
Amazon Web Services (AWS), the cloud computing giant, appears poised to make a monumental move that could directly challenge Nvidia’s seemingly unshakeable dominance in the artificial intelligence (AI) chip market. For years, AWS has developed its own custom AI chips, primarily for internal use, but that strategy is now set to evolve dramatically. This shift could redefine competition in the rapidly expanding AI landscape.
Recent reports indicate that AWS is in active discussions to sell its proprietary **Trainium** AI chips to other companies, specifically for integration into their data centers. This isn’t just a rumor; Peter DeSantis, Amazon’s AI chief, confirmed these early-stage talks to Bloomberg, though he refrained from naming potential buyers. Such a bold step signals a new era for Amazon’s hardware ambitions.
Amazon’s $50 Billion Vision Takes Shape
The groundwork for this strategic pivot was laid bare in Amazon CEO Andy Jassy’s annual shareholder letter earlier this year. Jassy explicitly stated the immense internal demand for Amazon’s homegrown AI chips, suggesting the company was contemplating selling them externally. He painted a compelling picture of their potential value.
Jassy elaborated, “If our chips business was a standalone business, and sold chips produced this year to AWS and other third parties (as other leading chips companies do), our annual run rate would be ~$50 billion.” He went on to say, “There’s so much demand for our chips that it’s quite possible we’ll sell racks of them to third parties in the future.” This visionary outlook underscores the significant market Amazon believes it can capture.
While a $50 billion annual run rate is substantial, it’s important to put it into perspective against Nvidia’s staggering current $326 billion revenue run rate. However, this projected figure isn’t insignificant; it’s comparable to the annual revenues of established industry players like Intel. This makes Amazon’s potential foray into direct chip sales a formidable challenge, rather than a minor disruption.
From Internal Use to External Sales: Why Now?
Historically, AWS has resisted selling its advanced AI chips directly to third parties, and for good reason. The primary benefit for AWS in developing its own silicon like Trainium and Inferentia has been the “waterfall effect” within its cloud ecosystem. By providing its custom chips on AWS, the company not only charges for the AI tokens processed but also for an array of complementary cloud services.
These essential services include storage, security, networking, and monitoring, all critical components that companies need to run their AI applications effectively. This integrated approach created a robust revenue stream beyond just the chip’s raw processing power. Selling chips directly means foregoing some of these downstream cloud service revenues, indicating a significant strategic re-evaluation.
Another crucial factor is the unprecedented demand for these chips. As Jassy highlighted in his shareholder letter, the current capacity for **Trainium** chips has been selling out almost instantly. Furthermore, the capacity for the next-generation **Trainium4**, which isn’t even slated for release for over a year, has also been completely pre-booked. This overwhelming internal demand, even before AWS formally integrated OpenAI models into its offerings, showcases the superior performance and efficiency of their designs.
Navigating the Manufacturing Landscape
The decision to sell chips externally introduces a complex set of challenges, particularly concerning manufacturing capacity. To meet both internal AWS demand and satisfy new third-party customers, Amazon would need to significantly ramp up production. This likely means securing more foundry space from manufacturing partners like **TSMC**.
However, this puts Amazon in direct competition with Nvidia for TSMC’s finite and highly sought-after fabrication resources. Nvidia has recently become TSMC’s largest customer, even surpassing Apple, making it incredibly difficult for any new player to elbow their way in. Securing sufficient manufacturing slots will be a critical hurdle for Amazon’s ambitious chip-selling strategy.
AWS spokesperson Doron Aronson reiterated the potential shift, confirming, “While we’ve historically declined requests to sell chips directly, Andy noted it’s quite possible we’ll sell racks of them to third parties in the future.” This direct confirmation from AWS leadership underscores the seriousness of their intent. As Jensen Huang, Nvidia’s CEO, ventures into the CPU market for AI, Amazon’s move is a clear signal of its own intent to carve out a significant share of the AI chip market, directly challenging Nvidia’s stronghold.
Source: TechCrunch – AI