Tech Company Blew Past Its $50 Billion AI Spending Target. Now It's Preparing to Spend Up to $95 Billion.

The software giant reportedly spent about $55.7B this year—far above what it planned—and now expects to increase AI spending further. A former Pentagon advisor warns this could signal Wall Street’s AI investments are entering a “dangerous new phase,” reflecting rising risk from overspending and escalating AI bets.

Analysis

The first-order issue is not “AI spending is high”; it is that the cost of defending share in generative AI is starting to look structurally inflationary for cash flow. For a mega-cap with fortress margins, the market can tolerate a step-up in capex for a while, but once incremental spend outruns near-term monetization, the equity multiple becomes hostage to FCF yield rather than revenue growth. That is usually when leadership shifts from the software layer to the infrastructure layer, because investors reprice the names that can actually convert AI demand into billings now.

Second-order winners are the picks-and-shovels complex: NVDA, AVGO, ANET, VRT, ETN, and power/data-center beneficiaries where order books can still outgrow supply. The risk is that the whole AI trade becomes self-financing through vendor financing and aggressive depreciation assumptions; if utilization lags, the “AI infrastructure” stack can look like a growth story with utility-like returns. That would compress multiples for the software cohort first, then eventually spill into the semiconductor names if hyperscaler capex pauses.

Timing matters: the next 1-3 months are about guidance and budget commentary, not realized earnings. If the company raises AI capex again but fails to lift revenue or margin targets, the stock should underperform on every rally, especially versus QQQ and SMH. Over 6-18 months, the key falsifier is clear evidence that AI workloads are lifting ARPU, attach rates, or gross margins enough to justify the spend; absent that, this is a classic capex arms race that ends with multiple compression, not just lower sentiment.

Key Decisions for Investors

• Short REZNF on strength or via put spreads into the next earnings/guidance update; target a 2:1 downside/upside skew if capex is revised higher again without a matching FCF outlook.
• Pair trade: long VRT/ANET basket vs short REZNF for 1-3 months; the infrastructure names monetize the spend immediately while the software side absorbs the margin drag.
• Avoid chasing the broad AI basket until there is evidence of utilization-driven ROI; if the market starts rewarding capex restraint, rotate from mega-cap software into semis/power with clearer order visibility.
• Set a watch item on REZNF free cash flow yield and capex as a percent of sales; if FCF yield falls below prior-cycle levels and guidance does not improve, extend the short for 6-12 months.
• If management demonstrates near-term AI monetization through higher operating margin or accelerated cloud/AI revenue, cover the short quickly; that would invalidate the capex-bubble thesis.