, a clear-eyed view of the VMware exodus, a handful of live demos, and a pretty compelling look at private AI running on infrastructure you already own.
Let me walk you through what we learned last week during the IT Press Tour event.
A VMware Veteran Takes the Wheel
Patel joined VergeIO about a month ago, and he made a point of saying it’s still early days. His résumé, though, reads like a tour of the modern data center. He started in product marketing at National Semiconductor in Santa Clara, went on to business school and enterprise sales, and landed at VMware in 2004, when the company had roughly 500 employees and had just introduced its virtualization hypervisor technology, ESX Server.
Over the years he built out VMware’s partner ecosystem, helped build the channel, ran the embedded OEM business, and started the telecom business through partnerships with Ericsson, Alcatel-Lucent, and Nokia. His last five years there were spent running worldwide sales for vSAN, taking it from zero to 25,000 customers. He told us the very first vSAN contract came from the U.S. Army, for a small edge deployment, and that the product overtook Nutanix for the top market share spot within about three years.
And with that illustrious look back at Patel’s background, the question in the room became — why join a competing company to VMware that many people haven’t heard of yet?
His answer was refreshingly engineer-ish. “It’s all about the product,” he said, and then he got specific. He described the company’s architecture as doing more than VMware ever did, with networking, storage, virtualization, backup and recovery, and deduplication all living in the kernel. Patel’s words: “You can collapse more and more data center functions than even VMware did.”
That’s a big statement coming from someone who helped build the original. It also lands differently when you know that VergeIO’s board also includes Kit Colbert, the former VMware CTO. Two people who spent years inside the company that defined server virtualization looked at VergeIO’s design and decided to put their names on it. For a company that has spent years building quietly, that kind of validation matters.
What’s the mandate? Patel said the board wants him to take the company from eight-figure revenue to nine and ten, and a fresh injection of capital means there’s money to try. “I want to take my old customers and bring them back to a better place than where they wound up right now,” he said.
Why the Architecture Is the Story
Here’s the thing about VergeOS. Most infrastructure stacks are assembled from parts: a hypervisor from one vendor, storage software or an array from another, networking from a third, then a backup product, then a DR tool. Five products, five upgrade calendars, five support contracts.
VergeOS takes a different road. It’s one infrastructure operating system, built as a single code base, that turns standard x86 servers into a complete environment, including the DR site. Crump pointed out that the whole thing runs in under 400,000 lines of code. Let that sink in for a minute, much like we did in the briefing room. That’s a small number for an entire infrastructure stack, and he was quick to explain why anyone should care, since users don’t buy line counts.
The payoff is efficiency. Customers see a 25 to 30 percent increase in VM density per physical server, according to Crump. Topgolf is the example he kept coming back to. The company runs 110 venues and shrank each one from six VxRail servers to three standard Dell servers. It also runs those 110 venues with a team of three people. Saratoga Casino has a similar story: three people managing four casinos that can never go down.
Licensing follows the same philosophy. It’s one license per physical server. No counting cores, no counting sockets, no spreadsheet gymnastics. VergeIO’s pitch is built on what Patel calls “simplicity, savings, and service,” and the company’s NPS scores back up the service part. Patel said he called a bunch of customers during due diligence, and they were raving. Crump had the best anecdote: one customer posted on LinkedIn that he had a new girlfriend, and her name was VergeIO. (Crump’s reaction was, roughly, “Okay, maybe not that.”)
Patel also offered a bit of history that explains the whole pitch. Mainframes got complicated. Then x86 servers gave everybody mainframe power for cheap, and by 2000 data centers were full of idle, overprovisioned boxes and admins sprinting around in sneakers. VMware cleaned that up, and then, in his words, “we did it to ourselves.” vCenter turned into a pile of add-ons, network virtualization became another silo, and backup appliances multiplied. Now, he argued, it’s time to collapse a few more layers.
300,000 Customers and an Open Door
Timing matters here, and Patel was candid about it. He estimates around 300,000 customers worldwide are looking to move away from the incumbent. Many aren’t thrilled with Nutanix either, for cost and complexity reasons, according to Patel. Broadcom, he said, wants to hold onto its top 10,000 customers, and Nutanix has to chase the enterprise too. That leaves a very large middle for everybody else.
VergeIO is going after that middle first, in small and mid-sized organizations. Patel said VMware itself began in the mid-market, and the enterprise didn’t take it seriously until version three or four. He expects the same pattern here. He also sees a renewal wave coming, since many companies signed two- and three-year deals in 2023 and 2024.
