Is This Another 2000? The Profits Holding Up This Market, and What Could Break Them

Research note: the ten Bubble Tracker gauges against their March 2000 values; profits, margins and capital spending for large US companies from SEC filings through September 25, 2026; S&P 500 earnings from Robert Shiller's data. Computed September 26, 2026. How we checked it · How we use AI.

• The resemblance is real. 8 of 10 Bubble Tracker gauges are at extremes; at the March 2000 peak it was 10.
• Growth is broad, but the dollars are concentrated. The seven technology leaders supplied 67% of the year's earnings increase, and about half of their own increase came from gains on investments rather than operations.
• The valuation rests on the profit share. If profits fell from 13.2% of GDP to about 10.5%, the market would cost as much per dollar of profit as in March 2000.
• What could go wrong: higher yields, slower infrastructure orders, input costs outrunning selling prices, and the build-out's depreciation. Higher yields need no fall in profits at all.

How close is this to 2000?

Close, on the measures of how much wealth sits in stocks. Households hold 48.2% of their financial assets in equities, against 37.4% in March 2000. Market value is 288% of GDP, against 164%. Five of the ten gauges are past their 2000 readings.

The historical charts show the scale of it. Market value against GDP and household equity allocation have passed their 2000 readings by a wide margin; cash relative to equity value has slipped below its dot-com reading. These measures overlap: rising share prices can push all three toward extremes at once.

Here is every gauge the tracker scores, set against its March 2000 reading. Six of the ten put the market's value or SPY's price in the ratio, so eight extremes are several views of one condition rather than eight independent confirmations.

So if the market is this stretched, why does it keep going up? My answer is profits. On ten years of smoothed earnings, today is nearly as expensive as the dot-com peak: the Shiller CAPE sits only 4% below its March 2000 reading. On the profits companies are earning right now, it is not. The market costs 21.8 times current corporate profits, against 27.4 at the top in 2000.

That gap is the whole story. In 1999 money kept pouring in after economy-wide profits had already peaked, in Q4 1997. Today the money is even more committed. But it is buying a profit stream that is a record share of the economy. What holds this market up is whether that stream keeps growing, so that is what the rest of this piece tests.

Are profits still growing?

Across 215 large US companies, earnings over the latest four filed quarters rose 37% on a year earlier, on revenue growth of 13%. Operating profit, which leaves out gains on investments, rose 26%.

The gap between those two numbers sits almost entirely in the seven technology leaders: Nvidia, Microsoft, Apple, Alphabet, Meta, Amazon and Broadcom. Their net income rose 67%; their operating income rose 34%. Much of the difference is paper. Alphabet booked a $99.0 billion gain on equity securities in its second quarter alone, mostly unrealized, and Amazon $53.4 billion of non-operating income, primarily from its investments in Anthropic. Those are marks set by private funding rounds, and they can reverse.

Outside the seven, the two measures agree. The other companies' earnings rose 20% and their operating profit 19%, and 74% of them reported higher earnings. The recovery is broad by company count and concentrated by dollars, and the broad part is made of operating profit.

What did the 2000 top actually look like?

The dot-com market had real earnings too. S&P 500 earnings rose 72% from 1995 to their peak. The index set its daily closing high in March 2000; its monthly-average price peaked in August. Trailing earnings reached their high in the third quarter. At the monthly-average price peak the index traded at 28.0 times trailing earnings; in June 2026 it traded at 25.2. On reported earnings, today is cheaper than 2000, but not by much.

Economy-wide profits were already falling, while S&P earnings kept climbing. The Federal Reserve's July 2001 report records technology earnings falling nearly half in the two quarters after their third-quarter 2000 peak.

Earnings subsequently fell 54%, and tech investment fell 16% from its peak. That is the uncomfortable precedent for my view: strong reported earnings did not prevent prices from turning.

The Bubble Tracker first counted 70% of its gauges at extremes in January 1997. SPY rose 96% from there to the March 2000 peak, then fell 49%. Months in that zone were followed by an average three-year return of −1.8%, against +30.6% when no gauge was extreme, but every completed month comes from that one episode. It says what happened once.

