The AI Bubble :: By Todd Strandberg

Several times in the past, I have written articles that predicted some form of financial calamity that turned out to be correct. In 1999, I predicted the dot-com bubble would eventually burst. Before the 2008 meltdown, I forecasted that banks would experience a downturn. In 2020, I wrote a chapter for one of Terry’s books that warned that we were headed for trouble. Because it takes several months before some books are published, the Covid-19 crash occurred before the book’s release date.

You would think the predictable conclusion of all stock market bubbles would warn traders away from repeatedly making the same mistakes. History brims with financial manias, from the real estate bubble collapse in Athens in 333 BC to the Mississippi Bubble in 1720 to the US Cotton Panic in 1837 to the French Credit Debacle in 1868 to the Great Crash in the US in 1929, the 1990 Crash in Japan, the 2008 housing debacle, and on and on.

“Since we’re currently living in a bubble related to AI mania, I feel it’s time to sound the warning once again. This time, I don’t think I’m calling for anything special — it’s like predicting cold weather at the South Pole. Because many stocks have reached dangerously high levels, heightened concern is warranted. If a crash occurs, all markets will be affected. During every major recession, the Dow Jones has historically declined by 50% on average.”

AI production is widespread. Nearly 90% of organizations now use AI in at least one function, and enterprise GenAI spending rose threefold to $37 billion in 2025. In 1999-2000, e-commerce accounted for just 0.6% of US retail sales, and most companies had “digital brochures,” not digital operations.

The “Magnificent 7” AI-focused tech giants account for roughly 31.5% of the S&P 500’s total value, with the top 10 AI stocks reaching up to 35%. By comparison, dot-com stocks peaked at just 25% of the index in 2000.

I’ve used AI programs for research and to edit old family photos. I’m amazed at what the photo-editing feature can do with images. It can take a photo that looks hopeless and turn it into something near picture-perfect, even enlarging it to fill my computer screen. I’ve also used AI to create images of my family and friends meeting various celebrities in humorous situations. I had a friend in Iowa meeting Elvis and Charles Manson on his front lawn. In one of the images, I gifted my mother an AK-47, which she proudly modeled for the camera.

I don’t see the need to pay for these services because I can always find an AI program that does it for free. It’s like the old saying, “If the milk is free, why buy the cow?” I’m actually rapidly running out of photos that could need a touch-up. A few months ago, the top image editor would let me do a dozen a day. Now the daily limit is only three. Since I only have 30 more pictures to process, I won’t need AI’s services in a month.

I’ve run some photos through several AI image editors. ChatGPT is the best in the business. It does an amazing job of cleaning up images. I could have a picture that looks foggy, and the program will make it look crystal clear. With these programs constantly improving, a year from now, another program could leave ChatGPT in the dust.

Because the free editing party could suddenly end, I’m scrambling to upgrade all my family photos. I do realize that, in two years, a new program will likely edit images with zero questions and do tasks 10 times better than today’s programs.

The marketing budget for AI companies is larger than Coca-Cola’s. If you walk down a city street, you can see Coke ads on nearly every major building. While 80% of the population consumes some type of Coke product each day, most people will never use one of these AI programs.

With trillions of dollars having been spent on AI, people are starting to ask what all this money has bought us. Just like with the dot-com boom, many new companies will spring from this technology.

When railroads came to America, there was a massive boom in railroad stocks. Most of them couldn’t make money, so they went out of business. The railroad boom blessed us with many miles of track that survived all these bankruptcies. Most AI companies will probably follow the same pattern and go out of business.

The Greater Fool

When you buy a share of stock, you’re purchasing a stake in a company’s underlying assets — the desks, chairs, computers, pencil sharpeners, and everything else the business owns. Under normal circumstances, a company’s stock price reflects the market’s collective expectation of the future income that company can generate using those assets. Stock valuation, in theory, is really just a bet on future cash flow.

But many of today’s companies are trading at valuations we’ve never seen before—multiples that would have seemed absurd a generation ago. A striking number of these companies have little to no operating income; some don’t even have a clear path to profitability, and yet their total market values stretch from hundreds of millions of dollars into the tens of billions.

