Why Amazon may be Big Tech’s AI dark horse
Few pundits place Amazon atop their list of companies that stand to benefit most from AI. That may soon change.
Even for Amazon, USD4 billion ($6.2 billion) is a big chunk of change. Yet when the company splashed that amount for a minor stake in Anthropic, probably the biggest challenger to OpenAI, it got a two-for-one deal.
Anthropic is galloping alongside Microsoft-backed OpenAI and Google’s Deepmind in a sprint to create artificial general intelligence (AGI). Amazon now has a horse in that race. A crucial part of the deal has Anthropic using Amazon’s Trainium GPUs to train its new models. That will prove helpful for Amazon, which is working on its next generation chips to whittle into at Nvidia's dominance.
“We get a very special type of feedback from them [Anthropic] that is very beneficial to the current generation of chip software, but also to the next generation’s hardware,” said Amazon Web Services (AWS) VP of compute David Brown. He was in Australia earlier this month to announce the full local launch of Bedrock, AWS's platform for AI app development which is already used by the likes of ANZ, Westpac and Telstra.
Nvidia’s near monopoly on AI chips comes not from hardware alone, but the ecosystem that's built up around it. Software development has a network effect to it: The more developers use a certain architecture, the harder it is to diverge from it. For one of the three top AI labs in the world to be using Amazon’s ecosystem is no small matter.
