The Market's Most Feared Month Is Here. These 3 AI Stocks Have Survived It Before.

Oracle shares surged 36% in one session on Sept. 10, 2025 despite an earnings miss, as remaining performance obligations jumped 359% to $455B—though the article flags $122B of debt and customer concentration risk within its $638B backlog. Broadcom rallied ~20% after results (revenue +22% to a record $16B; AI chip revenue +63% to $5.2B; FCF +47% to $7B) on a new custom AI chip customer identified as Anthropic, while Micron is highlighted for supply-constrained HBM pricing power. Overall, the piece argues September’s outsized moves come from contract/customer/pricing disclosures rather than AI narrative, suggesting modestly positive near-term read-through for AI-linked semis.

Analysis

The key distinction is between AI exposure and AI monetization quality. AVGO looks like the cleanest beneficiary because custom silicon turns hyperscaler capex into higher-margin content and can shift bargaining power away from merchant GPU suppliers over time; if inference workloads keep migrating toward ASICs, the second-order effect is slower share gains for NVDA in parts of the data-center stack. MU is also advantaged, but its upside is more tactical: HBM pricing power can add sharply to earnings for 1-3 quarters, then unwind quickly if capacity comes online or hyperscaler budgets flatten.

ORCL is a different animal: once the market focuses on backlog and not near-term earnings, the equity starts trading like leveraged long-duration credit. That creates a fragile setup where any delay in conversion, customer financing stress, or higher balance-sheet burden can compress the multiple faster than fundamentals improve. The falsifier is not one missed quarter; it is evidence that contracted demand is not translating into cash flow or that leverage remains sticky despite the growth narrative.

Consensus is probably underestimating how narrow this AI tape has become. The winners this September are not the loudest AI names, but the ones with verifiable pricing power, signed demand, or clear design wins; that is why the trade is more idiosyncratic than index-level. If hyperscaler capex decelerates from here, AVGO and MU can still work because they have observable unit economics, while ORCL is the name most likely to re-rate down as the market demands execution rather than story.

Key Decisions for Investors

• Long AVGO on 3-5% pullbacks into the next print; use a 1-3 month horizon and look for 8-12% upside if AI revenue guidance and customer commentary stay constructive. Exit if the new-customer cadence stalls or AI growth is presented as one-off rather than recurring.
• Pair trade: long MU / short NVDA into MU earnings. The edge is that HBM pricing power is more immediately monetizable than GPU scarcity, while NVDA already embeds a high bar for flawless demand; this should work best over the next 4-8 weeks. Falsify if HBM supply normalizes faster than expected or MU guide implies peak margins.
• Short ORCL or buy put spreads into the next earnings window. The thesis is that the stock is now a balance-sheet-and-conversion story, so any sign of slower backlog monetization or lingering leverage can drive downside over 1-2 quarters. Cover if backlog growth is matched by clearly improving free cash flow and debt reduction.
• Do not add broad AI-beta exposure until the next round of prints. Prefer names with contract visibility and pricing power over narrative-only exposure; the market is likely to punish anything that cannot show signed demand converting into revenue.