No company has ever been worth $6 trillion. Nvidia is about 5% away.
The stock ripped 20% in seven days, adding over $900 billion in market cap, and reports Q1 earnings on Tuesday. Wall Street expects $78.8 billion in revenue - a 79% jump from a year ago - and Nvidia has beaten estimates for eight straight quarters. At GTC in March, Jensen Huang told a packed house he expects $1 trillion in cumulative orders for Blackwell and Vera Rubin chips through 2027, double the $500 billion he projected at GTC just five months prior.
But the stock has fallen on three of its last four earnings days despite beating estimates every time, and Nvidia’s biggest customers are quietly building their own chips. Amazon’s custom silicon business just crossed a $20 billion annual run rate. Meanwhile, the CEO hitched a ride on Air Force One to Beijing this week, chasing a China market where Nvidia’s share has effectively fallen to zero under export controls.
At $5.7 trillion, either the AI buildout is just getting started or the market is paying for a future that’s already priced in.
Here's how the community voted
Nvidia guided Q1 at $78B and has beaten revenue by 3-4% in recent quarters, putting a $80B+ print in play. Blackwell chips are sold out through mid-2026, and Huang projected $1 trillion in cumulative Blackwell and Vera Rubin orders through 2027 at GTC, double what he estimated just five months prior.
Meta, Microsoft, Alphabet, and Amazon are set to spend a combined $725 billion on capex this year, up 77% from 2025, and the bulk of that flows through Nvidia’s supply chain first. The CUDA ecosystem locks developers in, and no competitor has cracked that moat. At ~27x forward earnings, the stock trades cheaper than the chip industry median of ~31x despite growing revenue 65% last fiscal year.
The China overhang may be lifting. The U.S. cleared H200 sales to 10 Chinese firms this week, including Alibaba, Tencent, and ByteDance, with Huang joining Trump’s Beijing trip to push the deals forward. If even a fraction of that market reopens, it is pure upside not priced into estimates.
Nvidia’s biggest customers are building their own chips. Amazon’s custom AI chip business hit a $20B annual revenue run rate in Q1 with nearly 40% sequential growth, Google just launched two new TPUs, and Microsoft and Meta are both expanding in-house silicon programs. Every dollar these hyperscalers shift to custom chips is a dollar Nvidia loses.
The stock has fallen on three of its last four earnings days despite beating estimates every time. At $5.7 trillion, the market has already priced in a blowout quarter. Options are pricing a ~7% move around May 20, and with the stock up 20% in a week, a “sell the news” reaction is the pattern, not the exception.
Export controls have effectively zeroed out Nvidia’s China revenue. Despite U.S. clearance for 10 firms, Beijing is blocking purchases to protect its domestic chip industry. Jensen Huang himself warned that Nvidia’s share of AI accelerators in China has fallen to zero - a meaningful chunk of the global market with no guaranteed path back.
Nvidia guided Q1 at $78B and has beaten revenue by 3-4% in recent quarters, putting a $80B+ print in play. Blackwell chips are sold out through mid-2026, and Huang projected $1 trillion in cumulative Blackwell and Vera Rubin orders through 2027 at GTC, double what he estimated just five months prior.
Meta, Microsoft, Alphabet, and Amazon are set to spend a combined $725 billion on capex this year, up 77% from 2025, and the bulk of that flows through Nvidia’s supply chain first. The CUDA ecosystem locks developers in, and no competitor has cracked that moat. At ~27x forward earnings, the stock trades cheaper than the chip industry median of ~31x despite growing revenue 65% last fiscal year.
The China overhang may be lifting. The U.S. cleared H200 sales to 10 Chinese firms this week, including Alibaba, Tencent, and ByteDance, with Huang joining Trump’s Beijing trip to push the deals forward. If even a fraction of that market reopens, it is pure upside not priced into estimates.
Nvidia’s biggest customers are building their own chips. Amazon’s custom AI chip business hit a $20B annual revenue run rate in Q1 with nearly 40% sequential growth, Google just launched two new TPUs, and Microsoft and Meta are both expanding in-house silicon programs. Every dollar these hyperscalers shift to custom chips is a dollar Nvidia loses.
The stock has fallen on three of its last four earnings days despite beating estimates every time. At $5.7 trillion, the market has already priced in a blowout quarter. Options are pricing a ~7% move around May 20, and with the stock up 20% in a week, a “sell the news” reaction is the pattern, not the exception.
Export controls have effectively zeroed out Nvidia’s China revenue. Despite U.S. clearance for 10 firms, Beijing is blocking purchases to protect its domestic chip industry. Jensen Huang himself warned that Nvidia’s share of AI accelerators in China has fallen to zero - a meaningful chunk of the global market with no guaranteed path back.
This poll has closed. New comments cannot be added.
There is a greater uncertainty projecting Nvidia because of the technological component. They transformed their business model significantly over the years and the creation of CUDA and powerful GPUs that exclusively run CUDA has helped them actually earn some "real" margin. This is in contrast to the older days where their hands were more tied and profitability swayed with console contract wins.
All that said, application specific computing has a risk. Bitcoin "mining" for example, became more efficient years ago on ASICs vs "general purpose" GPUs. You would lose money today in almost any scenario mining bitcoin on a GPU as a result of this.
I believe that dramatic change is a corollary for the world's AI processing ecosystem and hardware. There's a big incentive, both in terms of cost and performance, to lead to need-based ASICs. These needs could include state-sponsored projects (maybe they want to break encryption, etc.), hyperscalers looking to make their investments more efficient, new designs (Cerberus, etc.), and more. While Nvidia has the lead start, it's hard to beat application specific designs. The cost and timeline to build the ASIC is unknown to me, but seems more feasible today than it was 15 years ago. In part, it's because people are doing it, so an ecosystem will develop, but also because firms like TSMC have made it possible to build for smaller scale contracts (ironically, the exact reason Nvidia could break in).
While I'm not sure exactly how it'll work out, I think this has greater risk than Intel's X86 chipset design, because relatively fewer disparate items are needed to "work" to replace a GPU. X86 became a standard for many pieces, including the operation system, software, drivers, etc. from very fragmented sources and has always made it tougher to displace. With AI, the use cases might boil down to specific functions and programming that could be more efficient with application-specific designs and not the general purpose approach.
I am unable to quantify this as well, because I'm not in the AI space, but welcome others input. My general view is I doubt Meta, Google, Microsoft, Amazon, and state-sponsored entities are going to slow their ambitions to make ASICs that outperform what any general purpose GPU can do, and AI needs (by function) are probably fewer than what was needed of the X86 chipset, which has lasted for a long time.
This is not to say I expect any demise of Nvidia, but just a "lessened enthusiasm" if people's perception or estimates of organic growth rates were to pull back.