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

IBM released preliminary 2Q results ahead of July 22, guiding to revenue of $17.2B (+1% YoY) and adjusted EPS of $2.93, both below analyst expectations. Management attributed the shortfall to a capex reprioritization by AI customers—shifting spending from software toward servers, storage, and memory amid hardware shortages and rising prices. IBM shares fell 25% in Tuesday trading on the earnings-related shock, signaling near-term weakness despite strength in the broader memory/storage supply chain.

Analysis

The immediate market read-through is not “IBM-specific weakness” so much as a capital-allocation signal: enterprise AI budgets are still dominated by physical build-out, which tends to compress valuation for software-heavy vendors while extending the earnings cycle for memory and storage suppliers. That favors MU and SNDK first, with the higher-beta move likely in the next 1-3 quarters as customers pre-buy against feared shortages and price hikes. The second-order effect is that software spend can look artificially soft even when end-demand is healthy, so near-term underperformance in IBM-type software proxies may be a timing issue rather than a secular demand collapse.

The risk is that this is exactly how cyclical peaks announce themselves. When customers rush to secure supply, it often marks the point where ASPs and margins are strongest but incremental upside becomes more fragile; once foundry and NAND/HBM capacity catches up, the multiple usually de-rates before the earnings do. For IBM, the bigger concern is not the one-quarter miss but the possibility that its software mix remains structurally de-prioritized for several quarters, limiting operating leverage and forcing the market to value it more like a slow-growth services/infra hybrid than a resilient software compounder.

Contrarian view: the move may be overdone on both sides. IBM’s selloff could overshoot if management frames this as timing rather than demand destruction, while MU/SNDK may already be discounting a very strong cycle and ignoring the point where customers stop front-loading purchases. The cleanest falsifier is a quick reacceleration in IBM’s software bookings or evidence that memory lead times/ASP trends are flattening; that would argue the current positioning is too one-directional.

Key Decisions for Investors

• Pair trade: long MU / short IBM for 1-3 months, on the thesis that AI capex is still flowing to the hardware layer while IBM’s software mix faces temporary budget deferral; exit if IBM management restores software growth or MU commentary implies inventory normalization.
• Use SNDK as a higher-beta expression of the same theme, but only via call spreads or a small starter position; memory names can re-rate quickly, but the risk/reward worsens sharply if customer pre-buying has already pulled forward demand.
• Fade the knee-jerk IBM drawdown only if post-earnings guidance shows stable free cash flow and software bookings; otherwise keep IBM on the short/watch list for 1-2 quarters as a relative-growth laggard.
• Watch for a rotation into software ETFs or enterprise-app names if memory ASPs stop rising; that would be the first sign the current capex mix shift is peaking and the trade is moving from hardware winners back to software laggards.