Three of the biggest buyers Nvidia (NVDA 5.93%) All three companies are the least demanding of the chips. Amazon (AMZN 2.93%), letters (GOOG 0.80%)(GOOGL 0.82%)And Microsoft (MSFT 2.55%) Each is now designing their own AI processors, and each is building that silicon at speeds deep into data centers.
It’s easy to read that graphics processing units (GPUs) are the default choice for AI work and this is a problem for an enterprise. On Friday, the market acknowledged that the company faces some risks, with Nvidia shares falling 6% on broader semiconductor sales. But companies that make their own chips are buying record amounts of Nvidia, and that paradox sits at the heart of the story.
Image Source: Getty Images.
Amazon
Amazon has a very developed in-chip story.
The e-commerce and cloud computing company’s custom silicon business — the Graviton processor, the Triennium AI chip and the Nitro networking chip — topped $20 billion in annual revenue in the first quarter of 2026.
“If our chips business were a stand-alone business and sold chips made this year to AWS and other third parties, as other leading chip companies do, our annual revenue run rate would be $50 billion,” Amazon CEO Andy Jassy said on the company’s first-quarter earnings call.
Jassi added that the operation is now one of the top three data center chip businesses in the world.
However, none of this has dampened Amazon’s appetite for Nvidia. The company plans to spend about $200 billion in capital expenditures through 2026 — much of which leans heavily on Nvidia GPUs to serve customers that complement Amazon Web Services.
letters
Alphabet has been designing its tensor processing units, or TPUs, for more than a decade, and 2026 could be the year the work moves outside its own walls.
Google announced eighth-generation TPU systems as part of a strategy to no longer treat chips as internal-only. in May, black stone It announced a joint venture with Google to offer TPUs as a cloud service for rent, with an initial $5 billion commitment to bring 500 MW of capacity online by 2027. It was followed by an agreement to provide AI lab Anthropic access to 1 million TPUs and a previously announced lease agreement Meta platforms.
Now it’s ready to offer access to its custom chips outside of its own cloud, putting it in more direct competition with Nvidia than ever before. Even so, reports emerged this week that Google has signed a multiyear SpaceX cloud deal that includes access to 110,000 Nvidia GPUs; Alphabet keeps buying even as it builds.
Microsoft
Microsoft may be the laggard of the three in custom silicon.
Its effort is centered on the Maia accelerator, and a second-generation Maia 200 recently served in some data centers for some of the work behind Microsoft 365 Copilot and partner OpenAI’s models.
However, most of the AI work in Microsoft’s Azure cloud runs on Nvidia GPUs, so Maya is a way to recoup some of those costs over time, not a total replacement.
Meanwhile, the company’s spending is staggering: Microsoft expects to invest about $190 billion in capital expenditures during calendar 2026, and Azure — which saw revenue grow 40% in its fiscal third quarter (the period ending March 31, 2026) — appears to be capacity-constrained by the end of the year.
Today’s change
(-5.93%) $-12.96
Current price
$205.69
Key data points
Market Cap
$5.0T
Day limit
$204.33 – $214.87
52 week limit
$140.85 – $236.54
Vol
8.6M
Average Vol
166.8M
gross margin
74.15%
Dividend yield
0.14%
What this means for Nvidia
Combined, these three and the meta-platforms are on track to spend about $725 billion in capital expenditures in 2026 — about 77% more than last year.
Nvidia’s bear case is this: A growing share of that spending will flow to chips that its biggest customers design themselves, and they have a significant incentive to do what they can to become less dependent on a single supplier.
However, the bull case showed up in Nvidia’s own recent results. In the first quarter of fiscal 2027 (the period ending April 26, 2026), revenue rose 85% year-over-year to $81.6 billion, with data center revenue up 92%. Hyperscalers still make up half of that data center business.
“Demand has become parabolic,” Nvidia founder and CEO Jensen Huang said on the company’s fiscal first-quarter earnings call. He pointed to a fast-growing group of buyers — AI start-ups, corporations and governments — that “don’t build chips, don’t design their own chips.”
So, which way is this resolved? Maybe both at the same time, for a while. Custom silicon is real and growing, and it could erode Nvidia’s pricing power over time. But the overall pool of AI spending is expanding even more rapidly, and will continue to grow even as Nvidia loses its share at the margin. At a price-to-earnings ratio of 32, according to this article, the big risk is that domestic chips aren’t failing — they’re slowly winning as the market prices Nvidia for permanent dominance.
#Amazon #Alphabet #Microsoft #racing #design #chips #Heres #means #Nvidia #Motley #Fool