AI Stocks Just Lost Their Hedge. Bonds Are Quietly Back on the Table.

Twenty-two percent. That's what the Goldman Sachs AI basket has handed back since early June, and the drawdown has done something more interesting than bruise a few momentum funds. According to Fidelity's Jurrien Timmer, the selloff flipped the correlation between AI stocks and long-term Treasuries negative.

That's a big deal for anyone running a concentrated AI book. For most of the past two years, AI names and long bonds were cousins. Both took their cues from the same rate signal. When yields dropped, anything with a growth narrative caught a bid. When yields climbed, the whole trade sagged in unison.

Some hedge, right?

Now the link has inverted, and that changes the math on what a bond allocation is actually for.

The hedge nobody had

Here's the uncomfortable truth about concentrated AI exposure. It never really had a hedge. The 60/40 crowd assumed bonds would cover them, then 2022 showed up and stocks and bonds fell together while the Fed hiked. So the AI crowd went bare, or close to it, and told themselves the growth story was the hedge.

Timmer's numbers push back on that. If AI stocks and long Treasuries now move in opposite directions, bonds start earning their keep again. Admittedly, the effect is conditional. It works when yields fall on growth fears, which is exactly the scenario that would hit AI names hardest. It doesn't work when yields rise because inflation is sticky, and that's the version of 2022 nobody wants to replay.

Who wins, who gets caught

The winners are the boring ones, for once. Risk-parity funds, vol-targeting strategies, and any allocator who sat on a bond sleeve despite two years of being told it was dead weight. They get ballast that actually holds when the growth trade breaks. The losers are the people who swapped out of bonds for more AI beta because this time felt different.

But what happens if the next leg down in AI isn't about growth at all? Then the hedge fails right when you need it. That's the trap buried in Timmer's framing, and it's worth flagging before anyone treats long Treasuries as a magic shield. The note also points to a support level near 357 for concentrated AI holders. If that gives way, the bond cushion gets tested for real.

I'm not entirely convinced. A correlation that flipped in a few months can flip back just as fast, and bond demand is only as reliable as the next inflation print allows it to be. That's not a durable fix. That's a window.

What to watch

Two things. First, the path of the 10-year yield. If the next move lower comes from genuine growth cracks, the negative correlation holds and bonds do their job. If it comes from a Fed easing into sticky inflation, the hedge breaks and AI holders are on their own again. Second, inflation data over the next quarter. That's the variable that decides whether this is a real regime change or a brief detour.

The takeaway is simple. AI concentration has been an unpriced risk for two years, and the market just handed allocators a way to price it again. The skeptics who kept their bonds look right for a change. That doesn't make them geniuses. It makes them hedged.