How much of the valuation rests on the profit share?

Corporate profits now equal 13.2% of GDP. Hold market value and GDP fixed, and a fall to 10.5% would put the market on the same multiple of profits as in March 2000. The lower multiple today rests on that high profit share. It first reached 10.5% in Q4 2005, and only 45 of 303 quarters since 1947 have been at or above it.

Profits do not have to fall for prices to. The multiple can shrink on its own, and the grid below shows how much earnings growth it takes to absorb that. Earnings up 10% with the multiple down 15% leaves prices −6.5%.

Higher Treasury yields raise the hurdle for equities. Continued earnings growth could clear it, but today's profits alone do not guarantee that investors will keep paying the same multiple.

Whose profits are they?

The broad earnings growth describes the past year. Over the longer period, much of the increase in profitability belongs to the technology leaders. For the 96 companies present in both 2011 and today, net margins rose from 9.4% to 15.1%.

Companies that became more profitable also grew into a larger share of the basket. Apple, Amazon, Microsoft and Nvidia account for more than the entire margin increase in that matched group, with the remaining companies offsetting part of their contribution.

The others benefited from lower taxes. In the subset with complete tax and income data, lower taxes account for about 64% of the net-margin increase since 2011. That fits Michael Smolyansky's research on the contribution of lower interest and tax expense to corporate profit growth.

More recently, the other companies' net margin has barely moved: 9.8% in 2018, against 10.0% now. Their recent earnings growth is encouraging, but sustained margin expansion outside the leaders would strengthen the bull case further.

What does the AI build-out have to earn?

Microsoft, Alphabet, Meta and Amazon spent $511B on property and equipment over the last four quarters, much of it on data centers. Nvidia and Broadcom sell into that spending. Over the past year the four spenders' capex rose $219B and the two suppliers' revenue rose $167B.

The two chip suppliers produced 54% of the seven's operating profit growth over the year (33% of their net income growth, which also carries Alphabet's and Amazon's investment gains). A slowdown in infrastructure orders would put part of that growth at risk. One company's capital expenditure supports another's current revenue, although these totals do not trace individual purchases between the companies.

For the four spenders, the build-out creates a depreciation burden that could grow over the coming years. The table illustrates the effect of different asset lives on margins. It applies one life to mixed assets, includes spending beyond AI, and assumes no new capex; actual charges will depend on what was built and when it enters service.

Under the five-year illustration, depreciation would be $86B above the latest annual charge. Absorbing it would require revenue growth or savings elsewhere. The four still generated $150B of combined free cash flow over the last four quarters. I would watch whether that cash generation holds as investment continues.

What could break it?

The risks reach earnings through investment orders, customers' budgets and costs. Higher yields can reach share prices before any of those weaken. Each is taken in turn below, with the evidence for it, and then the readings that point the other way.

1. Yields

The ten-year Treasury yields 5.18%, and after expected inflation 2.85%, the most since November 2008. Our filings basket earns 3.6% of its market value a year. In 2011 it earned 6.8% while the ten-year paid 3.5%, a cushion of 3.3 points.

That cushion first disappeared in Q2 2023, and the basket's earnings yield has been below the ten-year in 14 of the 63 quarters since 2011. A bond paying more than stocks earn does not set a fair value, but it raises the bar that earnings growth has to clear. This is the risk that can lower the multiple before profits weaken at all.

2. Infrastructure orders

The build-out has a supply chain inside the index. Over the four quarters to Q3 2018, Microsoft, Alphabet, Meta and Amazon spent $56B on property and equipment, and Nvidia and Broadcom took in $32B of revenue. Today the figures are $511B and $392B.

The two lines have risen together. The suppliers produced 54% of the seven's operating profit growth over the past year, so a pause in data-center spending would reach the index twice: as lower capex at the spenders, and as lower revenue at the suppliers. These totals do not trace individual purchases, and both suppliers sell to many other buyers.

3. Input costs

Prices of processed goods that producers buy are up 12.3% on a year earlier; prices producers charge for final goods and services are up 5.4%. Since 2010, that gap has been wider only in 2021 and 2022.

