One Billions Isn't Enough

A hundred billion here and hundred billion there. Soon we will be talking real money. But now we are talking about real money when it comes to infrastructure investments for artificial intelligence.

Over the weekend, I was reviewing several billion-dollar market cap companies. Anyone who values public stocks is familiar with $5-10 billion valuation companies that are projecting delivering $1 billion in revenue or more. One billion dollars is a meaningful amount for 99.99% people of the world. However, after Nvidia's announcement today, I am revisiting all prior estimates again.

Nvidia's $500 Billion Investment

I've been using Nvidia's GPUs for two to three years. Every single A.I. app I use is powered by Nvidia's H100 chips because of how great they are. Much better than AMD chips in most cases. In fact, if you think about GPUs, Nvidia has almost no competition because they are in a league of their own.

Below is a revenue segment chart from FinChat for Nvidia. Since ChatGPT launched in December 2022, the demand for Nvidia GPUs has skyrocketed. At the same time, Nvidia's profit margins increased from 48.8% to 55.8%. That's unbelievable growth in the right direction.

What's interesting to observe, is that the rising tide has lifted all boats for Nvidia's other revenue streams. Gaming and automotive are all rising but data center revenue growth simply overshadows everything.

Nvidia's New Supercomputer

As an outsider, you would think that even data center revenue growth would have its limits. There are only so many customers and the market does not have infinite demand. But I was definitely wrong in these estimates. The demand for real intelligence, not artificial, is limitless in many ways. The challenge for humans is how to manufacture super intelligence at scale, while being profitable.

One way Nvidia will accelerate the demand for superintelligence is investing $500 billion in Arizona to manufacture A.I. supercomputers. This is a four-year project and expected to ramp up production for Nvidia's Blackwell chips at TSMC's plants in Phoenix, Arizona. NVIDIA is also building supercomputer manufacturing plants in Texas, with Foxconn in Houston and with Wistron in Dallas. This is a massive investment with a multi-year horizon.

If you look back at the revenue chart, you will see that data center revenue is currently at the $115 billion annual revenue run rate. But now Nvidia plans to invest $125 billion per year in production. With a four-year timeline, I think Nvidia's management will target to double their annual data center revenue to +$200 billion in the next 48 months. These are massive numbers and hard to comprehend.

Tariffs and Domestic Production

I think the major driver for this manufacturing plan is to drive revenue growth. Of course, the Trump Administration will incentivize new investment. That's one of the reasons why TSMC plans to invest an additional $100 billion in new semiconductor manufacturing in the United States. Again, these are massive numbers and investors haven't seen this level of investment ever.

Do I think Nvidia's announcement has to do with the recent tariff announcements? It's very likely because the press released was timed so well. But the real concern with A.I. and semiconductor manufacturing are the national security issues. Given how powerful A.I. has become, it is clear that the U.S. needs to command independence on this specific issue. Offshoring too much semiconductor manufacturing has become a national security for every day Americans.

All of the players and dollars mentioned above are bullish for the technology sector. I think demand will continue to rise, but there is one risk investors need to be aware of. Overbuilding for future demand can result in negative returns, or a permanent loss of capital.

Today it is easy to say that the demand for intelligence is infinite but that might not match the demand by customers. If Nvidia isn't able to double data center revenue in the next 48 months, then this $500 billion investment will breakeven at best. However, I wouldn't bet against Nvidia and think they will continue to deliver on the $100 billion run rate. A.I. chips will incrementally get better overtime and more companies will use the technology to supercharge their workforce.