AI and Fiber Overbuilds

Plus, links to a new podcast on John Deere!

YouTube

Spotify

Apple Podcasts

Today, we dropped a new episode with

Best Anchor Stocks

on Deere & Co. (parent company of John Deere). Links above. For the newsletter, I have some thoughts on a different topic.

---

AI and Fiber Overbuilds

Everyone knows the story of the dotcom bubble bursting in 2000. However, few know there was a simultaneous telecommunications bubble.

The idea was that internet usage was growing like wildfire due to the dotcom boom. Everyone would like access to the internet, so we need to build the infrastructure to allow people to get online. Everyone sounds like a large addressable market.

Check out this article from May 1999 in Forbes:

Traffic on the Web has indeed been doubling every three months. About 17 million homes already have two or more PCs. Communicating chips are now migrating off the desktop. Electrolux recently announced its "Internet refrigerator," an embedded PC replacing the scribbled note and door magnet. GE has an Internet microwave oven. EmWare, a software company, is working with Sybase, 3Com and Micron to bring vending machines on-line to make stocking and management more efficient.

And:

The global implications are enormous. Intel projects a billion people on-line worldwide. That's $1 trillion in computer sales -- and another $1 trillion investment in a hard-power backbone to supply electricity. One billion PCs on the Web represent an electric demand equal to the total capacity of the U.S. today.

Analysts believed that internet demand would grow indefinitely, leading to an indefinite boom for semiconductors, computer networking equipment, and electricity demand.

But in order to connect people to the internet, you need to build out the telecommunications infrastructure. Which led to an absolute flurry of internet fiber start-ups. From the Wall Street Journal in 2001:

All told, about 39 million miles of fiber-optic cable stretch underneath U.S. railroad beds, corn fields, natural-gas lines and roads, enough to circle the earth 1,566 times. Companies racing to build or expand nationwide networks laid some $90 billion of fiber during the past four years.

According to the article, telecoms “gorged” on $650 billion in debt. Wow.

I think we know how this story ends. What’s funny is that these analysts were correct about internet usage. 25 years later, most of the world is online with multiple internet-connected devices.

The problem was that it took 25 years to get here. And we got a lot more efficient with electricity/semiconductors and all the other infrastructure of the digital world (not going to pretend to be an expert on this stuff, but the predictions in the Forbes article were wildly off even though demand was higher than expected).

Here is what happened when the telecom bubble burst in 2001. From the same WSJ article:

The fiber glut underlies much of the uncertainty plaguing the telecom sector -- and has even spilled over into the economy at large. Billions of dollars in shareholder value have evaporated in some of the biggest owners of fiber networks, including Global Crossing Ltd., Williams Communications Group Inc. and Genuity Inc.

Some analysts tried to sound a warning. In October 1998, an Internet researcher at AT&T Labs named Andrew M. Odlyzko published a paper debunking the widely held view that Internet traffic was doubling every three months. Mr. Odlyzko laid out an argument that in fact Internet use was doubling only once a year. "It was an extremely convenient myth," says Mr. Odlyzko. "Every entrepreneur who was getting financing could quote it."

The stock prices of all these fiber players eventually plummeted, many going bankrupt. Only 2.6% of the fiber laid out was actually in use in 2001, according to industry analysts.

One of the CEOs is quoted in the article as saying “There wasn't enough revenue to go around.” This means they all spent so much on fiber but got years ahead of the revenue opportunity. Can an industry survive with so many years of negative ROIC? No, which is why telecommunications/fiber went into a steep downcycle in 2001. It had to rationalize.

Which brings me to AI. This estimate caught my eye in the Wall Street Journal on June 1st:

An oft-cited figure in arguments that we’re in an AI bubble is a calculation by Silicon Valley venture-capital firm Sequoia that the industry spent $50 billion on chips from Nvidia to train AI in 2023, but brought in only $3 billion in revenue.

Aren’t we making the same mistake as 25 years ago? If these estimates are correct, today the ROIC on all these AI projects is wildly negative.

Perhaps the AI sector (I’m not sure what that even means though?) will grow at a similar rate to the internet in 1999 over the next 25 years. This is a wildly optimistic outcome. If that occurs, a lot of companies will still go bankrupt or experience 90% drawdowns if they build supply 10 years in front of demand.

Obviously, Nvidia is not going bankrupt. But don’t think it isn’t at risk of a downcycle just because it has been invincible for two years.

This rumor also caught my attention:

CoreWeave, the cloud AI startup heavily backed by Nvidia and private equity firms, is reportedly seeking to become a public company in 2025 fresh off a $1.1 billion funding round

CoreWeave used to be a crypto miner but has pivoted to being a competitor to AWS, Azure, and Google Cloud for AI infrastructure services. It buys a lot of expensive Nvidia chips and is raising money quickly.

It just raised $1.1 billion in equity and has raised $7.5 billion in debt. If it IPOs, I’m sure it could raise over $10 billion.

The parallels to the telecom bubble seem obvious. It likely won’t follow the same path, but I think there is a major risk all these AI companies are building out way too quickly. This could eventually lead to a supply glut and a huge downcycle.

Of course, I could be wrong. Who knows what the next few years look like in this fast-moving industry. But if you are optimistic about AI stocks, I want to leave you with two quotes.

First, from the dark fiber WSJ article:

It's even worse for so-called "dark fiber," which hasn't yet been connected to the expensive electronic equipment needed to make it usable. A major brokerage house, say, that wanted to purchase its own strand of fiber could pick it up for $1,200 a mile, down from as much as $5,000 in 1997, according to Qwest.

Second, from this article comparing Nvidia to Cisco Systems:

Only 20% of the stocks trading with a price/sales ratio between 30 and 40 outperformed the market over the next 12 months. The more you extend your time horizon, the worse the results get. Over a 10-year period, that number drops down to 6%.

Investors better hope revenue (not just usage!) from all these expensive AI tools shows up quickly. Otherwise, this will end poorly.

---

Chit Chat stocks is presented by:

Public.com just launched options trading, and they’re doing something no other brokerage has done before: sharing 50% of their options revenue directly with you.

That means instead of paying to place options trades, you get something back on every single trade.

-Earn $0.18 rebate per contract traded
-No commission fees
-No per-contract fees

Options are not suitable for all investors and carry significant risk.  Option investors can rapidly lose the value of their investment in a short period of time and incur permanent loss by expiration date.  Certain complex options strategies carry additional risk.  There are additional costs associated with option strategies that call for multiple purchases and sales of options, such as spreads, straddles, among others, as compared with a single option trade.

Prior to buying or selling an option, investors must read and understand the “Characteristics and Risks of Standardized Options”, also known as the options disclosure document (ODD) which can be found at: www.theocc.com/company-information/documents-and-archives/options-disclosure-document

Supporting documentation for any claims will be furnished upon request.

If you are enrolled in our Options Order Flow Rebate Program, The exact rebate will depend on the specifics of each transaction and will be previewed for you prior to submitting each trade. This rebate will be deducted from your cost to place the trade and will be reflected on your trade confirmation. Order flow rebates are not available for non-options transactions. To learn more, see our Fee Schedule, Order Flow Rebate FAQ, and Order Flow Rebate Program Terms & Conditions.
Options can be risky and are not suitable for all investors. See the Characteristics and Risks of Standardized Options to learn more.

All investing involves the risk of loss, including loss of principal. Brokerage services for US-listed, registered securities, options and bonds in a self-directed account are offered by Open to the Public Investing, Inc., member FINRA & SIPC. See public.com/#disclosures-main for more information.

---