These are price indexes rather than margins, and the two cover different baskets. But a company buying at the first rate and selling at the second is being squeezed, and diesel at a record $6.53 a gallon is in nearly every freight bill. This pressure falls mostly on the companies outside the seven, whose margins are the thinner ones.

4. The leaders' customers

Advertising and cloud budgets are paid out of other companies' profits. The last time those profits fell, the platforms felt it. The other companies' earnings were down 11% on a year earlier in Q2 2023; the four ad and cloud platforms' revenue growth slowed from 33% in Q3 2021 to 7% in Q2 2023.

Today the customers are growing again, with earnings +20% and platform revenue +19%. If input costs turn into an earnings squeeze outside the seven, this is the route by which it reaches them. In 2023 the platforms' slowest revenue growth came in the same quarter as their customers' worst earnings.

What points the other way

The first reading below says companies have room to absorb cost pressure before margins give. The other three show a consumer in better shape than the saving rate alone suggests, with one exception: mood.

These readings, alongside growth across most of the companies we track, are why I still favor continued earnings expansion. The risks have plausible routes into profits; the current evidence has yet to establish a broad reversal.

Three ways this ends

Earnings carry prices higher. Earnings keep growing and valuations hold sufficiently steady for prices to advance. The bubble gauges could remain extreme through that rise. This is my base case.

Profits grow, but the market pays less for them. Earnings keep rising while higher yields or disappointment about AI returns shrink the multiple. The grid above shows how easily that produces a falling market with rising profits. It is the ending I think most investors underrate.

Margins crack. Depreciation outpaces revenue growth, infrastructure orders slow, or input costs squeeze other companies. A falling profit share erodes the earnings support for today's valuation.

What I am watching

In 2000, reported S&P earnings weakened after prices had turned. These observations could reveal deterioration before it becomes obvious in economy-wide profits, but none promises advance warning of a top. The thresholds are my judgement.

I remain bullish because earnings growth extends well beyond the seven leaders. I would reconsider if that breadth weakened, the largest spenders' cash generation deteriorated, or higher yields began to overwhelm earnings growth. A high bubble score alone would not change my view.

How we checked it

• The 2000 readings. Each gauge's March 2000 value is today's revised series filtered by publication date, using estimated lags where exact dates are unavailable. Later observations are excluded; subsequent revisions to earlier observations remain. These are not the original vintages investors saw.
• The gauges overlap. Six of the ten share the market's value or SPY's price as an input. Eight extremes are several views of one condition, not eight independent confirmations.
• Profit growth. The same companies' latest four filed quarters are compared with the four available one or three years earlier. Three-year changes are cumulative. The sample consists of large companies covered today and omits companies that disappeared. Earnings are net income, which includes gains and losses on investments; operating income leaves them out, and both are shown because in 2026 the two diverge sharply at Alphabet and Amazon. BEA economy-wide profits, which exclude gains on holdings, and this listed-company basket cover different populations; one cannot fully explain the other.
• The margin reconciliation. The 2011 to 2026 change in the basket margin is split into companies entering the basket, margin changes within the 96 matched companies, and changes in their revenue weights, and checked within and across sectors (SEC industry codes mapped to sectors). The tax split uses the 42 other companies that report operating income, pretax income and tax on the same four quarters; it measures a contribution on that subset and does not settle the cause.
• The depreciation schedule. Straight-line depreciation of each quarter's filed capex over four, five or six years, with no new spending, compared with the latest four quarters of reported depreciation. Useful lives, construction timing and asset mix all differ in practice.
• The 1995–2003 record. S&P 500 monthly-average prices and trailing reported earnings from Robert Shiller; BEA profits and tech investment via FRED. Shiller interpolates quarterly earnings to monthly values. The historical return comparison has one completed bubble episode, with overlapping months, so it does not establish a repeatable timing rule.
• Frozen and checkable. The page renders a snapshot taken at publication with fingerprints of every input file. The freeze script's check mode recomputes every part of the study, including the margin, build-out and macro findings, and names any that drift.