#87 Dot-Com Bubble vs. the AI Boom
Lessons from the post-dot-com decade: how to protect and position your portfolio today
The late-1990s dot-com bubble has become a classic example of exuberant markets driven by revolutionary technology – and of the painful crash that followed. Today, many draw comparisons between that era and the current excitement around artificial intelligence. Tech stocks have surged on AI optimism, raising the question: are we in an “AI bubble,” and if so, what happens when it bursts? Today we start exploring the similarities and differences between the dot-com frenzy and today’s AI boom, and – most importantly – examine what happened when the dot-com bubble burst. Which companies or sectors proved to be safe havens and why? Finally, we distill ideas for positioning a portfolio (over the next 3–5 years or longer) to prepare for a possible bubble burst (if we really are in one).
Welcome to Edelweiss Capital Research! If you are new here, join us to receive investment analyses, economic pills, and investing frameworks by subscribing below:
🚀 New Project
For those who don’t know yet, let me remind you that together with Leandro @Invesquotes (Best Anchor Stock), we’ve launched a new podcast.🎙️ New episodes every Tuesday at 18h in all platforms. For now, in Spanish only.
Subscribe/follow if you like it and share with anyone who might be interested.
1. The Dot-Com Bubble explained in 1min
The dot-com bubble of the late 1990s was marked by a surge of internet-related companies with sky-high valuations. From 1995 to early 2000, the Nasdaq Composite index quadrupled in value, and the S&P 500 roughly doubled, fueled by optimism about how the World Wide Web would transform business. Investors poured money into any company with a “.com” in its name, often with little regard for traditional metrics like earnings.
This euphoria hit its peak in March 2000. Soon after, reality caught up – growth expectations proved too optimistic, many unprofitable companies ran out of cash, and interest rates had been rising. The bubble burst spectacularly. Over the subsequent two years, the Nasdaq plunged nearly 80% from its peak, while the S&P 500 fell almost 50%. More than $5 trillion of market value evaporated, and scores of dot-com darlings went bankrupt. Iconic failures like Pets.com and Webvan became synonymous with the crash – Pets.com, for example, burned through capital with a flawed business model (e.g. costly shipping of pet food) and saw its stock collapse to cents per share before liquidating in late 2000. Webvan, an online grocery pioneer once valued over $1 billion, shut down in 2001 after expanding too aggressively and losing hundreds of millions,
Yet, amid the wreckage, a few companies survived – and some eventually thrived. No need to explain the Amazon.com is case. Priceline.com (now Booking Holdings) similarly survived. Even Yahoo!, which saw its stock skyrocket and then crash, did not disappear (though it lost dominance to Google in subsequent years). Established tech giants like Microsoft, Intel, and Oracle suffered severe stock declines but had solid earnings and cash reserves that helped them endure and eventually recover.
2. Parallels Between the AI Boom and the Dot-Com Bubble
Is history rhyming today? The recent AI-driven stock surge bears some undeniable similarities to the dot-com era. Once again, we see excitement over a “revolutionary technology” driving remarkable gains in a handful of tech stocks. In the late ’90s, it was the internet; now “AI feels like the new internet”, a transformative innovation expected to upend industries from healthcare to finance to entertainment. This narrative is extremely powerful, and capital is rushing in accordingly.
A small group of giant tech companies is leading the market, much as a few big winners led the late 1990s rally. Back then, Wall Street nicknamed the leaders the “Four Horsemen” – Cisco, Intel, Dell, and Microsoft – which symbolized the tech boom. Today’s market is similarly top-heavy: the “Magnificent Seven” of Apple, Microsoft, Alphabet (Google), Amazon, Meta (Facebook), Tesla, and Nvidia comprise over 30% of the S&P 500’s value. This narrow leadership means the market’s fate is heavily tied to a few momentum stocks – a risky scenario also seen in 1999. Back then, names like Cisco and Intel were trading at astronomical valuations and accounted for outsized portions of index performance; today, AI-chip maker Nvidia has played a similar role, skyrocketing as the poster child of AI. Notably, Nvidia’s share price climbed roughly 4,300% in the five years leading up to 2024 – a dizzying ascent that “stirred memories” of Cisco’s ~4,500% gain in the five years before its 2000 peak. Such exponential rises are classic signs of investor exuberance. To be fair, Nvidia’s fundamentals look stronger than Cisco’s ever did: revenues are up 7× in three years, versus Cisco’s 4× in the years preceding the crash. But semiconductors are an inherently cyclical business, and peaks always look most stable right before they turn.
