Since the financial crisis, US technology has dominated global investing. For fifteen years, skeptics warning of stretched valuations were repeatedly proven wrong. These companies combined visionary founders, monopoly-like positions, enviable margins, and massive growth, and they rode the era of zero interest rates to become the world’s largest firms. Capital, starved of alternatives, crowded into these “secular compounders,” and for more than a decade they outperformed everything else. The narrative became self-reinforcing: technology was not just a sector, it was the market.
But today, that trade has reached an inflection point. The boom now rests almost entirely on AI, and investors have turned it into a one-way bet. Fundamentals are flashing red: growth is slowing, competition is intensifying, and capital intensity is exploding to levels that recall some of the most catastrophic investment busts in history.
The parallels to the dotcom bubble are striking. In 2000, Microsoft and Cisco lost a third of their market value within a week, Amazon dropped nearly 80 percent in a single year, and the Nasdaq erased trillions in months. Strong companies didn’t protect investors from bad prices. Analysts at the time were slow to acknowledge reality – even late in 2000, Wall Street still forecast nearly 9 percent earnings growth for the following year. Investors holding “quality” at inflated multiples still faced a decade or more of dead money: Microsoft took fifteen years to regain its 2000 peak.
The same belief structure that powered the late 1990s is powering markets now. Then, American exceptionalism was the anchor: GDP grew 5 percent, deregulation juiced productivity, emerging markets collapsed, Europe and Japan stagnated, and the US was the sole engine of prosperity. Today, capital again pours into US equities under the “TINA” logic – there is no alternative. The second anchor was technological revolution. In 1999, IBM declared “The revolution has arrived.” Today, AI is framed in the same epochal language. The promise is not just new tools, but an economy transformed.
But the evidence shows the quality, valuation, and breadth of today’s bull market are not better than 2000 – in fact, they are worse.
Revenue growth at the tech giants has structurally decelerated. Microsoft, Alphabet, Amazon, and Meta already dominate their industries, capturing roughly 60 percent of sector profits. That concentration limits their ability to grow faster than the economy. Revenue growth, once 20–30 percent annually, is now trending toward single digits within five years – comparable to utilities and insurers, but with much greater cyclicality and risk.
Meanwhile, the supposed endless growth of digital advertising, which funds much of this investment, is fading. Advertising is already 70 percent digital, and its growth is expected to slow into the low single digits. Historically, total ad spend rarely exceeds 2 percent of GDP, meaning the pie has natural limits even if AI drives efficiency.
Competition, too, is nothing like the monopolies of the 2010s. Google no longer has the ad market to itself: ByteDance has overtaken Meta in global social media revenue, Amazon and Netflix are eating into the ad pie, and even Walmart and Uber are expanding aggressively. The cloud, once a three-player oligopoly of Amazon, Microsoft, and Google, now faces Oracle and other aggressive entrants. AWS margins have already compressed by seven percentage points. In China, datacenter overcapacity is so severe that utilization rates hover between 20 and 30 percent – eerily reminiscent of the fiber glut of the early 2000s.
That telecom glut is the most dangerous parallel. AT&T, Verizon, and WorldCom spent hundreds of billions laying fiber in the late 1990s, convinced by forecasts of exponential demand. Capital expenditures exploded, often to 60, 80, or even 100 percent of EBITDA, swallowing all cash flow. Debt piled up. When demand lagged, the sector imploded. WorldCom collapsed into bankruptcy, AT&T was forced into restructuring, and Verizon slashed dividends as its balance sheet buckled. Investors learned the hard way that building the infrastructure of the future can still destroy shareholder value if done at the wrong price and pace.
Today’s big tech firms are on the same trajectory, but at much greater scale. Microsoft, Alphabet, Amazon, Meta, and Nvidia are pouring unprecedented sums into AI datacenters, chips, and infrastructure. Combined CapEx for these five companies is projected to surpass $200 billion annually within two years. Microsoft and Alphabet are expected to spend more than 80 percent of EBITDA on CapEx, while Meta’s AI push already consumes around 120 percent of EBITDA. Even Apple, historically the most disciplined of the giants, is being pressured to join the arms race. Nvidia itself, the linchpin of the AI narrative, requires continual reinvestment as chip cycles shorten and each generation of GPUs rapidly cannibalizes the last. The more they build, the more they must replace. Depreciation charges are already soaring, and free cash flow margins are deteriorating across the sector.
This is not the profile of a “capital-light, cash-gushing” industry. It is the profile of an arms race, where enormous sums are spent in the hope that future monetization will arrive quickly enough to justify the investment. But just as with fiber in 2000, utilization is low, monetization is lagging, and the mismatch between spending and realized demand is widening. AI-related revenues are still under $20 billion globally, even as CapEx is projected to surpass $7 trillion by 2030. By comparison, internet revenues after the dotcom buildout reached hundreds of billions within a few years. Today’s gap between investment and realized income is even more extreme than the telecom bust.
Valuations only amplify the risk. The share of S&P 500 companies trading at more than ten times revenues is around 35 percent today, compared with 25 percent in 2000. Nvidia trades at more than thirty times revenues, while its market capitalization of $4.5 trillion represents about 15 percent of US GDP – three times the relative scale of Cisco at its 2000 peak. Cisco, then the essential infrastructure provider of the internet, collapsed when demand forecasts proved inflated. Nvidia now holds the same narrative role, but at a systemic scale.
The broader market is no safer. The S&P 500 trades at 23 times earnings, close to the 25 times of 2000, but with slower growth and fiscal deficits rather than a government surplus. In 1999, the US budget surplus was so large that Washington seriously considered ending Treasury issuance. Today, deficits are chronic and growing. Investors bullishly call this 1995 – the midpoint of the internet boom. But valuations, speculation, and fundamentals argue it is already 1999.
Even Microsoft illustrates the danger. In 2000, it traded at 60 times earnings while growing revenues at 35 percent. Today it trades at 35 times earnings with growth in the low teens. The multiple is lower, but the growth is far slower, leaving no margin of safety. And just as Microsoft delivered fifteen years of flat returns after 2000, investors today risk a lost decade even if businesses continue to grow.
Other bubbles show the same anatomy. The shale boom of the 2010s consumed enormous capital under the assumption of unending demand. When oil prices fell, overcapacity was revealed, margins collapsed, and investors were left with stranded assets and debt. The Nifty Fifty of the 1970s, the Japanese equity bubble of the 1980s, and the US railroads of the 19th century all looked like “this time is different” stories, with dominant companies at the center of revolutionary growth. Each ended with violent repricing once valuations, fundamentals, and capital spending collided.
The current AI mania combines the worst features of all of them. Like the dotcom era, valuations are detached from growth. Like telecom, CapEx is consuming EBITDA at unsustainable levels. Like shale, the buildout risks creating stranded assets. Like Japan, the concentration of wealth in a handful of names creates systemic fragility. And unlike all of them, today’s excesses are occurring in the world’s largest, most systemically critical firms – companies that dominate indices, ETFs, retirement accounts, and sovereign wealth portfolios worldwide.
The lesson is clear: essential does not mean investable at any price. AT&T’s networks were essential. Verizon’s lines carried the world’s traffic. Cisco’s routers powered the internet. WorldCom’s pipes were critical. All collapsed under the weight of overcapacity and mispriced expectations. The AI buildout, already consuming nearly all of big tech’s cash flow, risks replaying that story at a scale far larger than any bubble before it.