As Amazon, Alphabet, and Microsoft aggressively develop their own AI processors to reduce dependency on external suppliers, questions arise about Nvidia’s long-term dominance. Despite these tech giants’ internal chip advancements and significant investments, they continue to be major purchasers of Nvidia’s GPUs. This creates a complex dynamic where custom silicon could eventually challenge Nvidia’s pricing power, yet the rapidly expanding AI market might allow Nvidia to sustain growth, even with some market share erosion.
Three of Nvidia's largest customers—Amazon, Alphabet, and Microsoft—are simultaneously engaged in an intensive effort to develop their own artificial intelligence (AI) processors, aiming to reduce their reliance on the chip giant. These tech behemoths are aggressively integrating their custom silicon into their rapidly expanding data centers. This trend might seem like an imminent threat to Nvidia, whose graphics processing units (GPUs) have long been the industry standard for AI workloads. Indeed, market sentiment on a recent Friday saw Nvidia shares drop by approximately 6% amidst a broader semiconductor downturn, suggesting investor concern about these internal chip initiatives. However, the intriguing paradox at the heart of this story is that these same companies are also purchasing record volumes of Nvidia's chips.
Image source: Getty Images.
Amazon's In-House Chip Prowess
Amazon stands out with the most mature internal chip development program. Its custom silicon portfolio, encompassing the Graviton processor, Trainium AI chip, and Nitro networking chip, achieved an impressive $20 billion annual revenue run rate in the first quarter of 2026. Amazon CEO Andy Jassy highlighted the scale of this achievement during the company's first-quarter earnings call, stating, "If our chips business was a stand-alone business and sold chips produced this year to AWS and other third parties as other leading chip companies do, our annual revenue run rate would be $50 billion." Jassy further asserted that Amazon's chip operation now ranks among the top three data center chip businesses globally. Despite this remarkable internal success, Amazon's demand for Nvidia's GPUs remains robust. The company projects capital expenditures of around $200 billion in 2026, with a significant portion allocated to infrastructure that continues to heavily depend on Nvidia GPUs to power its Amazon Web Services (AWS) offerings for clients.
Alphabet's Expanding TPU Reach
Alphabet, through Google, has been designing its tensor processing units (TPUs) for over a decade. The year 2026 appears to mark a pivotal shift, as Google begins to extend the availability of its custom chips beyond its internal operations. Google recently unveiled its eighth-generation TPU systems, signaling a strategy to offer these powerful processors as a rentable cloud service. A notable development in May saw Blackstone announce a joint venture with Google, committing an initial $5 billion to bring 500 megawatts of TPU capacity online by 2027. This follows previous agreements to grant AI lab Anthropic access to as many as 1 million TPUs and a reported leasing deal with Meta Platforms. While Google's willingness to offer its custom chips externally intensifies its direct competition with Nvidia, recent reports indicate that Google also inked a multiyear cloud deal with SpaceX, securing access to approximately 110,000 Nvidia GPUs. This demonstrates Alphabet's continued dual approach of building its own while simultaneously buying from Nvidia.
Microsoft's Strategic Maia Accelerator
Microsoft appears to be in the earlier stages of its custom silicon journey compared to its peers. Its primary initiative revolves around the Maia accelerator, with the second-generation Maia 200 having recently commenced operations in some data centers. These chips are supporting critical workloads for Microsoft 365 Copilot and models developed by its partner, OpenAI. Nonetheless, the overwhelming majority of AI processing within Microsoft's Azure cloud still relies on Nvidia GPUs. The Maia accelerator, therefore, represents a strategic endeavor to gradually claw back some of that substantial spending over time, rather than a swift, wholesale replacement of Nvidia's hardware. Microsoft's capital expenditures are staggering, with an expected investment of roughly $190 billion in calendar year 2026, even as its Azure cloud service, which saw revenue grow 40% in its fiscal third quarter (ending March 31, 2026), continues to face capacity constraints through the year-end.
Nvidia's Future Amidst the Shift
Collectively, Amazon, Alphabet, Microsoft, and Meta Platforms are projected to spend approximately $725 billion on capital expenditures in 2026, marking an astonishing 77% increase from the previous year. This substantial spending forms the core of the 'bear case' for Nvidia: a growing portion of this capital could be directed towards custom chips designed by its largest customers, driven by a strong incentive to reduce dependence on a single supplier. Conversely, the 'bull case' for Nvidia is powerfully illustrated by its own recent financial performance. In its fiscal first quarter of 2027 (ending April 26, 2026), Nvidia's revenue soared by 85% year-over-year to $81.6 billion, with data center revenue alone jumping 92%. Hyperscalers still accounted for roughly half of this booming data center business. Nvidia founder and CEO Jensen Huang aptly described the situation, stating, "Demand has gone parabolic." He also highlighted a rapidly expanding customer base beyond hyperscalers, including AI start-ups, enterprises, and governments, many of whom "do not build chips, do not design their own chips." The resolution of this dynamic is likely a complex coexistence for the foreseeable future. Custom silicon is undeniably real and gaining traction, poised to potentially erode Nvidia's pricing power over time. However, the overall AI spending pool is expanding at such an accelerated rate that Nvidia could continue its growth trajectory even if it experiences some marginal loss of market share. Given Nvidia's price-to-earnings ratio of about 32 at the time of writing, the more significant risk might not be a dramatic failure of these in-house chip efforts, but rather their gradual success, while the market maintains a valuation for Nvidia based on an assumption of perpetual dominance.