Moreover, and most imporatnly, the source of the bubble is not within the big techs. the real risk lies here: we are again seeing investors pour money into anything AI-related, reminiscent of the dot-com craze. In the late ’90s, companies famously added “.com” to their names to attract investors, and many IPOs of firms with scant profits were oversubscribed. Today, simply slapping a ChatGPT partnership onto a company’s branding can trigger a “feeding frenzy” among investor.
The most explosive corner of the AI boom isn’t models or chips, but the so-called “Neoclouds.”
These are companies promising to rent GPU computing power “as a service” to hyperscalers and AI startups.
Growth numbers look spectacular:
In just two years, this once-niche layer has become the fastest-growing segment of the digital ecosystem. Analysts now project $65 billion in revenues by 2030, compounding at 30%+ annually — faster than the traditional cloud giants. But how much of this growth is real, and how much lives in Excel?
Take Iris Energy ($IREN), a retail-investor favorite. Roughly 90% of its revenue still comes from Bitcoin mining, yet some brokers already value it as the “next Nvidia of cloud.” The business model is extremely capital-intensive: billions in GPU purchases (+$2.9 B in PP&E expected between 2025–2027), paired with accounting assumptions that stretch credibility — a nine-year useful life for chips, and 80% data-center margins in a sector where 50% is exceptional. If those assets depreciate over two or three years, as technology cycles suggest, the economics collapse quickly.
This speculative fervor — when traditional valuation metrics lose relevance and the future is drawn in fluorescent ink — rhymes perfectly with the late 1990s, when many “.com” companies with no earnings were valued on eyeballs and clicks. The technology may indeed be transformative; that doesn’t mean any price makes sense.
Importantly, the belief that a transformative technology will revolutionize everything is a common thread in both eras. In 1999, investors genuinely thought the internet would reshape every corner of the economy (they weren’t wrong, but the timing and winners were unclear). Today, many believe AI will “reshape every corner of the economy,” boosting productivity and spawning new industries. In both cases, this optimism is rooted in truth – the internet did change the world, and AI likely will as well. However, history shows that markets tend to overestimate the speed and easy profits of such transformation in the short term. During bubbles, investors pay almost any price for companies perceived to be on the cutting edge.
“When you have a bubble… it’s rooted in some true, positive fundamental development… that creates enthusiasm for people to pay any price”.
It is not the goal of this post to argue whether we are or aren’t in a bubble — but several things smell undeniably fishy. Valuations have drifted far from economic reality, with private AI labs valued at hundreds of billions despite burning cash and public “AI infrastructure” stocks trading at 50–100× sales on projected numbers that may never materialize. Capital is flooding in “just in case”, as venture funds, corporates, and retail investors all scramble to secure exposure before missing out. Yet for all the noise, there are still few genuinely profitable use cases today — most of the current demand comes from AI companies buying from each other rather than from paying end customers. The sector also shows signs of perpetual disruptive churn, where each six-month innovation renders the previous one obsolete, eroding the odds of durable returns. Finally, creative financing is spreading fast — suppliers funding their own clients, circular contracts counted twice as revenue, warrants exchanged for purchases — all classic symptoms of late-cycle exuberance. None of these things prove we’re in a bubble, but together they make the air feel thin.
