Markets•
October 6, 2026•
7 min read

The Nasdaq Is at a Record. Can AI Earnings Carry the Weight?

AI demand is real. But with a narrow rally, Treasury yields above 5%, and earnings expectations already unusually high, demand is no longer enough.

By Nagham

The Nasdaq Composite closed at another record on Monday.

But underneath that headline, the market looked much less uniform.

Of the 101 securities tracked in the Nasdaq-100, 55 were below their 50-day moving averages. Among the 373 record closes for which breadth data were available since 2015, an average 71.5% of Nasdaq-100 securities were above their 50-day averages.

On Monday, only about 46% were.^[1]

That doesn’t predict a correction.

Breadth is a concentration signal, not a timing signal. A cap-weighted index can continue rising for months while relatively few companies do most of the work.

This market is increasingly dependent on a small group of mega-cap winners.

Nvidia gained 2.1% Monday and reached another record. Microsoft rose 1.5%, while Meta and Tesla gained roughly 2%. Meanwhile, the Magnificent Seven — Nvidia, Apple, Microsoft, Alphabet, Amazon, Meta and Tesla — reached a combined market capitalization of about $24.84 trillion at Monday’s close.^[2]^[3]

The concentration itself isn’t necessarily bearish.

It simply makes disappointment more expensive.

Then Look at the Bond Market

While technology stocks were making records, the 10-year Treasury yield closed Monday at 5.31% and pushed to roughly 5.35% Tuesday morning — a 24-year high.^[4]

Two markets. Two messages.

Nasdaq: record high

10-year Treasury: above 5.3%

Growth expectations remain extraordinary. So does the cost of capital.

This isn’t about saying a 5% Treasury is automatically better than owning stocks. It isn’t.

But higher long-term rates increase the discount rate applied to future corporate cash flows. The further those profits sit in the future, the more sensitive their present value becomes to rates.

There is another interesting comparison.

On October 1, the S&P 500 traded around 19.2 times forward earnings, implying a simple forward earnings yield of roughly 5.2%. That is now close to the yield available on the 10-year Treasury. The two figures are not apples-to-apples — corporate earnings can grow, while Treasuries have very different risks — but the comparison illustrates how much more demanding the valuation equation has become.^[5]

For readers watching TLT, it offers another window into that pressure.

TLT holds U.S. Treasuries with remaining maturities greater than 20 years, making it particularly sensitive to long-duration interest-rate risk. Persistent weakness in TLT alongside rising long yields tells us duration remains under pressure; stabilization alongside falling yields would tell us some of that pressure is easing.^[6]

AI Demand Isn’t the Question Anymore

Micron already gave us evidence that the demand is real.

Micron’s demand evidence

Customer commitments: $22B → $32B

Remaining performance obligations: $100B → $150B

Its customer commitments under long-term agreements rose to $32 billion from $22 billion in June, while remaining performance obligations increased to approximately $150 billion from $100 billion in the prior quarter.^[7]

That’s not simply AI enthusiasm.

Those are commitments.

But Micron’s numbers don’t tell us that every company selling into — or spending heavily on — AI will ultimately generate returns high enough to justify its valuation.

The Earnings Bar Has Already Been Raised

Q3 earnings setup

Expected S&P 500 EPS growth: +29.5% YoY

Estimate revisions during the quarter: +1.4%

Companies issuing positive EPS guidance: 62%

FactSet expects Q3 2026 S&P 500 earnings to grow 29.5% year over year.

More unusually, analysts raised Q3 earnings estimates by 1.4% during the quarter. Over the previous five years, estimates had typically fallen by an average 2.2% during a quarter.

Companies are optimistic too: 72 of 116 companies issuing Q3 EPS guidance — 62% — issued positive guidance, versus five- and ten-year averages of only 40% and 41%.^[8]

The market isn’t entering earnings with a lowered bar. The bar has already been raised.

AI companies now have to show not simply demand, but conversion:

Revenue. Margins. Cash flow. Guidance.

Then Comes Anthropic. Potentially.

Bloomberg reports that Anthropic is targeting an IPO as soon as mid-November, although the timing remains subject to change.^[9]

If it happens, the numbers investors will be asked to price are extraordinary.

According to Reuters’ review of Anthropic’s confidential S-1, 2025 revenue reached about $4.6 billion, twelve times the previous year’s level.

But the company also reported a $42 billion net loss.

That headline needs context.

Roughly $34 billion came from financing-related write-downs, leaving the underlying operating loss at a little more than $8 billion.^[10]

Still enormous.

But economically very different from losing $42 billion through operations.

An Anthropic IPO would not be a verdict on the entire AI sector. Its capital structure, compute requirements and business model make it unusual.

How much are you willing to pay for spectacular growth when producing that growth also requires spectacular amounts of capital?

So What Changes the Story?

I wouldn’t use one number as a magical line in the sand. But I would watch three.

Breadth

46% → 60%+

If more Nasdaq-100 securities reclaim their 50-day averages, the rally becomes less dependent on mega-cap leadership.

Rates

<5% / >5.5%

A move below 5% would ease some valuation pressure. A sustained move above roughly 5.5% would raise the hurdle further.

Earnings

Revisions

Continued upward revisions support the rally. A turn downward would matter more while breadth remains narrow.

Those aren’t market laws.

They’re watch lines.

Because this market doesn’t need AI demand to prove itself anymore.

Micron already helped with that.

The harder question is whether profits can keep pace with expectations — while money itself has become expensive again.

Take Notes

AI demand built the story.

Now earnings have to justify the price.

Want the story behind the market, not just the numbers?

Join The Sunday Letter

Sources & Method

Market data and company figures reflect information available through October 6, 2026. Breadth is used here as a concentration measure, not as a timing indicator. The 5% and 5.5% Treasury levels and 60% breadth level are watch lines for this analysis, not predictive thresholds.