The AI Trade Holds Up to Rising Yields — Until Earnings Stop Outrunning Price

The 10-year Treasury crossed back above 4.8% this week, oil is above $100, and the S&P 500 barely blinked. So when Bank of America steps in to tell clients that bond yields have to spike much higher before they derail the AI trade, the instinct is to file it under "sell-side reassurance" and move on. But the firm is making a genuinely mechanical point, and it's worth taking apart — because whether rising yields actually break the AI trade comes down to a mechanism most headlines skip.

Why yields are supposed to hurt AI at all

The threat is a discount-rate story. A stock is worth the present value of its future cash flows, and higher bond yields mean a higher discount rate — which shrinks that present value. The more of a company's value sits in far-off profits, the harder it gets hit. That's why the word that keeps coming up is "duration." Long-duration assets, the ones promising cash five or ten years out, are the first to break when rates climb.

The AI trade looks like the longest-duration trade in the market, which is why the fear smells reasonable. But here's the catch nobody states plainly: the AI trade is not one asset. It's a basket of companies with very different cash-flow profiles, and rising yields punish them very differently.

The part of AI that isn't duration

Nvidia is up roughly 24% over the last four months, and Microsoft is up about 25%. Both are earning enormous sums right now on the back of an AI infrastructure build-out that the big cloud providers are bankrolling at a scale that is frankly hard to get your head around — the five largest US cloud and AI players alone have committed to somewhere between $660 billion and $690 billion of capital spending in 2026, nearly double 2025 levels. That is cash landing today, not a promise for 2032. Earnings now means the discount rate matters less.

That is exactly the mechanism BofA is leaning on. The firm's strategists argue that sector earnings growth is outpacing share-price appreciation, which keeps driving price-to-earnings multiples down. Think about what that means. A stock that goes up 24% but sees its earnings go up more than 24% has gotten cheaper, not richer. The trade is de-rating while it grows. That is the opposite of a bubble dynamic, and it is why the firm's own price-based bubble-risk indicator shows no bubble — and why its financial-stress index shows derivatives markets staying sanguine through the higher yields.

The history supports the "bar is high" reading, up to a point. In the late 1990s, long-bond yields rose by more than 200 basis points and the Fed hiked rates by more than 100 basis points while the Nasdaq kept soaring. Stocks back then had clearly decoupled from fundamentals; BofA argues that decoupling is not present now. And the bond market is reacting more calmly than it did in 2022 — the 30-year's worst drawdown this year is about 7%, versus a 25% sell-off four years ago.

What would actually break it

Now the flip, because the reassurance only holds under one condition — and it's a condition you can watch in real time.

First, why yields are rising matters. This is the dangerous flavor of yield move. The 10-year is climbing not because the economy is booming and real growth is pulling rates up, but because oil above $100 is feeding inflation fears and the market is demanding more to hold long-dated US debt against a ballooning fiscal picture. That's a term-premium and inflation move, not a growth move. When rates rise on growth, there's an earnings offset; when they rise on inflation and fiscal supply, there isn't. The dot-com-era comparison therefore flatters the current setup: back then, rising rates largely came with a strong growth backdrop. This one is closer to stagflation-ish, and growth assets tend to fare worse in that flavor of tightening.

Second, and this is the signal that would change the whole reading: the moment price appreciation starts to outrun earnings growth, the AI trade stops being an earnings trade and becomes a pure rate trade again. That's the switch. As long as multiples keep compressing, the rate headwind is being absorbed by the cash flow — BofA's "high bar" holds. The trade breaks, mechanically, when the stock prices finally run ahead of the earnings that are supposed to justify them, or when the long end spikes enough to force the disorderly repricing that BofA itself flags as the risk — a "disorderly bond market tantrum" if upcoming Fed meetings don't give the market the tightening it seems to want. The 30-year is already above 5.1%. It doesn't have to rise much further to start demanding attention.

There's one more layer worth a sentence, because it cuts against the comfortable conclusion. The resilience is concentrated. The momentum over the past several months has been carried by a handful of mega-cap AI names while the broader market churns — the S&P 500 is up about 12% year to date, but most of that has come from the leaders. That means "the AI trade is absorbing higher yields" really means "a few giant companies with huge current earnings are absorbing higher yields." Broad, equal-weight resilience it is not. The calm is real, but it's a narrow calm, sitting on top of heavy put positioning in the options market and an options market whose apparent composure has not yet been tested by a violent long-end move.

So the useful takeaway isn't "yields don't matter." It's "yields matter, but only through a specific switch." Keep watching whether the AI names keep getting cheaper as they grow — that is the sign the rate headwind is being absorbed and BofA's reading holds. The moment the P/E stops falling and price starts outrunning earnings, the trade's duration comes back, and a 4.8% yield that was tolerable stops being tolerable quickly. That's the condition to evaluate, not the headline.