2 Growth Stocks I Plan to Buy if the Stock Market Crashes

The S&P 500's Shiller CAPE reached 40.3 on Sept. 15, its second-highest level since the 2000 dot-com peak, leaving the market vulnerable to a correction amid rising oil prices, inflation risks, AI-development concerns, and upcoming midterm elections. The article identifies DigitalOcean as a correction-buy candidate after its AI ARR rose 212% year over year to $234 million and backlog increased 12-fold to $894 million, though its 13.2x P/S multiple exceeds its 8.6x historical average. Lemonade is also highlighted: Q2 2026 in-force premiums grew 32% to more than $1.4 billion, revenue rose 79% to $294.4 million, and its 4.3x P/S is below its three-year average of 5.3x.

Analysis

The key distinction is duration and operating leverage: DOCN and LMND would likely underperform the S&P 500 in an inflation-led drawdown because both require investors to capitalize earnings materially into the future. For DOCN, the relevant risk is not merely slower SMB spending; it is whether GPU/data-center commitments become fixed-cost capacity ahead of demand conversion, creating gross-margin pressure if utilization lags. For LMND, lower claims-handling expense is valuable only if underwriting profitability persists through adverse selection, catastrophe losses, and reinsurance renewals.

A pause in frontier-model development is not uniformly negative for DOCN. It could shift AI spending from expensive model training toward inference and application deployment, favoring lower-complexity cloud platforms and open-source model ecosystems; however, AMZN, MSFT and increasingly GPU-cloud specialists retain far greater procurement leverage and can compress pricing if SMB AI demand proves attractive. The near-term market reaction should remain dominated by real yields and oil-driven inflation, while the 1-3 month company-specific catalysts are capacity utilization, AI ARR conversion, gross-margin trajectory, and LMND's loss-ratio development.

The contrarian risk is that a broad correction does not automatically create value in either name. DOCN needs evidence that incremental AI revenue earns returns above its infrastructure cost, rather than simply monetizing scarce capacity at temporarily favorable pricing. LMND's multiple can expand on durable combined-ratio improvement, but its long-dated growth framework is not independently investable until rate adequacy, retention, and reinsurance economics validate it over multiple underwriting periods.

Key Decisions for Investors

• Set a DOCN accumulation trigger only after a valuation reset toward its historical sales multiple range and confirmation that AI backlog converts without sequential gross-margin erosion. Initial position: 50 bps; scale to 100 bps after two quarters of stable or improving utilization/margin. Falsifier: AI growth accompanied by rising capex intensity, lower gross margin, or weaker SMB net retention.
• Use a 3-6 month DOCN / short IGV pair rather than outright long exposure if real yields remain elevated: DOCN has idiosyncratic AI conversion upside, while the hedge reduces broad long-duration software beta. Enter only after DOCN underperforms IGV by another 10-15% or valuation reaches the pre-defined range; stop if DOCN's AI revenue growth decelerates sharply versus software peers.
• Place LMND on a post-results long watch, not a preemptive dip-buy list. Initiate only if quarterly gross loss ratio and customer retention remain favorable through a meaningful weather-loss period and reinsurance costs do not absorb the operating-efficiency gains; target 12-18 months, with downside defined by a reversal in loss-ratio improvement or rate-approval friction.
• Avoid treating the cited AI infrastructure demand as a direct NVDA/AMD read-through. Monitor DOCN capacity deployment and supplier disclosures for evidence of sustained GPU purchases; absent that verification, a slowdown in frontier-model capex could be more negative for NVDA/AMD than for inference-oriented application platforms.