As we discussed migrations, something top of mind for me was the question, “What about Broadcom closing off the VDDK, the API many migration tools use to pull VMs off vSphere?” The VergeIO team said they aren’t affected today, calling the change a warning shot they’re watching. Patel’s view: for every speed bump, there’s a detour, and plenty of customers can live with a bit of downtime during a move anyway.
Crump framed the way customers show up as four doors:
- Exit VMware. Import running VMs with their storage and networking intact, and keep applications as they are.
- Fight data center inflation. Stretch the hardware you own by mixing server generations, running refurbished nodes, and using less RAM and storage per workload.
- Run every site the same way. Every location runs the same VergeOS and takes the same upgrade, automated with Terraform, Ansible, and Packer.
- Maintain AI sovereignty. Run models on GPUs inside your own tenant, so prompts and data never leave your network.
Most customers walk through one door and then discover the others. Crump’s advice on the first one was blunt: get VMware out of the picture first, and worry about the cool features later.
Leaving VMware One Server at a Time
The migration story deserves its own moment, because it’s where many readers will feel the pain. VergeIO’s in-place approach works like a careful game of musical chairs. Free up one host, move its VMs to the others, reimage the empty host with VergeOS, then move the next batch across, and keep going until the whole cluster has changed over. Crump said more than half of customers do this without buying any new hardware, though many add a server or two to make the shuffle easier.
Vincent showed how it looks in practice. A VMware connector talks to vCenter and refreshes the VM inventory. He picked a small Alpine VM, ran a differential backup into VergeOS, then imported it while preserving the MAC address and choosing a storage tier. The one tweak was switching the boot disk controller to SATA. He took a snapshot as a safety net, powered the VM on, and watched it boot. Because backups keep running while the source VM stays live, the only real downtime is the cutover.
How fast can it go when the pressure is on? Crump mentioned customers whose VMware licenses were set to expire days after they signed, and VergeIO moved thousands of VMs over a weekend. Not the way anyone wants to do it, but good to know it works.
We Specialize in Ugly Clusters
Hardware is where the VergeIO story gets practical, and a little funny. “We specialize in ugly clusters,” Crump said, meaning clusters where the bezels don’t match and the servers come from different vendors and different years. VergeOS happily runs Dell, HPE, Cisco UCS, and Supermicro side by side, older generations and refurbished nodes included.
Why does that matter now? Because buying new hardware has become painful. Patel noted that waiting 12 months for a server, or paying steep memory prices for one, is a real problem today. He described customers facing VxRail end of life who wipe the nodes, install VergeOS, pick up a third-party maintenance contract, and extend the hardware’s life by five years. Formica in the UK re-amortized its servers out to 2034 after moving over. Crump’s own marketing lab runs 13 servers, all 12 years old or older, and he said he refuses to let anyone upgrade them because it would ruin the point.
There’s a sustainability angle in there too, which a fellow journalist in the briefing room pointed out. With AI gobbling up power and water, a platform that keeps old servers useful for years is quietly a green story. Patel agreed, with the caveat that the pitch lands harder in Europe than in the U.S. right now.
Spinning Rust Gets a Second Act
This was my favorite part of the session, because it’s so unglamorous. Crump explained that a 15.36 TB NVMe SSD that cost around $4,000 to $5,000 last year jumped to something like $12,000 to $14,000 when his team went shopping for lab capacity. A decent 24 TB HDD? About $1,200. For the price of one NVMe disk, he could buy nearly 160 TB of hard disk capacity.
So, hard disks, in 2026? Ok, hear me out. VergeOS lets admins set a preferred storage tier per virtual disk. Keep the OS, logs, and tempdb on NVMe, and slide the bulk data down to hard disks. The slide showed a SQL Server VM’s media cost falling from $263 to $133, a 49 percent drop (using street prices from late September).
Vincent proved it live. A custom app hammered a SQL Server VM at around 80,000 rows per second while he changed the tier on the data disk. The change took roughly two seconds, and the transaction rate didn’t flinch. Crump was honest about the catch: much of that smoothness comes from VergeOS’s RAM cache, and the real data migration continues lazily in the background. “We’re not magicians,” he said. “We can’t make it as fast as flash.” He also believes auto-tiering is part of the problem, since nobody can easily undo what an algorithm decides. VergeIO would rather give the admin the lever.
Isolation, DR, and a Ransomware Tripwire
The virtual data center concept is worth a short mention. Each department or customer gets an isolated tenant with its own compute, storage, networking, and protection. If the user-data environment gets hit by ransomware, it doesn’t spread. Replicating a VDC also copies the network and storage settings along with the data, which is why Crump claims “incredibly high first-shot DR.”
