There has been a lot of talk about ARM (ARMH +0.00%) based servers in the datacenter, and Calxeda – founded in 2008 – was one of the leading "champions" of this move to ARM-based servers. Unfortunately, despite the fanfare that this company has drawn over the last couple of years, the company couldn't secure another round of funding to continue development of its low power ARM-based chips intended for scale out servers. As a result, the company is shutting down.
Why couldn't it secure more funding?
The most pressing question that any believer in the ARM server story should attempt to answer is, "why couldn't Calxeda get more funding?"
The reason that it's such an important question is that Calxeda was funded by some pretty big names, including ARM Holdings, Advanced Technology Investment Company (the folks backing Global Foundries), Highland Capital, and many others. In particular, with ARM so gung-ho about its datacenter opportunity, one would think that it would find the cash (or the stock – ARM's stock is currently pushing through to new all-time highs and would make great "currency" for this fund) to keep the dream alive.
Calxeda was primed for a buyout, but it never came
There was never any chance that a start-up with "more than $90 million" in funding could really succeed in the server space against Intel (INTC +7.88%) to whom countless others – including DEC, IBM, Oracle, and AMD – have fallen. The goal for any small start-up like Calxeda is to develop valuable technology that can be sold for more than the initial investment.
It is not inspiring to see that Calxeda's original backers – again, one of whom is ARM itself – unwilling to sink any more money into this investment. If there was really something unique and valuable that could come out of this, then additional investment would be warranted. But that's the problem – there wasn't much here. Calxeda basically took off-the-shelf ARM cores and built a system-on-chip around them and a custom-developed scale out fabric known as "Fleet Fabric". This "fabric" was supposed to allow Calxeda to connect "10's, 100's, or even 1000's" of these SoCs (known as EnergyCore) to enable great scale-out dense server solutions.
How much was that really worth?
The big question here is just how much is a successful fabric worth? Well, look no further to AMD's (AMD +6.31%) acquisition of SeaMicro for its "Freedom Fabric." AMD paid $334 million for the company and in return got a very nice fabric (that AMD plans to integrate into its own scale-out oriented chips) as well as a business that actually has a pretty sold number of customers actually buying its home-grown servers.
It makes sense that investors didn't want to sink more than $100 million into Calxeda. If SeaMicro – with a fabric and a business that sells entire boxes – went for $334 million, then Calxeda may have been able to sell itself for $150 million – not a terrible return on $100 million, but it leaves very little room for further investment with an expectation of meaningful return on that capital.
What happens now?
With Calxeda shutting down, the interesting thing to see will be who picks up the remains. If, say, Google or Facebook were interested in picking up the remains, then this may lend some credence to the rumors that these players are looking to develop their own server SoCs. Same thing goes for Qualcomm or Samsung. At any rate, Calxeda's assets probably won't see too much bidding action since if a company really did find these assets highly valuable, they probably would have done so before this shutdown.
Foolish bottom line
Competing in servers isn't easy and it isn't cheap; anybody thinking that $90 million or so was enough to go up against Intel was perhaps a tad too optimistic. There are still other players left wanting to challenge Intel here, and many will certainly try, but as we've seen with Calxeda, there will be many more broken companies on this trail to ARM servers before this is all said and done.
