원문 보기
When I look at AI, I think we just have to do some measurements. ChatGPT was launched in 2023. So, for 3 years now, we have been developing innovations that come from artificial or superintelligence. The public internet was launched around 1994, okay? And that's when mobile networks were being built .
The total costs of developing the Internet, mobile communications, and telephones in 2026 prices amounted to about 35 trillion. AI infrastructure spending is around 30 trillion. So it costs about the same . That's the thing. Cisco's peak was only in 2000, okay? And between 1993 and 2000, Cisco fell by 50% twice.
Why did Wall Street's most famous bull just compare Nvidia to Cisco, the symbol of the 2000 tech crash, and call it good news? You've seen Tom Lee's videos from Fundstrat where he says AI is now at the 1997 stage and the bubble is expected to peak in 2033, and you probably think that means another 6 years of easy money.
Hmm, listen, in this video let's try to test Tom Lee's 1997 Cisco analogy based on actual historical declines and current market rotation data. But the main thing you need to know right now is that even if Tom Lee is 100% right that we are still in a 1997 moment, the same year Cisco fell 45%. Even more interestingly, I will prove that the arguments of the "bulls" and " bears" have quietly merged into the same warning.
Oh, and by the way, it's a small channel that's still trying to survive. A like, a follow, or a share with friends—that's the only favor a cat can ask for. That's all. Now let's take a look at this timeline on your screen. In July 2023, Tom Lee compared Nvidia's stock growth to Cisco's profits from 1990 to 1997, arguing that a true 1999-style bubble meant that more than tenfold growth was ahead.
Now look at the mark on the right that indicates "now". It's been a full three years, but paradoxically, look how that digital clock above resets back to 1997. Lee says we are far from a bubble, and immediately adds that the peak of the bubble will likely be around 2033. Now look down. In our last report, Citadel's Ken Griffin called this almost certainly a bubble.
Ray Dalio calls this a classic bubble . And now the biggest " bull" on television is calling you the year of the peak of the bubble. The Bears and the Bulls now agree on how this movie ends. The only disagreement here is when exactly the fire will break out in the theater. If you follow this long enough, you'll notice that not only is Lee's clock standing still , but the special security test he created to prove we're not in the year 2000 has failed this quarter.
And worse, if you take his word for it and start investing as if it were 1997, historical price data points to a trap that destroys almost every retail investor who tries to profit from it . Before we continue, I want to emphasize: I'm not trying to be a Michael Bury-style "bear," but my duty is to help you take an objective look at how this story is changing .
You see, according to Lee's analogy, back in 2023 it was 1997. Three years later—it's still 1997. In July, Lee warned that Cisco shares had halved at least four times between 1993 and 2000 . In October, he said that Cisco had only fallen 50% twice. When the clock never moves forward and historical dips change depending on the interview, you are not looking at a valuation model.
You are listening to a fairy tale. Well, to be honest , investors are crazy about stories. I won't judge whether it's good or bad. We need them, but we shouldn't depend on them too much. More importantly, in July, Lee gave us a brilliant, testable rule to distinguish 1997 from 2000. In July, Lee claimed that Cisco's customers in 2000 were speculative fiber-optic companies that used unrealistic discounted cash flow models— just Wall Street jargon for valuing a company based on imagined future earnings.
In the meantime, look at the green column on the right. Lee noted that today's chip buyers are giant hyperscalers with solid balance sheets. This was true, but only until recently. Now look what happens when the calendar flips to October. As we demonstrated in our last review, Alphabet, Meta, and Amazon are now borrowing tens of billions of dollars or issuing stock just to keep building data centers.
Now look at that red box that appears at the bottom . Debt-funded neo-cloud companies like CoreWeave are currently selling off as investors retreat. They play the same economic role as credit fiber optic operators in 1999. If judged by Tom Lee's own balance test , the customer base is no longer safe .
It is actively being converted into the same borrowed funds that caused the crash of 2000 . But that's just the worst-case scenario, right? That is, a cat will usually jump anyway, even knowing that it will knock over things around it. The main thing is that the cat will still land safely, right? Well, now let's assume that we ignore the debts and believe that Tom Lee is absolutely right.
Let's assume that it is indeed early 1997, and AI stocks will increase several times by 2033. What was owning Cisco stock from 1997 to 2000 really like for your peace of mind and brokerage account ? Look at the red decline at the very beginning of this line graph. Using exact numbers from Lee's interview, Cisco was trading around $9 in early 1997.
Later that year, it fell to around $5. That's a 45% collapse in the very year that Lee says we're now repeating. Then watch as the line rises to $18 in 1998 , only to collapse again during the Russian debt default and the Long-Term Capital Management hedge fund crisis. So, eventually, in March 2000, it soared to $80.
