If an AI Bubble Crashes the Stock Market, Warren Buffett Says This Is the First Move Investors Should Make

The S&P 500 has risen more than 88% and the Nasdaq Composite 111% over the past three years, driven substantially by AI-related enthusiasm, but Goldman Sachs estimates AI users would need to spend at least $1 trillion annually to support hyperscalers' returns on investment. Drawing on Warren Buffett's dot-com-era warning, the article advises investors to prioritize durable competitive advantages rather than technology hype and to maintain a long-term buying approach. It notes that the S&P 500 has returned more than 700% since the March 2000 dot-com peak and that every rolling 20-year period since 1919 has delivered positive total returns.

Analysis

This is not a discrete fundamental catalyst; its investable value is as a positioning reminder. The vulnerability is concentrated in AI infrastructure names whose valuation assumes a rapid transition from capex-led demand to broad enterprise monetization. NVDA remains less exposed than many second-derivative beneficiaries because its near-term revenue is backed by hyperscaler procurement, but a slowing order cadence or lower cloud capex guidance would compress the entire AI hardware multiple before it materially impairs reported earnings.

Over the next 1-3 months, monitor MSFT, AMZN, GOOGL and META capex commentary, AI revenue disclosures, and NVDA lead times/order visibility. The key bear signal is not elevated capex alone, but evidence that inference revenue and enterprise software adoption are failing to offset depreciation and power costs; that would shift investor focus from growth to return-on-invested-capital and hurt high-multiple infrastructure suppliers most. Conversely, sustained hyperscaler capex revisions higher would invalidate a near-term short thesis.

BRK.A is a relative beneficiary if volatility creates forced selling in expensive growth: its liquidity, recurring insurance earnings, and ability to deploy capital make it a lower-beta quality compounder rather than a direct AI hedge. The contrarian point is that broad AI skepticism should not automatically translate into shorting NVDA: a valuation reset in speculative software, networking, and unprofitable data-center adjacencies can coexist with resilient demand for the dominant compute platform.

Key Decisions for Investors

• Do not initiate a broad AI short solely on this commentary; treat it as a risk-management signal rather than a new information event. Reduce crowded, low-quality AI exposure where revenue is not independently tied to signed backlog or recurring usage metrics.
• For a 3-6 month defensive rotation, pair long BRK.B against a basket of high-multiple, cash-burning AI software/infrastructure names rather than short NVDA outright. Target 2:1 upside/downside; exit if hyperscaler aggregate 2027 capex guidance rises by more than 10% while AI monetization disclosures improve.
• Maintain NVDA core exposure only with defined downside protection: buy 3-6 month put spreads financed by selling upside calls after sharp rallies, or trim if forward revenue estimates stop rising following hyperscaler earnings. A material decline in disclosed cloud capex plans is the thesis trigger for a larger underweight.
• Set an earnings-season watch item for MSFT, AMZN, GOOGL, and META: quantify incremental AI revenue versus incremental depreciation, power, and capex. If monetization remains opaque through two reporting cycles, favor quality cash generators such as BRK.B over AI-beta exposure for the following 6-18 months.