From Karl Marx’s vampire to Warren Buffett’s moat, the crumbling gothic castle has been capitalism’s favorite metaphor for 250 years

A two-day selloff exposed investor concerns that AI may weaken established tech companies’ competitive moats: Alphabet fell 12.6% after search revenue missed expectations and free cash flow was negative $9.1 billion versus an expected negative $2.23 billion, while ServiceNow dropped 27% and Intel was down roughly 25% premarket. Big Tech lost nearly $900 billion in market value as AI spending accelerated, and hyperscalers raised $326 billion in stock and bonds this year; the article also notes falling AI token prices and limited reported financial returns from AI outside tech.

Analysis

The useful distinction is not “AI winners versus losers,” but who can convert falling inference costs into durable gross profit. For model vendors, lower prices help adoption but weaken economics unless unit costs fall faster; token growth alone is not proof of a moat. For incumbents, the key risk is monetization substitution: AI can move activity into a less valuable channel before advertising or subscription models adapt.

The pressure is likely greatest for seat-priced software where agents can reduce paid user counts. That is not automatically a thesis against ServiceNow (NOW): systems of record and embedded workflows may retain bargaining power even if the interface changes. The key evidence is renewal pricing, net retention, and whether pricing shifts from seats to successful workflows. Reddit (RDDT) faces a different trade-off: restricting anonymous access may strengthen first-party data and ad targeting, but could also reduce discovery and audience growth. Track logged-out reach and monetization per user, not the policy change alone.

Over 1–3 months, earnings guidance and disclosed AI revenue relative to capex matter more than headline spend. Over 6–18 months, refinancing pressure in software credit could constrain private-equity-backed competitors and prompt consolidation; lender-level exposure is not established here, so this is not yet a JPMorgan (JPM) short thesis. Contrarian point: the selloff may overgeneralize from exposed interfaces to durable workflow assets, while underestimating the risk that infrastructure spending earns inadequate returns.

Key Decisions for Investors

• Avoid a broad short of technology on this evidence alone. Consider a beta-controlled long AMZN / short NOW pair only after verifying AWS growth and capex monetization against NOW renewal pricing and net retention; enter after post-earnings volatility settles. The thesis is falsified by AWS demand or returns weakening while NOW sustains pricing and retention.
• Treat GOOG as a monetization watch, not simply an AI-displacement short: require evidence that AI-assisted search can preserve revenue per query and free cash flow. A deterioration in both would strengthen the underweight; improving monetization would invalidate it.
• Monitor RDDT’s logged-in strategy through audience reach, ad monetization per user, and any disclosed data-licensing economics. Do not assume the access change is value-accretive until those indicators confirm it.
• For software credit, seek borrower- and lender-level exposure, maturity schedules, and covenant data before expressing a bank or private-credit trade; the article does not establish JPMorgan’s exposure.