IBM: What Its Earnings Misses Say About the State of AI Right Now

IBM pre-announced Q2 results with revenue of $17.2B (+1% YoY) and adjusted EPS of $2.93, both below analysts’ expectations, citing a capex shift by AI customers away from software and toward servers, storage, and memory. The stock fell about 25% in Tuesday trading, reflecting the earnings shortfall and the unexpected “magnitude of the capex reprioritization.” The note also implies a supply-driven price squeeze benefiting memory/storage peers (e.g., Micron up sharply), but highlights the cyclicality risk for those segments once supply catches up.

Analysis

This is less an IBM-specific disappointment than a live read on AI budget allocation: capital is moving from recurring software spend into constrained infrastructure inputs. Near term, that favors memory and storage names like MU and SNDK because procurement is being pulled forward into a tight supply window, which can keep ASPs and gross margins elevated for 1-2 quarters even if end demand is only moderate.

The second-order loser set is broader than IBM. Enterprise software vendors that were underwriting AI upside from incremental seat expansion may see a slower conversion rate if customers are funding infrastructure out of the same capex envelope; that can pressure FY26 guidance quality and multiple support across software-heavy large caps. The bigger risk is that this is inventory behavior, not secular demand: if lead times normalize or hyperscaler capex moderates, the pricing power in memory/storage can mean-revert quickly.

Contrarianly, the market may be over-reading the duration of the memory trade. What looks like durable AI demand may just be a temporary front-load ahead of supply tightness, and the easiest money is often made before peak cycle prints. Falsifiers to watch: falling DRAM/NAND contract prices, shorter lead times, or any sign that enterprise software budgets reaccelerate on the next round of CIO spend surveys.

Key Decisions for Investors

• Long MU on any post-news consolidation; 1-3 month horizon. Risk/reward remains favorable while pricing remains tight, but trim if management commentary starts to point to inventory bulge or weakening lead times.
• Pair trade: long MU / short IBM for relative-value rotation. Thesis is budget substitution from software to infrastructure; exit if IBM's next guidance update shows software stabilization or MU warns on ASPs.
• Add tactical exposure to SNDK only on pullbacks, not strength-chasing. This is a high-beta cyclical expression; best entry is after a 5-10% retrace if channel checks still show tight supply.
• Avoid buying IBM weakness until the market sees whether the miss is one-quarter noise or a structural AI monetization problem. A quick rebound above the pre-warning range would falsify the bearish read-through.