Why Tech Investors Are Reevaluating AI Investments

The article argues AI remains a positive long-term story for Big Tech, but near-to-medium-term investor expectations may shift as heavy infrastructure spending lifts depreciation costs and can limit share buybacks, contributing to slower EPS growth and lower valuation multiples. It also highlights that the risk profile differs across AI ecosystem players, from OpenAI to Anthropic.

Analysis

The market usually misprices AI capex as a pure growth story; the second-order effect is a transfer of cash flow from shareholders to vendors and infrastructure while the hyperscalers’ own per-share economics deteriorate. Higher depreciation is especially toxic because it shows up with a lag, so reported EPS can look fine for a few quarters even as economic return on capital falls and buybacks become less elastic. That combination tends to compress multiples first in the names with the heaviest capex intensity and the least visible monetization.

Relative winners are the picks-and-shovels layer: GPU suppliers, networking, power/cooling, and some software names that monetize inference usage without absorbing the capex burden. Relative losers are the platforms funding the buildout, especially where AI monetization is still experimental and capital returns are the main defense of the stock. Over 1-3 months, the key catalyst is guidance: any upward revision to AI infrastructure budgets without a matching revenue bridge should widen the gap between the infrastructure basket and the hyperscaler basket.

The contrarian miss is that investors are still treating buybacks as durable even though they are the easiest lever to crowd out when management prioritizes compute. If utilization or pricing for AI services disappoints, the market will re-rate these stocks as utility-like capex businesses rather than software-like compounding machines. Falsifiers are simple: slower capex growth, improving AI monetization per dollar of spend, or a clear re-acceleration in buybacks; absent that, the multiple risk is down over 6-18 months.

Key Decisions for Investors

• Enter a relative-value pair: long NVDA/AVGO vs short a basket of MSFT, GOOGL, META, and AMZN into the next earnings/guidance cycle; target 3-6 months, with the trade working if hyperscaler capex keeps rising faster than AI revenue and buybacks stay muted.
• For more defensive exposure, rotate part of QQQ exposure into SMH or SOXX rather than owning the mega-cap platform names outright; this keeps AI upside while reducing direct exposure to depreciation and buyback crowd-out at the platform level.
• Use a trigger-based short on the hyperscaler basket only if forward capex guides rise again without a commensurate increase in operating income or free cash flow; that setup implies 10-15% multiple compression risk over 6-12 months.
• Watch for buyback authorization commentary in the next 1-2 quarters: if repurchases accelerate despite AI spending, cover any short exposure because the market will likely treat that as a signal of monetization confidence.
• If you want a cleaner contrarian long, prefer AAPL over the AI capex-heavy cohort on valuation discipline alone; it benefits from capital return scarcity if the market starts punishing D&A-heavy EPS growth.