>> Calxeda, a maker of ARM-based server SoCs, said it will close its doors.
This company might have sold its business or got acquired than just closing doors. I felt they had some really great business model out there. It is always hard to disrupt hardware business as you need to have a lot of funds to do that.
It's funny, with 320bit core architectures attracting less-than-stellar demand, the pressure on all those 64-bit ARM-based SOCs scheduled to ship next year and beyond mounts. Calxeda had a lot of ambition in this market and I for one am surprised to hear of this announcement. Nonetheless, the staikes remain high in this new segment--there could be a lot to gain but its a matter of biding one's time properly which Calxeda was not able to do.
Rick, you wrote:
Calxeda could not hold on until the emerging market becomes large enough to sustain it.
It's true. And I imagine that "emerging market" also includes geographically emerging market such as China. I remember earlier this year, I bumped into a group of Calxeda's senior executives including its CEO, in a hotel in Beijing. It was clear that the startup had high hopes for a growing number of Internet-based service companies in China to embrace Calxeda's solution based on ARM 32-bit cores.
But again, that market probably didn't grow fast enough for the startup to hang its hat on.
Are you seeing any significant market for 32-bit ARM server SoCs?
Where would such a market be? The nature of what servers do will place a premium on address space. The vastly greater address space possible in 64 bit architectures will make them preferable for most server tasks I can imagine,
I'd call Calxeda's 32 bit offerings proof-of-concept, demonstrating performance in a ballpark with comparable X86 chips but greater power efficiency. It would all have been well had ARM had 64 bit CPUs available in a shorter time frame.
They didn't, and it killed Calxeda. Investors were unwilling to provide further capital to keep Calxeda going. It would lengthen the payback period and increase their risk. Calxeda would need to establish a clear leadership in the 64 bit ARM server market, grow rapidly, and ship a lot of silicon, and do so in the face of competition from bigger, better heeled vendors looking to play in the same space. The folks providing the funding would see a better bet in cutting their losses on Calxeda, and investing in someone else when ARM actually had 64 bit designs ready to turn into chips.
I might use a 32 bit ARM chip as something like a peripheral controller in a server, but I wouldn't use if for a CPU, and CPUs are the point of this exercise.