The artificial-intelligence renaissance of the last few years has transformed data center design at both cloud providers and inside self-managed facilities, putting more and more of a premium on specialized chips that can handle that research. Megh Computing, founded by ex-Intel chip engineers, hopes to capture some of that opportunity.
Portland Business Journal reported Wednesday that Megh has raised a modest $1.5 million seed fund from lead investor Portland Seed Fund and a few other participants. Based in Portland, Megh has developed software tools designed to be used with FPGAs (field programmable gate arrays), chips that can be reconfigured to focus on specific workloads as needed by customers.
General-purpose server processors from companies like Intel are good enough for the vast majority of cloud workloads, but newer areas like artificial intelligence and autonomous vehicles require the heavy application of processing power designed for those tasks. FPGAs are attractive because they can be reconfigured to support different tasks as needed, but the programming required to do that isn’t easy.
That appears to be what Megh has in mind, providing software libraries — collections of code that developers can use in their own applications — to users deploying FPGAs in their data centers or renting them on the cloud. Megh’s software will also improve the performance of FPGAs compared to the approaches used by developers who might lack expertise in this area, Megh Computing CEO Prabhat Gupta told Portland Business Journal.
Gupta was general manager of Intel’s project to meld its Xeon general-purpose processors with FPGAs, which eventually led to its acquisition of Altera in 2016. Cloud providers are also working on their own custom chips for these types of applications, as Amazon Web Services vice president of infrastructure and commercial support Peter DeSantis outlined at our GeekWire Cloud Tech Summit last month.