IBM's Krishna tries to reassure investors that AI won't disrupt company's software unit

IBM CEO Arvind Krishna said only ~2% of IBM software could be replaced by AI-constructed applications, arguing the remaining software is “infrastructure” that supports AI-ready data and hybrid cloud—despite Wall Street skepticism after a disappointing Q2. The article highlights a sharp software demand issue tied to mainframe hardware capacity and AI-driven memory price spikes, with z mainframe infrastructure revenue down 42% and transaction processing software down 9%, while IBM still expects a $1B free-cash-flow tailwind in 2026. Guidance for software revenue growth has shifted toward 6%–8% for the year (from double-digit expectations), and analysts at Jefferies note the stock is recommended even as ~75% of Q2 deal slippage may return before year-end.

Analysis

The market is likely underpricing the earnings-quality issue rather than the headline AI risk. If a small amount of legacy software is already vulnerable, the bigger problem is that IBM’s profit pool is concentrated in mature, maintenance-heavy products where even modest churn can keep growth stuck in the mid-single digits and cap the multiple. That matters because software is the highest-margin layer, so every point of lost mix pressure has an outsized effect on FCF durability and on how credible the 2026 bridge looks.

Second-order, this is a relative winner for modern cloud and workflow vendors that can prove faster AI integration and lower switching costs. Enterprise buyers will use any perceived obsolescence in legacy platforms to renegotiate contracts, stretch refresh cycles, or reallocate spend toward infrastructure, storage, and hyperscaler-native tools; that is a headwind for incumbent software vendors with aging installed bases, even if they are not named in the commentary. The mainframe softness also suggests a longer lag between hardware momentum and software monetization, which is negative for IBM’s operating leverage over the next 1-3 quarters.

The near-term risk is a reflexive bounce if the slipped deals truly reclose before year-end, but that is a timing story, not a fundamental reset. The bigger 6-18 month catalyst is whether AI-assisted modernization meaningfully accelerates application replacement; if it does, IBM’s software mix can decelerate even if headline revenue holds up. I would watch for a second straight quarter of software growth below management’s original double-digit framework or any guide cut tied to transaction-processing softness; that would falsify the stabilization thesis and likely force another leg down in the stock.

Key Decisions for Investors

• Sell IBM on any post-earnings bounce over the next 1-2 sessions; structure as a 1-3 month short with a target of 8-12% downside if software growth stays in the mid-single digits and deal slippage does not fully convert.
• Pair trade: long MSFT or NOW / short IBM for 1-3 months to express the view that AI spending favors modern software platforms over legacy maintenance stacks; use a relative-performance stop if IBM outperforms the pair by ~5% on a closing basis.
• If you want convexity, buy IBM 3-6 month put spreads only on a rally back into resistance; risk/reward is better than outright puts because the maintained guide may support short-term range trading.
• Watch-item, not a trade yet: add IGV to the list of software names vulnerable to valuation compression if IBM is read as a canary for legacy app replacement; reassess if the ETF fails to reclaim its 50-day within 2-3 weeks.
• For SBUX exposure, treat the software spend reduction as anecdotal rather than material; no standalone position, but it reinforces the broader theme that aging enterprise apps are easier to trim during cost reviews.