VMware went public in August and the shares nearly doubled on the first day. The market has decided that running several operating systems on one machine is a permanent and defensible business, and priced it accordingly.
The technology deserves it. Server consolidation is the rare enterprise purchase that pays back inside a year and that everyone can see working. A rack of twenty machines each running at eight percent utilisation becomes three machines running properly, and the savings on hardware, on power, on cooling, and on the floor space itself are not marginal. Anyone who has walked into a server room and felt the heat coming off equipment doing almost nothing understands the pitch immediately.
My doubt is not about the technology. It is about which part of it stays valuable.
The hard problem VMware originally solved was making virtualisation work on processors that were never designed for it. The x86 instruction set has corners that resist being run under supervision, and getting around them required genuinely clever engineering, and that cleverness was the moat.
Intel and AMD have now put the necessary support into the silicon. The corners have been smoothed out in hardware. That does not make writing a hypervisor easy, but it moves it from remarkable to merely difficult, and difficult things attract competitors.
You can see them arriving. Citrix bought the company behind the open-source hypervisor over the summer. Microsoft has said its own version is coming with the next server release, and Microsoft’s version will be included rather than sold, which is the manoeuvre that ended a great many independent software businesses in the nineties.
There is a pattern here worth naming. A layer of software becomes essential, somebody builds an excellent independent version, and then it is absorbed into either the operating system or the chip, because whoever owns those has both the motive and the distribution. It happened to disk compression, to memory management, to the TCP stack, to the firewall. It is happening to the browser slowly and to the media player less slowly.
The hypervisor looks exactly like the next one, and I would expect it to be free within about three years.
What is not obviously commoditisable is everything above it. Moving a running machine between hosts without dropping connections. Deciding where workloads should sit. Knowing what is running where across a few thousand instances, and restarting the right thing when a host dies at four in the morning. That is unglamorous, it is where the operational value actually lives, and it does not come free with an operating system.
If VMware is still valuable in five years it will be as a management company that happens to ship a hypervisor, rather than a hypervisor company. The share price currently seems to be paying for the second thing.