Information Technology: Creative Destruction on Speed

Intel and AI-linked semiconductors are benefiting from a stronger AI-capex backdrop, with S&P 500 Semiconductors 2026 earnings growth revised up to 86.5% from 65.0% at the start of the year. The group now trades at a 20.9 forward P/E, slightly below the S&P 500’s 21.1, despite revenue growth forecasts of 55.7% for 2026 and 34.0% for 2027. By contrast, application software has de-rated to a 23.4 forward P/E, the lowest since 2014, as investors price in AI disruption.

Analysis

The key market implication is not simply that chips are winning; it’s that AI capex is increasingly self-funding through the hyperscalers’ own equity and cash flows. That changes the distribution of outcomes inside tech: capital is likely to keep rotating from low-beta software monetization stories into the infrastructure layer where pricing power, utilization, and backlog visibility are still improving. In the near term, that supports a higher multiple floor for compute, networking, and memory, while pressuring anything whose value proposition can be substituted by copilots or agentic workflows.

The second-order effect is that the market is underpricing the duration of this cycle. When semis trade near the broad index multiple despite sharply superior growth, the real risk is not fundamentals rolling over tomorrow but investors chasing revised estimates too late, after the re-rating has already happened. That argues for being long the segments where estimate revisions are still accelerating, rather than owning the crowded consensus winners that have already de-risked via multiple expansion.

On the other side, application software likely faces a longer, uglier revaluation path than most expect. The winner-take-most dynamic in AI means budgets can shift from seat-based software spend toward model access, infrastructure, and workflow consolidation, which compresses vendor pricing power before revenue growth visibly breaks. The contrarian mistake would be to assume lower multiples in software automatically mean value; in several verticals, the cheapest names may still be value traps if renewal economics and feature displacement deteriorate faster than expected.

The most attractive setup is to express the view through relative value, not outright beta. The best risk/reward is to own the AI compute stack while fading software breadth, because the market is still rewarding revenue growth that is likely to become less durable. The main reversal risk is a capex pause from hyperscalers or a brutal semis inventory correction, but that looks more like a months-long digestion process than an immediate catalyst unless AI deployment data stalls abruptly.

Key Decisions for Investors

• Go long AMZN and GOOGL over a 1-3 month horizon as the cleanest large-cap proxies for continued AI capex monetization; risk/reward is favorable because their own cash generation can fund the cycle, but trim if cloud growth decelerates meaningfully.
• Add a tactical long in TSLA for 3-6 months only as a secondary AI beneficiary; this is higher beta and more speculative, so size smaller and use upside to finance via call spreads rather than cash equity.
• Initiate a long semis / short application software relative-value pair for 2-4 months: long SMH or best-in-class compute suppliers, short a basket of high-multiple software names most exposed to workflow substitution; this is the cleanest expression of the AI displacement trade.
• Avoid chasing valuation in legacy IT hardware such as IBM; if you want AI exposure, prefer names with accelerating estimate revisions and end-market leverage rather than statistical cheapness.
• Use pullbacks to accumulate AI infrastructure exposure, but set a hard risk stop if hyperscaler capex commentary turns cautious for two consecutive quarters; that would be the earliest signal the market is too far ahead of actual monetization.