Here’s the problem: the prices of these stocks are no longer tethered to future earnings potential in any meaningful way. Instead, they’re propped up almost entirely by hope — the belief, or perhaps the wager, that someone else will come along later and pay an even higher price for the same shares. It’s less an investment thesis than a game of musical chairs, where everyone assumes they won’t be the one left standing when the music stops.

This is the essence of the “greater fool” theory. An investor who was foolish enough to buy a stock trading at 200 times earnings suddenly looks wise — even brilliant — the moment he finds someone even more foolish willing to buy it from him at 1,000 times earnings. The original buyer wasn’t right; he simply wasn’t the last one holding the bag.

The Old King: Cisco Systems

Historians widely regard Cisco Systems as the poster child of the dot-com bubble because it represented the bubble’s ultimate high and its most devastating crash.  Unlike many dot-com companies that were purely speculative ideas with no revenue, Cisco was a real, highly profitable powerhouse selling the “picks and shovels” (routers, switches, networking hardware) that built the early internet. Yet, despite being a genuine market leader, its stock became swept up in an unsustainable valuation loop.

In March 2000, Cisco briefly surpassed Microsoft to become the world’s most valuable company, peaking at a market capitalization of around $550 billion. Its price-to-earnings (P/E) ratio climbed past 400. Investors assumed internet traffic would grow exponentially forever, pricing Cisco as if it would capture almost all of that growth without slowdown or competition.

When the dot-com bubble finally popped, Cisco’s stock suffered a massive loss. It collapsed by roughly 88%, dropping from its peak above $80 per share down to under $10 by 2002. The crash wiped out over $430 billion in market value.

Cisco’s stock declined not because the firm ran out of customers. To fuel its massive growth rates, Cisco engaged heavily in vendor financing—loaning money to cash-strapped dot-com startups and telecom firms so they could buy Cisco hardware. During the late 1990s, demand for routers and switches was so intense that Cisco had backorders lasting months. Fearing shortages, clients began placing identical orders with multiple suppliers or ordering twice what they actually needed, intending to cancel late orders once the first batch arrived.

The New King: Nvidia

If any single company deserves to inherit Cisco’s crown from the dot-com era, it’s graphics card maker Nvidia. And it hasn’t just matched Cisco’s old bubble valuation — it’s blown past it, tacking on an extra zero and then some. Nvidia’s explosive growth has convinced investors to push its market cap beyond $6 trillion, an almost incomprehensible figure for a single company. Because its earnings have kept pace with its stock price, at least on paper, the company still trades at a forward P/E of only around 35 — a number that looks almost modest next to its market cap, even if the underlying assumptions baked into it are anything but modest.

Still, Nvidia has its own unique vulnerabilities lurking beneath the surface. The company relies on a small handful of hyperscalers — Microsoft, Alphabet, Meta, Apple, Tesla, and Amazon — for roughly 70% of its revenue. That kind of customer concentration means any shift in how the AI business model works, or in how much these giants are willing to spend, could hit Nvidia’s bottom line hard and fast.

This isn’t hypothetical: a few years back, when the cryptocurrency market fell into a bear market, Nvidia felt it directly. Since most of the mining and transaction-processing hardware powering crypto ran on Nvidia GPUs, the company saw a sharp, sudden downturn in sales once that demand dried up.

Another looming concern is Nvidia’s extraordinarily high product pricing. The company has maintained exceptional gross margins — somewhere in the 70-78% range — largely on the back of extreme GPU scarcity. Some of its top-tier GPUs now sell for as much as $6 million apiece. To put that in perspective, the very same package sold for roughly $40,000 just a couple of years ago — a staggering increase in a remarkably short window. Meanwhile, many of Nvidia’s traditional rivals are racing to develop their own in-house chips, a trend that could eventually erode Nvidia’s near-monopoly on the AI hardware market.

