AI leaders argue software will adapt – not die – but valuations are stretched

A roughly $1 trillion rout in U.S. software names weighed on markets even as AI founders and VCs at Web Summit Qatar argued SaaS will persist through AI integration. Founders highlighted stretched valuations but expected normalization (Glean valued at $7bn; Miro at $17bn), investors warned of a deflating AI bubble, and structural signals included reports OpenAI may lose ~$14bn this year and Forbes' estimate that >$340bn chased startups in 2025 with over 65% into AI, suggesting potential valuation resets and delayed IPO activity.

Analysis

Market structure: The immediate winners are compute and platform providers (NVDA, GOOGL, AAPL) that supply GPUs, cloud infra and integrated devices; losers are high-multiple, low-differentiation SaaS and unprofitable AI pure‑plays that will face multiple compression. Expect share gains for vertically integrated suppliers (chip fabs, cloud) and pricing power on constrained GPU supply—NVIDIA/TSMC can sustain 10–30% price premiums near-term. Flow effects: equity outflows into safety lift short-term Treasury yields and USD; options vol on tech will stay elevated (+20–50% IV spikes) as headline risk rises.

Risk assessment: Tail risks include US/China export controls cutting China GPU revenues (20–40% hit to NVDA growth in worst-case), major AI regulatory action limiting data monetization, or a sudden capital pullback from venture funds forcing markdowns. Time horizons: days—volatility spikes and liquidity squeezes; weeks/months—valuation reset of 20–40% across public AI names; quarters/years—survivors consolidate 3–5x revenue growth. Hidden dependency: private AI valuations hinge on continued cheap capital; a tightening of financing costs would rapidly widen default and dilution risk.

Trade implications: Direct plays favor long NVDA and AAPL and selective long GOOGL for cloud exposure; short expensive SaaS names like CRM and unprofitable AI IPOs. Use relative-value pair trades (long NVDA vs short CRM) and structured options to buy convex upside while capping premium (9–12 month call spreads on NVDA; 3–6 month puts on CRM). Rotate from speculative AI into hardware + profitable SaaS; act on 10–20% pullbacks or post-earnings dislocations with 6–12 month horizons.

Contrarian angles: The consensus underestimates persistent ARR value—quality SaaS that embeds AI will outlast hype and re-rate once multiple compression stabilizes; a >15% dip in MSFT or CRM could be a buying window for differentiated franchises. Historical parallel: dot‑com bust vs today—difference is positive revenue and real capex demand (GPUs), so survivorship outcomes will be more concentrated but economically meaningful. Unintended consequence: a concentration into NVDA could spark regulatory scrutiny or supply bottlenecks that reverse gains quickly.

Key Decisions for Investors

• Establish a 2–3% NAV long position in NVDA with a 6–12 month horizon; implement via a 12‑month call spread (buy 450C / sell 650C) sized to equal 2% NAV and set a hard cut if NVDA falls >25% from entry.
• Add 1.5–2% NAV long exposure to GOOGL for cloud AI demand (buy shares or 9–12 month ITM calls) and add 1% NAV to AAPL on any >10% pullback; trim positions after +40% appreciation or if guidance weakens.
• Establish a 1–1.5% NAV short/hedged position against CRM: buy 3–6 month puts 10–15% OTM sized to 1% NAV (or short equivalent notional) targeting 20–40% downside in 3–9 months; cover if fundamentals (ARR growth) beat by >200bp.
• Reduce pre‑IPO/unprofitable AI private exposure by 30–50% and redeploy into hardware/cloud winners; if OpenAI/Anthropic reported FY loss exceeds $10bn or IPOs delay >6 months, increase defensive cash by 2–3% NAV.
• Hedge macro risk: shorten duration by ~25–50 bps (sell 5–10% of long-duration bond holdings) and buy 90‑day Nasdaq 100 puts equal to 1–2% portfolio notional (or a VIX call spread) to protect against a headline-driven selloff.