And immediately after that, look at the chasm on the far right side. It collapsed by almost 90%. So, to actually save money as predicted by Lee in 1997, two brutal tests had to be passed. First, you had to survive at least one approximately 50% crash without panic selling. Second, you had to magically guess the peak and sell by 2001.
And most average investors failed test number one because of borrowed money. We just saw this happen in real time. A surge in leveraged products in South Korea recently turned a normal semiconductor pullback into what Lee himself called forced deleveraging. This means that brokers automatically liquidated investors' stocks on the day because they were buying stocks with leverage.
And just in time, the SEC has just approved triple- leveraged funds on Bitcoin and Ethereum for the public . If you hold a triple leveraged fund, for example, as I know there are people who are currently recommending buying 3X SOXL, then during a 1997-style 34% pullback, your account is practically wiped out.
You won't live to see 1998, let alone 2033. Even a cat only has nine lives, and I guarantee you that Wall Street's credit products are designed to burn through all nine before lunchtime. Now, before you think I'm just kicking Tom Lee's ball of yarn off the table for fun, we need to look at where his data is absolutely correct, because under the 1997 headline, Lee also made a massive portfolio change back in July that almost no one talked about.
Look at the left side of the screen. Back in July, Lee warned that chipmakers, especially memory companies, were facing the risk of a "bullwhip effect" due to customers double- ordering during shortages. This is exactly the same shift from hardware manufacturers to software and margin beneficiaries that we outlined in our last video.
And look at that red bar, memory stocks fell a whopping 25% last week alone. Now , to your right, Lee notes that corporate earnings are really strong, with S&P 500 earnings growth in the third quarter at about 23.8% . Meanwhile, hedge fund net exposure is at its lowest level since the tariff collapse in April 2025 , meaning large institutional managers are underinvested and holding cash on the sidelines.
This is exactly the moment that makes me stop and question my own caution about the market in general. With third-quarter earnings up nearly 24% and hedge funds hiding in cash since the tariff panic, the bear could easily be overtaken by a catch-up rally. I used to build financial models at those Fortune 50 companies before I traded my office cubicle for a scratch pad.
And the most difficult discipline in investing is separating a bad company from a bad stock price. Technology developers are a great business, which, frankly, makes betting against their momentum terrifying. But look at these two bars competing at the bottom of the graph from October 1st to October 19th.
The standard cap index, dominated by giant AI developers, rose just 0.51%. Meanwhile, the balanced S&P 500 index, where every company from railroads to banks has the same weight as Nvidia, rose 2.56%. For the first time, it finally outperformed the tech-heavy index by a factor of five. So, here's what this gap between 0.5% and 2.5% really means.
Tom Lee looks at the low hedge fund position and the 25% drop in memory stocks and interprets it as a temporary slump to buy more AI hardware, which could still be true. But the data suggests that something much larger is happening. This five-to-one lead over the equilibrium index is not a random spike.
Combine this with what we saw in the BLS data , where profits for non-financial corporate divisions increased by 43%, and the picture becomes unmistakable. Institutional money is already quietly flowing out of the companies that build servers and into 493 other companies that are using the technology to increase their profits.
And remember , who else reads Cisco's chart like that? Even Tom Lee quietly shifted his own priorities toward software and users back in July. When a strategist goes on TV and screams at retail viewers about 1997, while his own research warns of the bullwhip effect in memory chips and favors lower- level beneficiaries, look at what his research does, not what the TV line says.
So, how to invest your money today to avoid falling into a 1997-style 45% “air hole” ? This is not personal financial advice, but essentially the same boring scenario that I keep repeating. Size each AI stock you hold as if it were to fall 50% tomorrow. Remember, Tom Lee's Cisco analogy involves falling by half at least twice before reaching a peak.
Look at the value of your tech assets now, divide it in half, and honestly ask yourself: if it fell to this level next month, would you start panic selling? If the answer is yes, your position is too large. Short it today while prices are high. On the other hand, complete the reorientation to beneficiaries.
The fact that the equal-weight index is outperforming the capitalization-weighted index indicates that the train has already moved. Direct your new contributions to broad equal-weight funds or profitable, margin-enhancing non-tech companies. Additionally, apply the cash flow rule to your large tech stocks .
Leave the hyperscalers who still pay for their operations solely through operating cash flow. Automatically downgrade or cut the stake of any giant that is forced to borrow money or issue new shares to continue growing. And finally: stay away from the recently approved triple- leveraged crypto and tech products, and limit your crypto exposure to the low single digits of your capital— an amount you can afford to lose completely, regardless of what the CEO of Bitmain or any fund promoter says.
If you want someone to check the real numbers behind your favorite Wall Street stories and you don't mind the occasional "hair on the balance sheet," like and subscribe now. And now I want you to express your opinion in the comments. If you believe that we are truly in 1997 and that there are still years of growth ahead, what specific number on the company's balance sheet will definitively tell you that we have already entered 1999? Think about it, and see you in the next video.
Comments 0