Broadcom’s VMware strategy looks ever more shaky - and less relevant

Yes, it’s that darned AI again

Opinion For many, VMware by Broadcom has meant misery by the boatload. The virtualization platform's new owners have embarked on price hikes for the big and forcible eviction for the little. The dividing line isn't clear. A 24,000 VM migration by share repository Computershare seemingly triggered by the gouge suggests things might not go to plan.

The Broadcom BCM2836 chip in the new Pi 2

It looks a lot like VMware just lost a 24,000-VM customer


Broadcom's equation is that the bigger its customers are, the more reluctant they'll be to eat the huge cost of migration. When your customers have to pay to leave you, you can start thinking in terms of how much pain you can inflict on them and still keep them around. That seems less like a functioning market, and perhaps more an abusive relationship. An equation of value replaced by an equation of pain.

While Broadcom's model of corporate growth may not be explicitly built that way, it is very susceptible to it - and, conversely, susceptible to blowing up in its face if it gets those equations of pain wrong.

That model goes back to Avago, the company that bought the original Broadcom in 2016 and took its name. Avago traced its origins to Hewlett Packard's semiconductor division in the 1960s. After much corporate shape shifting, it found itself in the 2000s with key mobile phone technologies just as that sector took off. Judiciously buying out its competition in that niche but lucrative market gave it an effective micro-monopoly.

Which is great, but once you own a niche market, how do you grow? As lucidly detailed by the Asianometry YouTube channel, maximizing profits, minimizing costs and divesting non-core interests gave the company the leverage to borrow lots of money and acquire others with similar overwhelming presence in other niches. Repeat as fast as you can to pay off the debts incurred in each acquisition.

This behavior has resulted in Broadcom becoming what's best described as a "publically traded private equity fund masquerading as a semiconductor company." This is where the equation of pain kicks in: you'll need cash fast, so hiking your product price to breaking point is a must. You have a near monopoly, so the cost of escape for your customer base is so high they'll have to pay you instead. It worked by buying CA, it worked with Symantec. It may not be working with VMware.

As noted earlier by The Reg, a hypervisor virtualization stack is inherently resistant to the sort of lock-in to which other classes of software and specialist hardware are prone. A virtualized environment's job is to be a proxy for standard hardware. From the management and deployment perspective, there are differences that do indeed create pain points on migration, but these do not impinge elsewhere on an organization's operations or the experience of its customers. In fact, virtualization combines the inherent anti-lock-in of a standardized hardware market with the ease of deployment and scaling that comes with software. The best of both worlds - or the worst, depending on who's feeling the pain.

With Computershare, it seems that it won't be feeling the pain for long. A 24,000 VM migration that pays for itself in under a year is remarkable in any context, but one where the vendor felt able to multiply prices by a factor of ten or more is one where that vendor either got the equation of pain radically wrong, or one where the vendor didn't want the business. Which means the vendor is supremely confident of milking its biggest customers.

They may not agree. In the right light, a 24,000 VM migration looks awfully like proof of concept for a 240,000 VM migration.

Even given that Computershare was already running an environment that supported multiple technologies, the claims are that even on the old pre-hiked prices the migration would have been worthwhile. No matter what market share says, this does not feel like a monopoly too painful for clients to escape.

In the end, none of this may matter. While all the VMware shenanigans were playing out, ChatGPT happened, making Broadcom's position as co-developer of Google's TPU AI accelerator chip beyond fortuitous, and giving it a current market cap of $654 billion or more than nine times what it paid for VMware. From being a venture fund masquerading as a chip company, Broadcom has become a chip company with a vestigial venture fund.

A company this big cannot grow by acquisition. There are no targets left, and everything looks very different. The strategic choices facing the board are not so much how to manage VMware most efficiently for the next few years, but whether it's worth spending the time on. In terms of its potential to significantly grow Broadcom, it may no longer merit attention. Worse, if the equation of pain really is flawed, sticking to plan A makes no sense at all.

The odds are shortening on VMware being divested so Broadcom can concentrate on fighting Nvidia and developing its AI silicon share. For a user base weary and wary of the changes so far, this will be further unwelcome uncertainty with faint rays of hope attached. The equation of pain has changed, but the pain itself has not. ®

More about


Send us news

Other stories you might like