3. But this time MIGHT be different
As the CEO of Anthropic recently argued, the lack of cash flow in leading AI labs doesn’t necessarily mean the economics are broken — only that the technology is still in its investment phase. His logic (not entirely convincing, but directionally fair) is that as long as each new model meaningfully improves on the previous one and adds value to users, current losses are a reflection of ongoing reinvestment rather than structural unprofitability. The problem, of course, comes if that improvement curve flattens or the technology becomes commoditized: when differentiation fades, pricing power disappears, and the “AI infrastructure” built on high-margin assumptions could hit a wall. Still, there’s no doubt that AI is already creating real value — the open question is who captures it. If most of that value accrues to consumers through cheaper or free tools, then society will benefit enormously, but shareholders may not. In that sense, the analogy with the dot-com boom cuts both ways: the internet transformed the world, yet it took years before investors could reliably make money from it.
Another key distinction is that today’s AI wave is being financed by highly profitable incumbents — Microsoft, Amazon, Google — whose core businesses throw off enormous cash flows and can subsidize years of experimentation. That’s a very different setup from 1999, when loss-making startups depended on a constant inflow of speculative capital to survive. The weak link, however, remains at the core of the value chain: the model developers themselves still burn billions and haven’t demonstrated a clear path to self-sustaining profits. As long as Big Tech is willing to foot the bill, the ecosystem can keep expanding. But if enthusiasm fades or capital tightens, the system’s dependence on corporate sponsorship could quickly become its own vulnerability.
Many also argue that debt hasn’t yet entered the picture the way it did in the 2000 bubble, when telecom companies levered up massively to lay fiber and build network infrastructure. I’m not so sure. It’s true that the Big Tech players haven’t yet leaned heavily on debt — though Meta is starting to — but a quick look down the AI value chain tells a different story. The trillion-dollar infrastructure being built to power AI — data centers, chips, energy — is already being financed with debt backed by long-term contracts that assume exponential demand. In other words, the leverage may not sit on the balance sheets of Microsoft or Google, but it’s very much there — embedded in the suppliers and intermediaries building the ecosystem on borrowed money. That kind of hidden leverage rarely stays hidden forever.
4. How to prepare your portfolio for a potential AI bubble burst
Let’s assume you’re the “loser” — the one who refuses to sing along with the AI siren song (at least not at these prices). You’re fine being the person with “mediocre” returns while friends boast triple-digit annualized gains… because you care about surviving the cycle and compounding after the music stops. How do you position for the next few years?
After the 2000 peak, leadership flipped. Speculative, high-multiple tech was crushed; stable cash generators and cheaper “old economy” names took over. Defensive sectors (consumer staples, healthcare, utilities) held up or even outperformed. “Value” beat “growth” for years. Commodities and EM had a strong run as capital rotated away from U.S. tech into resources and catch-up growth. The lesson is simple: when a dominant growth narrative unwinds, bread-and-butter cash flows and lower expectations become scarce—and get re-rated.
Where to look if the music stops
Staples.
Many global consumer staples have been dead money for years. Nestlé, PepsiCo, Coca-Cola, Unilever — all high-quality franchises with global scale, distribution moats, and predictable demand — have underperformed in relative and absolute terms. These are businesses that compound quietly in the background: they sell affordable indulgences, reinvest in brands, and pass on inflation with a lag. Yet they’ve been de-rated as investors chase tech, AI, and “growthier” stories. Valuations are now near decade lows relative to the market (in some cases due to terrible management strategies… it’s no news, but these are day 2 or 3 companies), despite pricing power and steady cash generation. In every major cycle, defensives like these have done their job: boring when euphoria reigns, but resilient when narratives break. They rarely make headlines, but they preserve purchasing power and give you liquidity to buy risk assets when others are forced to sell. This kind of “anti-bubble” exposure tends to outperform not because margins expand, but because multiples stop compressing while everything else collapses.
Retail defensives.
