The 10-year Treasury yield has climbed from roughly 4.96% in mid-August to as high as 5.31%, sitting at 5.27% as of this week. The 30-year has moved from 5.29% to a peak of 5.66%, now 5.64% -- levels that put both well above where they sat for most of the last few years. The textbook assumption is that this is exactly the environment high-multiple tech can't survive. Over the last seven weeks, that's not what happened.
A stock's price is, in theory, the present value of its future cash flows. The further out those cash flows sit, the more a higher discount rate shrinks their value today -- which is why "long-duration" growth stocks (companies priced mostly on earnings years from now, not this quarter) are considered the most exposed when yields rise, and why defensive, bond-proxy sectors like utilities are supposed to hold up better since their cash flows are steadier and nearer-term.
Pulling the real numbers instead of assuming the mechanism played out as expected: QQQ (Nasdaq 100) rose 4.1% over the same stretch the 10-year and 30-year were climbing to their highs. NVIDIA, the single largest AI-capex bet in the market, rose 5.5%. ARKK -- the ETF built almost entirely out of the kind of long-duration, far-out-earnings growth names the textbook says should get hit hardest -- rose 8.0%, outperforming everything else in this comparison. The one name that actually underperformed was XLU, the Utilities Sector SPDR -- the classic defensive, rate-sensitive "bond proxy" -- down 6.9% over the identical window.
That's close to the opposite of the textbook script: the most rate-sensitive growth basket led, and the defensive sector that's supposed to be the safe harbor lagged everything.
The more precise story is in the shape of the move, not just the level. The Fed funds rate actually eased over most of this window -- from 4.09% in mid-August down to 3.63%, before ticking back up to 3.75% most recently -- while the 10-year and 30-year climbed the entire time. Short rates flat-to-down, long rates up sharply is a bear steepening: the market isn't pricing a Fed actively fighting inflation with higher policy rates, it's pricing something at the long end specifically -- more Treasury issuance, inflation risk further out, or both. That's a different trade than "the Fed is tightening," and it explains why it hasn't automatically translated into a tech selloff: the earnings conviction behind the AI buildout (the same Nasdaq-100-at-highs, NVIDIA-near-highs move we covered this week) has, so far, mattered more to tech multiples than the discount-rate math on its own.
"So far" is doing real work in that sentence. The mechanism isn't wrong, it's just been outweighed -- which means it can stop being outweighed. Two things would flip it: the long end continuing to climb well past these levels without a growth or earnings story strong enough to offset it, or a genuine AI-capex earnings disappointment that removes the thing currently winning the argument. Neither has happened yet. The Layer Rotation Heatmap and Momentum Map are the places to watch for the first sign of that actually turning -- a real rotation out of the AI-compute layers into defensives, not just a single red day.
Want to see it yourself?
Launch App Free →