IBM −23%: The Memory Squeeze Arrives in Enterprise Budgets

Companion to today's decision-levels Pulse — the split tape just produced its first mega-cap casualty.

What is the buzz

IBM did something companies almost never do: it pre-announced a miss. Preliminary second-quarter results — revenue of $17.2 billion, up one percent but roughly $650 million short of expectations, adjusted earnings of $2.93 against the $3.01 the street carried — released a week before the scheduled print. The stock fell 23%, its steepest single-day decline in decades. Infrastructure revenue fell seven percent on a mainframe shortfall, and the transaction-processing software attached to those mainframes fell with it.

So what

The causes IBM's CEO named are two stories this diary has been tracking for weeks — now printing in one company. First: enterprise customers spent June front-running memory, server and storage price increases, pulling capital forward before the supply squeeze bites. That is the DRAM cycle — the same force that halted Korea's market on Monday a week ago and repriced the memory complex — arriving in corporate IT budgets and eating the mainframe renewal cycle on its way through. Second: customers shifted spend toward AI while the legacy stack shrank. Budgets are not growing into both worlds at once; the build-out is being funded partly out of the old stack's pocket.

Legacy weakness, not software weakness

The headline says IBM cited "weakness in software" — but look inside the print. Red Hat grew eleven percent. Distributed infrastructure grew 37 percent. Generative-AI consulting signings kept rising. What fell was the mainframe Z system and the transaction-processing software chained to it. The split inside IBM is the same split the tape has been trading all week: modern software holds its bid, the legacy stack loses budget to the AI build-out. This is what the SaaSpocalypse thesis looks like when it stops being a framework and starts being a pre-announcement — the displacement pressure lands first on the oldest revenue, not on software as a class.

Now what

The tape's first reaction was sell everything, ask later. The disruption fear spread across the whole software complex within the hour — ServiceNow fell as much as eight percent, Salesforce and Adobe five to six — the market suddenly asking every enterprise-software name the question IBM had just answered badly: how much of your revenue is legacy? Then, within two hours, it started sorting. The names on the disrupted side of the question stayed down hard. The names on the disruptor side reversed to green: Datadog +2%, Palantir recovered its entire morning loss — and Cloudflare traded through the 280 all-time-high line named in this morning's Pulse to a new intraday record, in the middle of the panic. The cloud basket dipped below its 24 pivot and reclaimed it; the semi complex never gave back the floor, holding above 560 throughout.

That two-hour sequence is the whole thesis compressed into an afternoon: the market sells software as a class on legacy fear, then re-prices it name by name into disrupted and disruptor. What to watch from here: whether ServiceNow, Adobe and Salesforce close at their lows — the market's first legacy-exposure list, written in real time. IBM's full report next week — the name was already on the Print Record forward calendar, and its reaction history just got a new data point a week early. Tomorrow ASML, Thursday TSMC: the referees on whether the memory demand IBM's customers front-ran is a squeeze or a boom. And the AI Build-Out layers on the Rubin view — if enterprise budgets are being pulled toward memory and servers, Layer 3 is where it shows up first.