On ransomware specifically, VergeOS has a feature called IO Fortify that watches net-new data growth. Because deduplication sits at the core of the OS, growth is normally slow, so a sudden jump of three or four percent is a red flag worth an alert. Snapshots are independent full copies rather than a chain of dependent metadata, so you can delete thousands of them without trouble, or even delete a damaged volume and slot the snapshot back in. VergeIO is fully supported by Veeam, and Crump was quick to say that a copy on something without the same logo as production is still best practice.
Chat With Your Infrastructure, Then Keep Your Data
Now to the part that got the most attention in the room. VergeIO has built a strong AI story into a platform that, a few months ago, was mostly about supporting GPUs. Since January, the company has worked closely with NVIDIA, and Crump said VergeIO has the company’s blessing for its approach. VergeOS can pool GPUs, do vGPU, and support NVIDIA’s MIG, which hard-allocates slices of a physical GPU to specific virtual data centers. Whatever’s left stays available to everyone else.
Crump described two paths. The first is “chatting with your infrastructure,” aimed at IT professionals. The second is offering a private AI service to the whole organization that replaces public tools like ChatGPT or Claude for sensitive work.
The first demo used Pi, an agent framework, running a local model on the cluster’s own GPUs. Vincent asked it to create 20 Ubuntu VMs, each with one CPU and 512 MB of RAM. About six minutes later, all 20 were running in VergeOS, built through the Verge CLI with nothing leaving the environment. Then he asked it to shut them down and delete them, and it did.
Crump’s idea for how this lands with admins was clever. Imagine a VMware administrator with 20 years of experience who can simply talk to VergeOS in VMware’s vocabulary, and the AI acts as a Rosetta Stone. He even joked that his team spent all this time on a new GUI and in a year nobody may use it.
The second demo covered private data analysis. Vincent ran an open-source model of roughly 35 billion parameters, served locally, and uploaded six webinar attendance exports (simulated data) for analysis, trends, and a suggestion for who sales should call first. The model reasoned through it and returned charts. “None of that data left your environment,” Vincent said. “It’s all local.”
The hardware may surprise you. His lab server has two consumer GPUs at roughly $600 each, and the model splits across them for about 110 tokens per second. That’s not a hyperscaler’s data hall, and it didn’t need to be.
Crump also had a neat way to describe the regional split. In Europe, he said, wait to hear the words “data sovereignty” and raise your hand. In the U.S., wait until somebody complains about token costs, then raise your hand. Patel thinks sovereignty could end up a bigger driver in Europe than at home, and he pointed to NVIDIA’s earnings commentary that a growing share of its sales now goes to enterprises, not only hyperscalers. VergeOS also controls what can leave the building through its networking layer, and for dark sites, Crump imagined a level-two support engineer available around the clock.
Their own use of AI is worth a mention too. The company runs models in an air-gapped private cloud to protect its IP. Patel said the engineering team now writes test harnesses and regression tests in two days that used to take a month, and that version 26.2 shipped last week. Oddly, nobody has asked him for more headcount. What the team wants is budget for beefy servers. Patel believes the company is doing the work of one, two, or three times its size.
Europe, the Roadmap, and the Honest Hurdle
VergeIO has set up a dedicated EMEA sales team led by Rémi Bargoing. Patel said the model will be channel-heavy, with local hires, local languages, and local support. He expects early adoption in the Nordics, the Netherlands, and the UK, and Germany to become the biggest market a bit later. Then he added that this might not play out the usual way, because VMware-style virtualization is a well-understood product category and adoption could arrive in many countries at once. Asia and Latin America will get the same local treatment.
As for what’s next, Crump said the themes will stay consistent: easier and more complete exits from VMware, better ways to extend the life of old hardware (including monitoring and life calculations, since ten-year-old servers do eventually break), stronger multi-site capabilities, and more preparation for AI. Patel said an ambitious 12-month roadmap is “still in the head of our founder,” with answers to be shared a year from now.
We asked the VergeIO team what would stop them from winning more deals. The answers from Crump and Patel were disarmingly honest. It isn’t price, and it isn’t features, since he says VergeIO is right there with everybody else. It’s getting into the conversation. “People don’t know we’re here,” Patel said.
After this session, that’s a little harder to say about VergeIO. A company with a VMware veteran as CEO, a former VMware CTO on the board, a unified architecture that makes old hardware useful again, and a private AI story that runs on a pair of $600 GPUs is a fair bet to show up on more shortlists. If your VMware renewal is coming up, put VergeIO on the bake-off list and see for yourself.
As Patel put it: “We thought the market was finished, and now it’s opened up all over again.”
##
