IBM's Worst Day in 115 Years, and the Tag That Almost Fired
Some days the market hands you a live-fire test of your own system. Today was one of them.
IBM had the worst single trading day in its history as a listed company — worse than Black Monday 1987 — after pre-announcing a Q2 shortfall a week ahead of its scheduled July 22 earnings call. The stock closed around $217, down roughly 25%. And IBM has been sitting on my watchlist tagged WOULD OWN AT ~$210 since I finished the dossier in May. The trigger is now three percent away.
I'm not acting. Why not is the most useful thing this system has taught me so far.
What I observed
From IBM's 8-K filed July 14, 2026 (Exhibit 99.1, Arvind Krishna's letter to investors) — preliminary figures; the quarter isn't fully closed:
- Revenue $17.2B, up 1% year over year — against analyst consensus of roughly $17.9B (consensus per financial media, secondary source)
- Software up 5%; consulting flat; infrastructure down 7%
- Diluted EPS $2.93 operating non-GAAP (up 5%) vs. consensus ~$3.01–3.02
- Year-to-date free cash flow $4.8B on $7.8B operating cash flow
- Red Hat growth accelerated to 11%; distributed infrastructure grew 37% with a ~$500M backlog
Management's explanation, in its own words: clients shifted late-June capex toward supply-constrained servers, storage, and memory ahead of expected price increases; IBM "did not anticipate the magnitude of the capex reprioritization," and numerous large deals slipped past quarter-end — teams "faltered" and didn't adapt fast enough.
Market context, from secondary sources: the close near $217 came on 5–6x average volume, down about a third from the June 2 high of $332.46 in six weeks. In April, management had reportedly guided toward 10%+ software growth (needs verification against the Q1 transcript). At least one securities-litigation investigation was announced the same day — a monitoring item, not evidence of wrongdoing.
Why it matters
In the naive version of my system, this is the moment everything was built for: the dislocation arrives, the pre-committed price is nearly touched, and I act calmly while everyone else panics. Three separate checks say otherwise.
The trigger's assumptions just got hit. My $210 tag was derived from a fair value range built in May, and that range rested on specific inputs — software compounding at high single digits or better, durable free cash flow, a mainframe cycle that decays gracefully. Today's letter stresses those inputs directly: software grew 5%, infrastructure fell 7%. A trigger price is downstream of a fair value range, which is downstream of a thesis. When a material event hits the thesis inputs, the trigger doesn't get executed — it gets suspended. A WOULD OWN AT tag is only as current as the dossier behind it, and a material event stales the dossier on the spot, regardless of its calendar age.
This is informed selling, not forced selling. One of my catalyst patterns is buying technical unwinds — but the mandatory check is verifying the selling is technical. Here the company itself told the market the quarter missed. I don't get to pretend otherwise because the price is attractive.
Two behavioral rules block action anyway. No decisions in the heat of a market move, and no action inside a pre-earnings window. Full results and — more importantly — full-year guidance arrive July 22. A pre-announcement is a partial information release by design. Acting on partial information eight days early isn't conviction; it's impatience wearing conviction's clothes.
The only question that matters for the refresh is whether this quarter reflects deal timing or demand destruction. Management says timing: capex reprioritized, deals slipped rather than died. The bear reading is that AI is starting to structurally substitute for the software and consulting spend IBM sells. Nobody can distinguish those from a pre-announcement letter — and management's framing comes from the party with maximum incentive to call the miss temporary, from a team that reportedly called this exact dynamic "de minimis" in April, ninety days before it produced the worst day in company history. Their story gets weighted as a hypothesis to test against the July 22 numbers, not accepted as an explanation.
My take
Working through this forced me to restate my bull case honestly, and two of its four pillars didn't survive my own filters. Reputation is an output of a moat, not a moat — and this quarter is a live counterexample, because toll-booth demand doesn't get deferred the moment competing hardware priorities appear. Watsonx fails my rule that a story must be grounded in current economics: IBM doesn't disclose it as a clean revenue line. The dividend is capital allocation, not defensibility. What survives is narrower and stronger: a high-margin recurring toll booth — mainframe-attached transaction-processing software plus Red Hat/OpenShift as the hybrid-cloud substrate for enterprise AI workloads — with a cyclical hardware kicker. Red Hat accelerating to 11% in the worst quarter in company history is the one verifiable AI-adjacent datapoint in the whole story.
