Nvidia just earned $58 billion profit in three months. It also sold off its entire China unit. It is true, and not true in the same instant.

Last Wednesday, Nvidia dropped numbers that made you do a double take. For the period ended April 26, revenue rose 85% year-on-year to $81.6 billion and compared with Wall Street's expectation of $78.9 billion. The net was $58.3 billion, beating consensus forecasts of $42.9 billion among analysts polled by FactSet.

For context, Nvidia ranked near the top of a class where few belong with its $58 billion quarterly profit. That is in the leagues of companies like Apple, the most profitable company on earth for much of the last decade.

At the foundation of everything is data centers. The division pulled in $75.2 billion of revenue, up 92% year over year primarily from its Blackwell 300 chip line and the networking hardware that glues thousands of chips together. If that prospect isn't enticing enough in the broader context of Nvidia's Q2 revenue guidance of about $91 billion, it means the company is going to be even bigger next quarter than it was this one.

Then there is China.

Nvidia is out of the $50 Billion market China accounted for about 20% of Nvidia data center sales a year ago Today that number is zero. Not down. Not declining. Zero.

It is worth understanding the events that led here. In April 2025, sales of the H20 chip to China were prohibited, only for them to be re-allowed in July. December's approvals were not without strings attached, however; the US government will require 25% of revenue share.

The shipments were subsequently halted by Chinese customs. Ten big Chinese tech companies – including Alibaba, Tencent and ByteDance – were eligible to purchase huge amounts. None of the deliveries happened. China is unwilling to authorize the imports because it wants to develop its own semiconductor industry.

Jennifer Huang states that the total AI chip market in China is just $50 billion. For a company of Nvidia's size, that's about $38 billion a year in revenue lost due to 20 percent of data center revenue. Right now, the company is expanding enough to be able to offset that loss. The question is how long, and whether Huawei's Ascend chips can close the performance gaps in the timeframe that Nvidia has been forced out.

Two Problems Money Cannot Completely Solve

The earnings were nearly flawless. The guidance was strong. But there are double structural constraints which reset unable to extrapolate indefinitely this trajectory.

The first is power. It takes between 12 to 24 months to build a data center. High-capacity grid connections can take between three and seven years to obtain. The backlog at the US grid is already > 2,600 gigawatts. Just 5 gigawatts of the 12 gigawatts of AI data center capacity set to come online by 2026 are currently under construction, while others remain stalled behind power and transformer delays.

Organizations such as xAI, Meta, OpenAI and even Oracle are constructing their own power systems from the ground up completely disconnecting from the grid in deep debt. Chip supply limits from TSMC helping set standards for manufacturing and ASML machines needed to produce leading-edge chips could take another two or three years to solve, Huang said.

The second is debt. In 2025, big tech firms issued $121bn of US corporate bonds, more than quadruple their average in history. That figure is set to hit $175 billion by 2026, according to Bank of America, after Amazon's record-setting $54 billion bond sale in March. Excessive borrowing raises capital costs for all of the affected companies, such as Nvidia customers. Slowing data center investment If borrowing becomes prohibitively expensive.

Where this actually leaves things

This Nvidia report still matters if you have been watching the crypto and AI intersection because it provides insight on how to think about the infrastructure arms race in dollars. Major four cloud providers set to spend about $700B on infrastructure fueling AI in 2026, up by more than 60% from already record levels seen last year Light sheds light Huang sees $1 trillion addressable market for Blackwell and Vera Rubin chips through 2027

The same constraint is the availability of chips. Demand, there is no issue with. What the results from Wednesday did was to answer the only question that really matters: Is this spending on rugby-land bubbling up into real revenue at the company building the picks and shovels?

Hard to argue with that, at $81.6 billion in a quarter.

Why are quantum stocks surging today?

Quantum stocks rallied today because the U.S. government is investing $2 billion in grants to nine companies working on quantum computing.

The National Institute of Standards and Technology said it signed letters of intent with the companies. The U.S. government will also take minority stakes in each business, but those stakes will not give it control.

The biggest award is going to IBM (IBM), which is set to receive $1 billion from the Commerce Department. IBM stock traded about 7% higher after the announcement.

Join the Conversation!

We'd love to hear your thoughts and comments. Join our community and stay updated with the latest trends and discussions in crypto.

Twitter | Instagram | Telegram Channel | Linkedin | Facebook