The Invisible $3 Trillion Financial Time Bomb

The amount of money that AI firms have spent is already shocking, but they have expenses that don’t show up on earnings reports. Meta has a $26 billion partnership with Blue Owl Capital to build data farms it will use when they’re finished. The $26 billion is currently not on Meta’s balance sheet. It is a future obligation that will eventually become an ongoing expense.

AI companies face roughly $3 trillion in total off-balance-sheet obligations, prompting economists to question whether these firms can sustain these financial commitments during a market downturn. Central to these concerns is CEO Jensen Huang’s proposal to guarantee the resale value of Nvidia’s GPUs. Given how quickly AI technology depreciates, analysts remain skeptical about how such guarantees could be implemented in practice.

A major risk of these undisclosed agreements is that they only come to light when problems arise. Capital is easy to secure during economic booms, but when market conditions worsen, lenders rush to pull out simultaneously. Blue Owl has already drawn media attention as investor anxiety grows over capital redemption risks. In response, the firm was forced to gate several illiquid funds—including one small fund whose valuation crashed from 100% secure to zero value in a single announcement.

In a notable financial move, Nvidia met with major banks and private equity firms to discuss raising $500 billion for AI infrastructure. Under the proposed model, investors would fund data center projects that purchase Nvidia chips, while hyperscalers commit to leasing the facilities. Nvidia would then reinvest its earnings back into those hyperscalers, fueling further demand for its hardware.

Nvidia is also backing $105 billion in financing for OpenAI data centers. Nvidia is willing to act as the banker in these deals because all the resulting GPU sales flow back to the company. However, if the data center starts losing money, Nvidia would be on the hook to cover the shortfall.

The Disaster Everyone Knew Was Coming

What sets the AI bubble apart from nearly every bubble that came before it is that most people already know we’re in one. Roughly 54% of fund managers now describe AI stocks as being in bubble territory. Numerous articles in the financial press have plainly stated that AI valuations have reached extreme levels.

Yet many people refuse to sell their winning stocks because they don’t want to pay steep capital gains taxes. Others simply believe stocks are meant to be held for life. And because some investors have already seen massive gains, a 30-50% correction doesn’t worry them.

This summer, I spoke with a gentleman who owned Nvidia stock. He has no plans to sell because he doesn’t want to miss out on the next double. The company’s value already equals about 5% of the nation’s GDP. It’s hard to imagine it reaching 10%. Of course, there’s no real limit to how high a stock can climb — the day may come when it reaches the full $36 trillion of our GDP.

GameStop, a video game, consumer electronics, and gaming merchandise retailer, has no real long-term future — at some point, all video games will simply be sold online. At its peak, GameStop operated 7,535 stores worldwide during fiscal year 2016. By January of this year, that number had fallen to 2,206.

One reason I would never short a stock is that losses on a short position are theoretically unlimited once the trade turns against you. In 2021, after retail investors on Reddit noticed that short interest exceeded 100% of GameStop’s available shares, the stock price skyrocketed from $17.25 to over $500 per share. Retail buying drove the price up, forcing short sellers to buy back shares at enormous losses. Five years later, the stock is back down to around $17 per share.

GameStop offers a valuable lesson for AI stockholders, many of whom are similarly long-side leveraged. An investor trading with 3x leverage only needs the stock to decline by 33% to be wiped out entirely. This has already played out in South Korea, where the top AI index crashed by a record amount due to massive long speculation.

AI stockholders who assume they can quickly exit once market conditions turn may be setting themselves up for a 1987-style crash — an event where a downturn that would normally unfold over months instead hits the market in a single day.

Conclusion

I’m not going to predict that the AI bubble will lead to a meltdown of the entire economy. We’re currently in the 18th year without a recession, even though one typically occurs roughly every five years.

The cryptocurrency bubble was the first thing that made me realize centuries-old financial rules seem to be broken. We saw the buildup with crypto, but never really saw the bust — it just kept expanding instead. We now have nearly 3 million different cryptocurrencies in existence. Bitcoin looked like it was on the verge of crashing several times, yet each time it dropped, stabilized, and then climbed back to a new high.

