The companies getting the most from AI are rethinking how work gets done

McKinsey survey data shows heavy AI adoption—nearly 9 in 10 firms use AI for at least one function and 44% are scaling enterprise-wide—but only 37% report meaningful EBIT impact, unchanged from last year. AI is consuming more budget (28% say AI is >10% of IT spend; 60% plan to increase next year) and 20% cite token/AI operating costs as constraining use. “AI high performers” (6% attributing ≥5% of EBIT to AI) are more likely to redesign workflows (nearly 75% vs 55% last year), implying returns depend on process change, not just spend.

Analysis

The market is still pricing AI as a growth engine, but the next leg is likely a margin question, not an adoption question. Near term, the cleanest beneficiaries are the toll collectors on compute, cloud, and implementation: semis, hyperscalers, and systems integrators that get paid before customers prove ROI. That argues for relative strength in NVDA, MSFT, AMZN, GOOGL, ORCL, and consultancies like ACN/IBM, while AI-featured software with lofty multiples faces tougher scrutiny if the promised EBIT lift stays absent.

The second-order effect is a budget squeeze inside corporate IT: if AI is already consuming a double-digit share of spend, CFOs will demand workflow redesign, vendor consolidation, and usage-based pricing. That tends to widen the gap between large enterprises and smaller firms because the former can amortize transformation costs across bigger revenue bases and existing data/process depth. Over 1-3 months, this can compress multiples for high-duration SaaS; over 6-18 months, the winners are the firms that can actually replace labor or redesign processes rather than simply add a chatbot layer.

Contrarian view: the consensus is overestimating how quickly AI converts into operating leverage for end users and underestimating how much value accrues to infrastructure and services. The risk to the bearish thesis is real if upcoming earnings show AI attaching to core workflows and not just pilots; that would force a re-rate in software names. Falsifier to watch: a clear uptick in AI-driven gross margin or EBIT commentary across enterprise software over the next two reporting cycles, versus continued capex growth without profitability inflection.

Key Decisions for Investors

• Relative value: long SMH or NVDA / short IGV for the next 1-3 months. Thesis is that spending continues to flow to compute while monetization skepticism hits app-layer software; target 2:1 if the market keeps rewarding capex beneficiaries and punishing names that cannot show EBIT conversion.
• Pair trade: long MSFT or AMZN vs short a basket of high-multiple SaaS names most exposed to AI valuation premium (e.g. CRM, SNOW, DDOG) into earnings season. Use this as a tactical expression of 'AI spend without profit' risk; stop if those names show explicit margin expansion from AI.
• Selective long ACN and IBM on pullbacks only if management commentary supports workflow redesign and implementation backlog. These are the clearest second-order winners if enterprises move from experimentation to transformation; otherwise treat as watchlist, not a conviction buy.
• Underweight smaller enterprise software vendors that depend on AI attach rates rather than hard ROI. The mismatch between adoption and EBIT is a setup for multiple compression over 6-12 months if customers start forcing vendor rationalization and lower token/usage spend.
• Set an alert for the next two quarterly cycles: if AI spend keeps rising but no broad EBIT inflection appears, add to short exposure in software and reduce exposure to names trading primarily on AI narrative rather than current cash flow.