Inside Amazon.com’s chip lab in Austin, Texas, half a dozen engineers on a Friday afternoon put a closely guarded new server design through its paces.
The server was packed with Amazon’s artificial intelligence chips that compete with those from market leader Nvidia, Amazon executive Rami Sinno said, during a visit to the lab.
Amazon is developing its own processors to limit its reliance on costly Nvidia chips – the so-called Nvidia tax – that power some of the artificial intelligence cloud business at its Amazon Web Services, the main growth driver.
Through its homegrown chips, Amazon wants to help customers compute complex calculations and process enormous amounts of data more cheaply.
Its rivals Microsoft and Alphabet are doing the same.
Sinno, the director of engineering for Amazon’s Annapurna Labs that is a part of its cloud business AWS, said Amazon’s customers were increasingly demanding cheaper alternatives to Nvidia.
Amazon bought Annapurna labs in 2015.
While the company’s AI chip efforts are nascent, Amazon’s workhorse chip Graviton that performs non-AI computing has been under development for nearly a decade and is on its fourth generation. The AI chips, Trainium and Inferentia, are newer designs.
“So the offering of up to 40 percent, 50 percent in some cases of improved price (and) performance – so it should be half as expensive as running that same model with Nvidia,” David Brown, vice president, compute and networking at AWS said.
Sales at AWS, which accounts for just under a fifth of Amazon’s overall revenue, surged 17 percent to US$25 billion ($38.3 billion) in the January-March quarter, compared with a year earlier.
AWS controls roughly a third of the cloud computing market, with Microsoft’s Azure holding about 25 percent.
During its recent Prime Day, Amazon deployed a quarter million Graviton chips and 80,000 of its custom AI chips to handle the surge in activity across its platforms, the company said.
The shopping event generated a record US$14.2 billion in sales, according to Adobe Analytics.