Dear Income traders,

Tech leadership just hit one of its most extreme readings since 1999. DataTrek flagged the chart below: State Street Technology Select Sector SPDR ETF (XLK), the tech-sector ETF, has outrun the S&P 500 by around 26.7 percentage points over the past 100 days.

That sits above their three-standard-deviation line (a move so far from normal it barely shows up once a cycle), and the only bigger spike on record was March 2000, at 42.3 points, right before the dot-com top.

No, that doesn't mean tomorrow is March 2000.

Markets aren't lazy enough to repeat on command. But it does mean tech is doing a massive amount of the index's heavy lifting.

Here's the thing: that 1999 comparison is the easy part. What matters more is what's actually doing the lifting. This isn't the broad mega-cap tech complex of old, it's semiconductors, almost by themselves.

The numbers back it up.

Semiconductors now make up close to 45 percent of XLK, and the top three names alone are about 39 percent of the fund. The hyperscalers (the giant cloud companies like Microsoft, Amazon, and Google that run the world's data centers) are actually getting hit by the rotation.

That's an important distinction, because "tech leadership" sounds monolithic and it isn't. Semis are the engine. The mega-cap cloud names are quietly spinning their wheels in debt issuance and heavy capex (the money they're pouring into building data centers and buying chips).

Take Micron Technology (MU). It's doing well because the world is desperate for memory and because it has real pricing power: it can charge close to what it wants. Micron just posted record revenue, up about 57 percent from a year ago, and still says it can only meet roughly two-thirds of the demand in front of it.

That's not the most diversified strength, but it is what it is. Strong leadership can persist longer than people think, and narrow semi leadership has run for months at a stretch before. But when the rubber band gets this stretched, and the leadership inside the leadership keeps narrowing, you stop pretending it's business as usual.

Concentration cuts both ways. When leadership is this narrow, the premium you collect for selling options can look friendly right up until one crowded trade starts moving against everyone at once.

It Wasn't a Tech Day. It Was a Rates Day.

This wasn't just a "tech is extended" day. Kevin Warsh's first decision as Fed chair landed hawkish, and the market heard the renewed focus on price stability loud and clear.

Look at what the Fed actually did. Back in March, the average policymaker still penciled in a rate cut for this year. Yesterday nine of them flipped to projecting a hike, the two-year Treasury yield (a clean read on where short-term rates are headed) jumped to 4.21 percent, its highest in more than a year, and the dollar posted its best day in almost a year.

Hawkish isn't shocking, but it isn't what the market wanted, and it raises the odds of a rate hike this year. Remember when cuts were a foregone conclusion, including by yours truly? Then Iran and the AI buildout happened, and the math changed.

That's the bigger read-through for today.

Concentrated tech leadership can keep the index levitated, but if rates lean on long-dated assets again (the ones most sensitive to where rates go, like high-growth tech), that same crowded leadership turns from support into an air pocket.

Add triple and quad witching today (the quarterly day when stacks of futures and options all expire at once, which can yank prices around), and round numbers become magnets, then trap doors.

Cheaper Compute Is Coming for the Simple AI Trade

There's a second force pulling in the same direction: open-source AI models keep getting better and cheaper.

At some point that breaks the clean "whoever builds the best model wins everything" story. I'm starting to wonder what happens to OpenAI-style economics if the whole market keeps racing toward cheaper and cheaper compute.

Microsoft (MSFT) just signaled this to the entire enterprise market. It rebuilt Copilot as a multi-model platform that can route across its own in-house models, Anthropic's Claude, and even cheaper open-source engines it is now weighing, like DeepSeek. The reason is simple: frontier-model token bills don't scale at those prices, and the cheap open-source option reportedly runs about a third of Claude's cost on the input side.

To be clear, that's not bearish on AI. It may be bearish on the simple version of the AI trade. If model capability turns into a commodity and compute costs fall, the value keeps spreading outward: into infrastructure, power, industrials, financials, energy, cooling, capital equipment, and the companies that use AI to fatten their own margins.

That's more fuel for the broadening trade. Industrials and financials both traded today like they want to break out, not roll over. That's exactly the behavior I want to see if leadership is finally widening out beyond the usual narrow tech names.

Your Playbook Into the Long Weekend

So the playbook is simple. Remember there is no Friday session this week, today is triple witching, and then the market goes dark until Monday. Everything below gets managed into the close, not parked overnight into a three-day gap.

Respect the stretch. A three-standard-deviation move is not where you get heroic with size.

Watch rates first. If short-term yields and the dollar keep climbing, intraday tech leadership gets harder to trust.

Watch the broadening. Industrials and financials breaking out tell you this isn't a one-lane AI trade anymore.

I also picked up some low-delta puts on Invesco QQQ Trust (QQQ), the Nasdaq 100 ETF. Those are cheap, far-out-of-the-money puts that act as low-cost insurance if the index rolls over. Haven't done that in a while.

Check out Turbo Income for how we trade these sorts of names.

Here for a good time AND a long time,

Hans