A subset of defensives — think Walmart, Costco — have already re-rated, reflecting their proven execution and consumer relevance. They are not cheap, but they illustrate an important point: defensiveness doesn’t mean low growth, it means predictable growth. These 2 examples are trading at ctazy valuations, but there might be cases out there with higher growth prospects and more reasonable valuations. Happy hunting.
Healthcare.
Healthcare is quietly back to being one of the cheapest “expensive” sectors. After the COVID boom and the subsequent unwind in diagnostics, vaccines, and medtech, valuations have normalized. Large-cap pharma trades at 13–15× earnings; quality medtech and tools companies at 18–22× — levels that, historically, have offered mid-teens returns once sentiment stabilizes. The long-term tailwinds — aging populations, chronic diseases, innovation in biologics and oncology — haven’t disappeared, but investor attention has. Many pipelines look healthier today than in 2019, yet the market prices them as if growth had ended. This is the classic setup where healthcare acts as ballast: non-discretionary demand, patent-protected pricing power, and low economic sensitivity. It also tends to benefit from rotation when capital leaves speculative sectors. The internet bubble burst in 2000; healthcare rose nearly 30% over the following two years while the S&P 500 fell ~16%. That’s not coincidence — it’s the structural defensiveness of cash flows tied to human needs, not advertising clicks or GPU cycles.
Commodities and Emerging Markets.
The mirror image of a tech unwind is often a resource or EM rally. After 2000, global capital left overbuilt internet infrastructure and poured into physical assets — energy, metals, agriculture — fueling the so-called commodity supercycle. Emerging markets, particularly resource exporters, benefited enormously: Brazil, Russia, and parts of Asia delivered triple-digit returns through the mid-2000s. Fast-forward to today: EM assets have been left for dead. China’s stagnation, weak governance, and a (former) strong dollar have crushed sentiment; LATAM and EMEA have seen outflows for a decade. Yet if AI demand drives real-world infrastructure (data centers, energy, materials), some of that capital could rotate back. Commodity leverage is where cyclical operating leverage hides. The challenge is selectivity: governance, FX risk, and political noise can erase gains. Still, some high-quality EM companies trade at single-digit earnings multiples. That’s a powerful contrast to U.S. tech at 30–40×.
A final note
I’m not suggesting that anyone should rush to fill their portfolio with energy producers, consumer staples, healthcare stocks, or emerging market indices. This is not a shopping list, and certainly not a model portfolio. It’s just food for thought — a way to think about positioning before the cycle turns, not after. Every market regime rewards a different kind of investor: the trick is not to predict the next one, but to make sure you can survive all of them.
The goal isn’t to abandon innovation or short-term momentum entirely. AI will reshape industries, just as the internet did. The point is to recognize that great technologies and great investments are not always the same thing, especially at euphoric valuations. In periods like this, the best opportunities often hide in the places investors have stopped looking — businesses that are dull, under-owned, or temporarily out of fashion.
If you step back, this entire debate — AI versus “old economy,” growth versus value — misses the larger question: how durable is the cash flow that underpins your portfolio? In 2000, it was the boring companies — the ones nobody bragged about at dinner parties — that quietly compounded through the wreckage.
So, don’t take this as a call to go defensive or to bet against AI. Take it as a reminder that capital preservation is a competitive advantage, and optionality is built through prudence. In every bubble, there’s a moment when sitting still feels foolish — some times for years, before it looks wise. Your job isn’t to call the top; it’s to make sure that, when the music stops, you still own businesses that make money, not stories that need it.
So, what do you think? Are we in the middle of a bubble—or just an early chapter of a genuine technological revolution? And if the music stopped tomorrow, which sectors, industries, or companies do you believe would hold up best?
If you enjoyed this piece, please give it a like and share!
Thanks for reading Edelweiss Capital Research! Subscribe for free to receive new posts and support our work.
If you want to stay in touch with more frequent economic/investing-related content, give us a follow on Twitter @Edelweiss_Cap. We are happy to receive suggestions on how we can improve our work.
Great comparison between the two events and wise advise for those who care to follow.
Very useful 😃