Why Amazon Paid $350 Million for Annapurna Labs

The cloud giant expects to make that back and then some

When Avigdor Willenz named his startup after the Himalayan range he had scaled on his trip to Nepal in the 1980s, he didn’t picture the mountain of cash that it would fetch from Amazon. In fact, he went through every conceivable effort to avoid an acquisition.

Instead of turning to VCs for the initial financing, Willenz bootstrapped Annapurna Labs with $10 million of his personal wealth and convinced a group of close acquaintances to match the sum. The move was openly aimed at removing the need to clash with investors when an outside bid would eventually arrive, since the founder planned to turn any such offer down.

Willenz successfully predicted that his startup would attract attention from potential buyers, but he didn't anticipate that the interest would materialize into an exit. Something had clearly changed since he declared to have “no intentions” of selling in a 2012 profile for the Hebrew-language Globes.

That change holds the key to why Amazon swooped in, and why now.

When Annapurna hit the scene, its offering — which never had the time to launch but has been confirmed as ARM-based silicon for midrange network servers — stood out. It posed a unique low-power threat to Intel’s dominant controllers, which have until then only been contested in the very bottom end of the market by technology from Willenz’s first semiconductor company, Galileo, now part of Marvell.

But in 2015, an ARM-based network chip that is actually useful outside small business environments is not nearly as impressive as it might have been been a couple years ago. Just in October, Applied Micro unveiled a series of 64-bit embedded processors fabricated from the ARMv8 design that is already shipping to customers.

Which begs the question why Amazon spent $350 million on Annapurna instead of putting the same money towards a supply contract with another chip maker, if it’s indeed so interested in moving off Intel processors. The answer is the usual one: cost.

Switching to ARM has the obvious advantage of lowering power consumption, as pundits were quick to point out following the deal, which can quickly add up across the cloud giant’s infrastructure empire. But that’s only the beginning of the potential savings. Moving away from Intel also opens the door to avoiding the steep profit margins it charges on each processor, yet merely signing up with another big name chip maker that also has stakeholders to please doesn’t come close to maximize the full potential for cost savings.

Bringing the design process in-house, however, does. With the talent and intellectual property that it’s gaining through Annapurna, Amazon can develop its own chips and outsource the fabrication to a manufacturer at a fraction of the cost of buying a ready-made processor. The company apparently sees the resulting savings reach well above $350 million as it continues expanding its infrastructure.

As for Willenz and his team, they've managed to avoid a collision course with the giants of the ARM ecosystem and make a hefty profit in the process. The venture may not have turned out as originally envisioned, but $350 million is difficult to argue with.

Banner via TechReport