The AI Doesn't Care About Your Rate Hikes

Micron just reported. Sit with the numbers for a second before you reach for an explanation.

Revenue of $41.46 billion. Adjusted EPS of $25.11. Beat by $4.62 a share, and $5.77 billion on the top line. Gross margins at 84.9% — a company record. Free cash flow at a record, too. And then the guidance: fiscal Q4 revenue of $50 billion, with EPS of around $31. This is a memory chip company. A business that, twelve months ago, printed $9.3 billion in quarterly revenue. The year-over-year comparison is so violent it doesn't look like the same industry — because it isn't.

The stock is up 727% over the past twelve months. Up 270% since January. After the close, it jumped another 13% in after-hours to $1,185, closing in on its 52-week high of $1,213. The session had shaved it down to $1,047 intraday while investors debated whether the print would justify the altitude. It did. It did more than justify it.

Here's the uncomfortable question nobody on the earnings call bothered to ask: at what interest rate does this stop?

---

The backdrop matters now more than it has in two years. BofA flipped its entire Fed call last Monday — from no hikes this year to 75 basis points of tightening, with 25bp moves penciled into September, October, and December. BofA economist Aditya Bhave's diagnosis was direct: "The Fed's inflation problem has gotten unambiguously worse." The Fed's own June dot plot revised the 2026 PCE forecast up to 3.6% from 2.7% — a move too large to treat as noise. Nine of 18 FOMC members now expect at least one hike this year.

And then there's Warsh. His first statement as chair ran 114 words — less than half the length of the April statement — with forward guidance noticeably absent. The man obliterated the prose. He said more by deleting sentences than any Fed chair has said by adding them. "The commitment to deliver is strong, unanimous, and unambiguous, and that's an important message we've missed for five years, and we're going to fix that," he told reporters. That last clause is doing serious work. Five years of above-target inflation. Five years of explanations, frameworks, transitory narratives, and reactive pivots. Warsh walked in, nuked the statement, and filed five task forces. Whether that's courage or theater won't be clear until September. But the signal is priced — CME FedWatch puts the probability of a December hike at 87.9%.

The Micron result sits across the table from all of that and doesn't flinch.

---

HBM3E and HBM4 are fully booked through calendar 2027, with demand extending into 2028. The company said fiscal Q3 reflected a market where demand for memory chips continues to outpace supply. CEO Sanjay Mehrotra on the call said market tightness is "locked in to persist beyond calendar 2027." The supply-demand geometry here isn't conventional. It isn't a commodity cycle wearing a growth costume. Memory process technology is getting more complex with each node — transitions are leading to slower bit growth, increasing cleanroom space requirements, while HBM's growth further pressures non-HBM supply. The hyperscalers are consuming memory faster than the fabs can build capacity to make it.

HBM4 12-high is ramping twice as fast as HBM3E 12-high, and Micron has already shipped more than $1 billion in HBM4 revenue. Management declined to give guidance beyond Q4 but said margins won't significantly shrink even in a downside scenario, given the structure of its strategic customer agreements. Price floors on those SCAs mean that even if data center demand cooled, you'd be looking at a very high-margin environment for Micron under the agreements.

Lock-in. Price floors. Demand visibility into 2028. These aren't chip company talking points. They're characteristics of a utility.

---

The intellectual problem is this: we now have two completely different stories about the economy running simultaneously, and they're both true.

Story one. Headline CPI rose 4.2% year-over-year in May, with the Cleveland Fed now projecting headline PCE could come in at nearly 4% for the same month. Housing-driven disinflation has mostly run its course. Energy costs tied to the Iran conflict have added pressure. The Fed has a real problem and Warsh, unlike his predecessor, seems constitutionally incapable of looking away from it. BofA's three-hike call isn't reckless — Deutsche Bank joined with two hikes of its own in a June 19 note. The rate structure is moving.

Story two. An AI capital expenditure cycle so structurally committed that hyperscalers are signing multi-year agreements with price floors embedded in them, bidding for memory that won't even be manufactured yet, and funneling cash into capex at a pace that makes the dot-com buildout look like a mild enthusiasm. Micron guided $50 billion for a single quarter. That's $200 billion annualized revenue for a memory business. SK Hynix is up 826% over the past twelve months. Nvidia is the gravity well and everyone else is in orbit — but the orbits are getting denser.

Higher rates slow housing, discretionary spending, small business formation, leveraged buyouts, car loans. They do not, in any near-term measurable way, slow a sovereign wealth fund wiring $10 billion to a hyperscaler to build out inference capacity. The capex cycle is funded. The contracts are signed. The cleanrooms are under construction. Micron has begun construction on a second cleanroom at the Tongluo site to support EUV equipment, with Singapore becoming another centre of excellence for advanced packaging. Interest rate sensitivity doesn't apply to committed infrastructure spend the way it applies to a 30-year mortgage.

---

This is what makes today's macro environment genuinely strange. The Fed is tightening into an economy where the dominant growth engine is rate-insensitive at the margin. Rate hikes will bite — they always do — but they'll bite the parts of the economy that aren't driving the earnings multiple: housing, consumer credit, small caps. Meanwhile MU is guiding $50 billion, NVDA is still the most important company in the world by any honest accounting, and the index concentration risk grows one quarter at a time.

The 10-year Treasury yield slid to 4.398% on Wednesday even as the 2-year hit its highest level since February — the curve pricing a short-term hawkish impulse without a conviction view on where growth lands. The market isn't wrong to be confused. There's genuine uncertainty about whether Warsh hikes into a slowdown or hikes alongside an AI boom that insulates corporate earnings from the usual transmission channels.

The answer probably involves both happening at once: the Fed tightens, the rate-sensitive economy wobbles, and the AI infrastructure complex reports another five straight record quarters and announces a dividend.

Today was a data point. Tomorrow, May PCE prints — and with Cleveland Fed models projecting headline PCE near 4%, the question of whether September is a hike or a hold becomes a lot less academic. Warsh wanted credibility. The data is giving him a stress test to earn it.

Micron, for its part, just printed a quarter that renders the concept of cyclicality temporarily obsolete. Whether that remains true at 4.5% Fed funds is the only question worth asking.