IBM Trigger Review: The Alert Fired Because the Thesis Broke — NO BUY #

16 July 2026 | Price: ~$206 (new 52-wk low $204.44 intraday) | Trigger: IBM-GRAHAM alert @ $210 (set 2026-06-17) | Verdict: NO BUY — kill condition hit before entry

Follow-up to: 2026-06-17-ibm-quantum-hype-not-a-buy.md

Why the alert fired #

Not a quiet drift into the buy zone. On July 14, 2026 IBM pre-announced Q2 via 8-K and fell ~25% in one session — the worst single day in the company's 115-year history (67M shares vs 10.6M avg volume). HSBC downgraded to Reduce with a $191 PT (below the current price); Oppenheimer downgraded the same day it had reiterated Outperform hours earlier.

The June alert's premise was "good company, wrong price — wait for the price to come to you." The price came down because the company stopped being the June version of itself.

The prelim numbers vs the alert's own gates #

The June alert note set explicit conditions. Scored against the July 14 preliminary 8-K:

Total revenue decelerated +9.5% (Q1) → +1.3% (Q2). Infrastructure swung +21% → −7% (z17 cycle rolled over). Revenue $17.2B missed consensus by $660M; adj. EPS $2.93 vs $3.02 — first miss after 5+ straight beats. CEO Krishna: "We did not adapt and move quickly enough" — clients redirecting budgets to AI infrastructure (servers/GPUs) away from traditional software and the mainframe stack.

This is the exact shr-043 falling-knife pattern (RI.PA analog): the "E" under the P/E is falling toward the price. A trough multiple on a decelerating base is not a margin of safety.

shr-042 verification: IBM-specific, not complex-wide #

The deeper IBM-DEEP alert assumed a drop to these levels would be a broad AI/quantum-complex selloff. Verified false: ~70–75% of the June 17 → July 16 decline is IBM-specific (the July 14 pre-announcement), ~25–30% is sector beta (Accenture guidance cut June 18, quantum-basket unwind, Fed hawkishness). Context: the quantum hype pop (ATH $332 on June 2) had already unwound before the June 17 blog; this leg was new, negative, fundamental information. Sector backdrop is genuinely ugly too — ACN −50% YTD, MSFT worst month since 2000, ORCL worst since 1990 — but IBM is the one that confirmed the "AI capex crowds out software" fear with a hard print.

The valuation math is stale #

• June math: fwd P/E 15.6x at $210 on FY26 EPS $13.45 → Graham IV at g=5% ≈ $249 → ~16% MoS. Dead.
• Post-warning FY26 consensus is being cut toward ~$12.0–12.8 (unconfirmed until revisions settle). At $206 that's fwd P/E ~16–17x — F6 fails again.
• Graham IV at g=3% on ~$12.4 ≈ $180 — below the current price. At g=5% ≈ $229, but g=5% can't be assumed when the latest quarter printed +1.3% and both growth segments decelerated. Break-even growth at $206 ≈ 4% — no longer conservative, now a bet.
• Local screen: 2/7 (June: 3/7). Caveats: yfinance dividendYield=20.0 is the known scaling bug; earnings-stability fail may be the 4-yr-window artifact. The price filters (P/E, P/B×P/E, current ratio 0.80) fail on real data regardless.

shr-020 red-flag summary #

• RED: price action (worst day in history, new 52-wk low, no stabilization); earnings/news (prelim miss + execution-failure admission, full report pending); analyst direction (2 downgrades in 24h, estimate cuts in progress, lowest PT $191 below price).
• AMBER: dividend (covered ~60% of SBC-adj FCF, but growth effectively stalled — third kill condition is borderline); balance sheet (ND/EBITDA 3.35x rising, negative tangible book — known); short interest (4.05% of float, low, but +12.2% MoM = informed bears adding, shr-023); insiders (zero open-market sells all year — but also zero buys during the crash; the Feb directors who bought at $233–237 are underwater).
• GREEN: nothing new.

Verdict and what happens next #

NO BUY at $206. The alert's own kill condition (Software <8%) is hit, the shr-043 gate fails on both segments, and the entry-zone math was built on an EPS that no longer exists. Buying now would be catching a falling knife six days before the full report — the same mistake RI.PA taught.

Actions taken today:

1. IBM-GRAHAM ($210) alert removed — fired, thesis broken, keeping it would be stale automation (shr-041).
2. IBM-DEEP ($185) re-armed with a rewritten note — the old note's core assumption ("a drop this deep = broad selloff, not IBM-specific") is now verified false. New gate: no entry on price alone; requires the full Q2 report + FY26 guidance (July 22) to show Software ≥8% cc durable, Consulting stabilizing, FCF guidance ≥$12B, dividend maintained — then a full re-score on post-revision EPS.

What would reopen the file (July 22 checklist):

• Software segment guide confirms ≥8% cc for H2 (Red Hat +11% is the bright spot — if it's the compounder and Transaction Processing was the drag, the mix story matters).
• Infrastructure miss framed as one-time capex timing, not demand destruction.
• FY26 FCF guidance (was ~$13.5B) lands ≥$12B and the dividend narrative holds.
• Consensus FY26 EPS settles, so the IV table can be re-derived on real numbers.

Funding note: even on a "buy" verdict, this had no authorized funding — the EUR 250 YCSH opportunity reserve is earmarked for MSFT-DIP350/VALUE310, DG-ENTRY, DPZ-ENTRY only (and DPZ is currently TRIGGERED with first claim on it). An IBM entry would have required a new charter-level allocation decision. Moot today, but worth remembering if July 22 revives the name.

Data: yfinance + Finnhub pulls 2026-07-16; SEC 8-K 2026-07-14 (prelim Q2) + Exhibit 99.1; CNBC/Forbes/MarketScreener July 14–15 coverage. Segment figures are preliminary until the July 22 full report.