Google and Amazon Eye AI Chip Supremacy: A New Era or Empty Ambition?

Google and Amazon are on the brink of shaking up the AI chip market, and Nvidia might want to brace itself. Both companies recently signaled plans to sell their own custom AI chips directly to customers. It sounds ambitious, but is it realistic? Let's dig into the details.

Setting the Stage

Nvidia has long been the undisputed king of AI chips. Their stock has soared on the back of this dominance, with the AI industry seemingly tethered to Nvidia's offerings. Yet, as of their Q1 earnings reports, Google and Amazon have planted a flag in the sand. They're ready to sell their AI chips, Google's TPUs and Amazon's Trainium chips, to a broader audience.

Andy Jassy, Amazon's CEO, made it clear in his recent shareholder letter. He's eyeing a future where Amazon sells full racks of Trainium chips beyond its own cloud services. In fact, Jassy hinted that this could happen in the next couple of years. Meanwhile, Google intends to start delivering TPU chips to select customers' data centers within the year, with bigger financial impacts expected by 2027.

The Nvidia Challenge

But here's the catch. Google and Amazon are also among Nvidia's biggest customers. They've been buying Nvidia chips to power their own data centers. So, how awkward will it get when these tech giants become competitors? Both companies insist they'll continue working with Nvidia. However, Nvidia's shares took a 4% hit following these announcements, a hint of market jitters, perhaps?

Alvin Nguyen, a senior analyst at Forrester, says Nvidia should be "concerned but not worried." The company has a strong ecosystem of hardware and software. Simply put, selling chips isn't just about the tech. Google and Amazon will need to offer extensive support and education to make their chips viable in the market.

A New Era or Just Talk?

Custom silicon is indeed becoming a significant part of the AI narrative for both Google and Amazon. They're pitching their chips as cost-effective alternatives to Nvidia. Yet, the path to widespread adoption might be tricky. Google's and Amazon's chips are closely tied to their own data center environments, which could limit appeal.

Patrick Moorhead of Moor Insights &. Strategy points out that this customization might be a hurdle. The broader AI industry is diversifying, using multiple suppliers. OpenAI, for instance, has deals with Nvidia, AMD, and Broadcom. So, will Google and Amazon's chips add value beyond their ecosystems, or are they just diversifying for the sake of it?

Beatriz Valle from GlobalData described this move as "extraordinary." She believes it's irreversible. Is it a strategic masterstroke or just another chapter in the complex saga of tech giants jostling for position? The gap between ambition and execution is enormous. Only time will reveal if Google's and Amazon's chip dreams will come true or crumble under the weight of Nvidia's legacy.