Bible prophecy is the only thing that explains how the system keeps holding together. Jesus promised that when He returns for the Church, it would come during a time of general economic tranquility. Despite a long list of looming dangers, we largely continue to live in a carefree world.

“But as the days of Noah were, so shall also the coming of the Son of man be. For as in the days that were before the flood they were eating and drinking, marrying and giving in marriage, until the day that Noe entered into the ark, and knew not until the flood came, and took them all away; so shall also the coming of the Son of man be” (Matthew 24:37-39).

We’ve dodged several economic bullets over the past few years. If the AI bubble comes and goes without leaving any lasting damage, I’ll take that as a sign that the rapture is drawing very near.

“Therefore be ye also ready: for in such an hour as ye think not the Son of man cometh” (Matthew 24:44).

 

Donation Drive September to November

It has now been 39 years that Rapture Ready has been an online resource for end-times information. I started the site sometime in September of 1987 while I was an airman stationed at Offutt Air Force Base. My computer was an Apple II clone made by Franklin Computer Co. It featured all green text, and there was no mouse.

A few weeks ago, I was talking with Terry, and I actually thought we were only coming up on our third decade. I did some quick math in my head and realized the big 4-0 was coming up. I was also shocked that I’ve been retired from the military for nearly as long as my 20-year service.

There was no grand opening for the site because, for the first few years, few people cared what I posted in the global newsgroups system–now extinct. When the graphic Internet came along with the 1994 release of Netscape Navigator, I switched to posting the site on a stand-alone domain.

The changes I find most amazing are Rapture Ready’s growth. In the early days, I could go on a two-week trip and come back without anyone noticing my absence. Today, I have to check something related to the site at least six times a day. The traffic went from zero to one thousand in the ’90s, to tens of thousands in the 00s, and to hundreds of thousands in the past few years. Today we measure our traffic in millions of hits.

I was recently watching a YouTube video, which made me realize that the current AI craze could be very helpful to the site. It has long been obvious that once the rapture takes place, there will be no one left to manage the Rapture Ready domain. I once jokingly asked for someone to volunteer to stay behind to manage the site.

For decades, computer programs have required manual labor to tell them what to do at each stage. We input the information, and the program would produce results. Now, there are AI agents that will quickly be able to act as free agents. They will do work that once required humans.

Computers would have to advance by many orders of magnitude before I would ever let one run the site unsupervised. There is one simple task that AI could do at any time. Once the rapture occurs, there will be a massive surge in our server traffic. An AI agent could help by significantly increasing the site’s bandwidth.

We are currently on a server that can handle ten times the daily traffic load. We could program AI to upgrade our host plan if traffic surges past a given level. We could tell AI to go crazy with credit cards if news of millions of people missing hit the news feeds.

Another way AI is helping RR is by enabling a vast amount of money to be spent on creating better chips. Servers are getting more and more powerful because AI is used to advance the technology that makes them run. It used to take several years for a new generation of a software program to be released. Because AI helps design itself, these releases come out every few months.

We could be approaching the point where bandwidth is no longer a factor. It has been several years since I’ve had to deal with any limits on my personal internet use. We will eventually reach the point where web hosting providers will have equally endless resources.

Our first use of AI will likely be a plugin that will answer people’s questions. Visitors who are new to prophecy don’t know where to go. RR runs on WordPress, and the latest version has options for AI providers like OpenAI, Claude, and Gemini.

We have a page that explains how to create mirrors of Rapture Ready. Because the post-rapture world will be a time of panic, people might be too distracted to share the contents of this site. Also, time will be very limited. All prophecy sites will likely be scrubbed from the internet within a few days.

As we get closer to the tribulation hour, every Christian needs an end-times mindset. If you truly believe the rapture is drawing near, you should be more focused on endeavors that will contribute to the Kingdom of God.

There are three ways you can contribute to RR: Send a financial gift via our PO Box address; make a one-time donation; or make monthly donations. If you’re interested in signing up, find the links on the Donation Depot page. Donations can be sent to our PO Box at:

Rapture Ready
P.O. Box 969
Benton, AR 72018