Nvidia generated $99.2 billion in net income over the trailing twelve months. Not revenue. Profit. That single number demolishes half the AI bubble commentary published in the last six months, yet almost nobody writing "Is the AI Bubble About to Pop?" bothers to mention it.
The number that matters: Nvidia's forward P/E ratio sits at roughly 24x today, against Wall Street consensus for 57% earnings growth this year. Its PEG ratio is 0.69. For context, the company's five-year average trailing P/E is north of 70. The stock is, by its own historical standards, cheaper now than at almost any point in the current AI cycle. That's a strange fact pattern for a bubble.
The Dot-Com Analogy Falls Apart Under Scrutiny
Every bubble article reaches for the same playbook: compare AI to the dot-com era, cite price-to-sales ratios above 30, and warn that history repeats. There's a kernel of truth in the comparison. Palantir's P/S ratio near 112 is genuinely absurd. Some neocloud providers like CoreWeave carry existential risk if demand cools. Nobody should pretend every AI-adjacent name is a sound investment.
But the aggregate picture is entirely different from 1999. Fidelity's research team noted in early 2026 that today's AI spending is "overwhelmingly being funded with cash, not debt, by companies that regularly generate high volumes of free cash flow." Goldman Sachs data shows AI capex currently equates to roughly 0.8% of GDP, compared to peaks of 1.5% or greater during previous technology booms over the past 150 years. We are not even close to historical excess by that measure.
The hyperscalers are spending aggressively, yes. Alphabet, Microsoft, Meta, and Amazon are on track to spend nearly $700 billion combined on capex in 2026. That figure stuns at first glance. But Fidelity's analysts also flagged something crucial: as of early 2026, they are seeing none of the typical warning signs of a true bubble, including no shrinking free cash flows driven by aggressive AI spending, no deteriorating leverage ratios, and no compression of multiples from power or computing bottlenecks.
Earnings Don't Lie. Narratives Do.
The S&P 500 is on track for its third consecutive year of double-digit earnings growth. FactSet projects 14.4% earnings growth for calendar year 2026, well above the 10-year average of 8.6%. Revenue growth of 7.2% is similarly above trend. The estimated net profit margin for the index is 13.9%, which would be the highest annual margin on record.
Critically, the earnings story is broadening. The Magnificent Seven are expected to grow earnings by 22.7% in 2026, roughly in line with 2025's pace. But the other 493 companies in the S&P 500 are projected to deliver 12.5% earnings growth, up from 9.4% in 2025. This is the opposite of what happens in a bubble, where gains narrow to fewer and fewer names until the music stops.
Goldman Sachs projects S&P 500 EPS of $305 in 2026, with revenue growth of 7% and 70 basis points of margin expansion. Their base case embeds a forward P/E multiple of 22x at year-end. Is that elevated? Historically, yes. Is it justified by the earnings trajectory? The data says it is, particularly when earnings growth is accelerating rather than decelerating.
Where the Real Risk Lives
This isn't to say the market is without vulnerabilities. The S&P 500's forward P/E of 21.5x is above both its 5-year average of 20.0 and its 10-year average of 18.8. Concentration risk is real; the five largest companies account for 30% of the S&P 500 and 20% of the MSCI World Index, the greatest concentration in half a century. An MIT Media Lab report from August 2025 found that 95% of organizations investing in generative AI are seeing zero return so far. That's a meaningful gap between spending and outcomes.
Ray makes a fair point when he flags the free cash flow trajectory at the hyperscalers. Amazon may face negative free cash flow of $17 to $28 billion in 2026, and Alphabet's free cash flow could plummet nearly 90% to roughly $8 billion from $73 billion in 2025. Those are not trivial numbers. If you stare only at the cash flow statement, you can build a bearish case.
But zoom out to the earnings trend and the picture changes. Google Cloud's backlog surged 55% sequentially and more than doubled year-over-year, reaching $240 billion. OpenAI hit approximately $20 billion in annual recurring revenue, tripling year-over-year. The demand signals are not speculative; they represent real customer commitments backed by contracts. DA Davidson analyst Gil Luria called the current investor skepticism around AI stocks "probably healthier than any previous cycle I've seen." Healthy skepticism is the opposite of bubble psychology.
The smart move here is not to flee AI exposure. It is to be selective. Strip the Magnificent Seven from the S&P 500 and the forward P/E drops to a more modest 20.3. Value stocks, small caps, and dividend payers have been systematically neglected as capital chased the AI trade. That creates opportunity on both sides: own the AI winners with real earnings power, and diversify into the parts of the market that have been left behind.
Bottom line for your portfolio: The AI bubble narrative sells clicks, not returns. A bubble requires speculative excess untethered from fundamentals. What we have instead is a capital investment cycle backed by the most profitable companies in history, growing earnings at 20%-plus annually, funding the buildout largely from cash. The market is telling you something. Most people aren't listening.