Before the call, on purpose, here's the provisional scenario map — anchors from primary sources (939.9M shares per the Q1 10-Q; FY2025 FCF of $14.7B; ~$217 ≈ 13.9x trailing FCF), everything else flagged as analyst assumption:
| Case | Condition | FY2026 FCF | Multiple | Implied value/share |
|---|---|---|---|---|
| Bull | Core criteria pass cleanly; slipped deals verifiably closing | $14.5–15B | 17–18x | ~$262–287 |
| Base | Criteria pass but softer; FCF guide roughly held | $13.5–14B | 15–16x | ~$215–238 |
| Bear | FCF guide cut, or substitution reading gains evidence | $12–12.5B | 12–13x | ~$153–173 |
Weighted 25/50/25 — the bear weight deliberately elevated because of the April calibration miss; trust is priced — gives a fair value near $223. My margin-of-safety rule (15–25% below fair value) puts the buy zone at roughly $167–190.
That's the uncomfortable arithmetic: at $217, a 25% crash brought IBM to approximately fair — not cheap. The market didn't overreact by my math; it repriced. And my old $210 trigger is provisionally too high under these assumptions, not too low. My gut this morning said "the trigger is 3% away, this is the moment." The arithmetic says the moment is another 12–23% lower, if it comes at all. One of those is analysis; the other is anchoring on a stale number.
Six weeks ago IBM traded at $332 and the tag kept me indifferent. Today it traded near my trigger and the same system forced the pause. Both halves are the system working.
What I'm watching
July 22, 5PM ET. Criteria written now, before the call, so the numbers get scored against pre-commitments instead of whichever picture I walk in wanting to see:
- Full-year FCF guide — the single most load-bearing number; the toll-booth thesis is the FCF. Pass: maintained within ~5% of the prior guide. Fail: cut materially or replaced with vague language.
- Software decomposition — pass: management names precisely where the Q2 software gap sat and the FY software guide survives at high single digits. Fail: explanations stay at "clients reprioritized" altitude.
- The slipped deals — pass: deals signed in July, backlog conversion quantified. Fail: "we remain confident in our pipeline" without numbers. Vibes are a fail condition by definition.
- Infrastructure trajectory — pass: the −7% reconciles with the z17 launch comp and the FY guide holds its April framing. Fail: the decline is steepening beyond the mainframe cycle's normal shape.
- Management calibration — pass: an unprompted reconciliation of April's "de minimis" with July's outcome. Fail: it goes unaddressed. This one determines how much weight their next guide deserves.
- The AI substitution question, asked directly — pass: engagement with segment-level evidence. Fail: deflection to the blended "gen AI book of business" metric, which doesn't answer the question.
Scoring, pre-committed: criteria 1–3 are the core. All three pass → the timing story wins and a new trigger gets derived from a refreshed dossier. Criterion 1 fails, or two of the core three fail → the refresh starts from rebuilt assumptions and no trigger is set until it's done. Mixed → the tag stays suspended and IBM goes to the back of the queue. Under no scenario does a buy decision happen on July 22 itself — the no-heat-of-the-moment rule applies to good news too.
2026-07-14 — IBM — No action. Q2 pre-announcement (8-K, Ex. 99.1). Stock −25% to ~$217 vs. WOULD OWN AT ~$210. Trigger suspended, not executed: material event stales the May fair value range. Selling is informed, not technical. Pre-earnings window active through July 22. Bull case narrowed to mainframe transaction-processing + Red Hat toll booth with cyclical hardware kicker. Provisional scenarios bull ~$275 / base ~$227 / bear ~$163, weighted ~$223; buy zone ~$167–190; prior $210 trigger flagged as likely too high. Pass/fail criteria for the call logged above, before the call. Dossier refresh mandatory after it.
Nothing here is investment advice. This is a public record of one person applying their own written framework to their own decisions — the point is the process, not the picks.