I’m a VC who bets on AI. What keeps me up at night isn’t the idea of these companies failing—quite the opposite

Software stocks have lost over $1 trillion in market cap since the start of the year amid fears AI agents can replace multiple SaaS subscriptions. The author warns VCs that AI could commoditize current portfolio companies and advises treating AI as a horizontal layer, investing in firms that own customer trust/distribution, are embedded where real money moves, and accumulate proprietary data. The piece recommends diversified bets across verticals such as fraud/compliance and highlights emerging markets—particularly Africa—and companies like Sabi, Smile ID and Orca as opportunities with defensible, hard-to-replicate regional data and trust.

Analysis

Treat AI as an accelerant to platform consolidation rather than a pure-product winner-takes-all story. Over the next 12–24 months expect enterprise buyers to reallocate spend toward vendors that can stitch AI into workflow endpoints they already trust; that reallocation can shave 200–400bps off growth rates for mid‑market pure‑SaaS vendors and will show up first in higher churn and lower net retention numbers.

The most durable economic moats will be transactional rails and regional orchestration layers — not model IP. Firms that own payment flows, settlement timing, identity verification, or cross‑border logistics capture telemetry that compounds into asymmetric ML features and optionality (e.g., financing, insurance, dynamic pricing). This favors large cloud/platform integrators and entrenched payments/security vendors while putting pressure on single‑product apps with low switching costs.

Emerging markets provide an outsized asymmetric return profile because fragmentation creates switching friction and regulatory frictions that raise the cost for global incumbents to replicate incumbents’ data sets. Expect winners in these markets to convert 10–30% of adjacent service economics (financing, storage, transport coordination) over 2–5 years, but also price in idiosyncratic political and FX tail risk.

Key catalysts to watch in the next 6–18 months are (a) bundling announcements from hyperscalers, (b) enterprise procurement changes reflecting AI‑driven workflow pilots, and (c) quarterly NRR/churn inflection in public SaaS names. Reversals arrive if open‑source agents commoditize orchestration layers or if interest‑rate normalization restores valuation premium to subscription growth, which would materially reduce the opportunity to short structural SaaS fragility.

Key Decisions for Investors

• Long MSFT (12–18 months): implement a modest call‑spread to express a sustained re‑rating if bundling accelerates enterprise wallet capture. Target a 1–1.5x asymmetric payoff (pay small premium vs uncapped upside) and size to 1–2% of liquid portfolio; hedge with a small short on high‑beta software exposure.
• Short IGV (iShares Expanded Tech‑Software ETF) via 6–9 month put spread: play continued multiple compression in narrow SaaS stories as buyers consolidate. Keep max loss defined by sold puts and size cumulative exposure to no more than 2–3% notional — reward/risk ~2:1 if multiple pressure continues.
• Pair trade — Long GPN (Global Payments) / Short PYPL (PayPal) (6–12 months): favor companies that monetize settlement and platform fees over consumer wallet plays likely to face margin pressure from embedded fintech in platforms. Size as a dollar‑neutral pair; target 15–25% relative return if fee capture increases.
• Long AFK (VanEck Africa ETF) or targeted EM fintech exposure (2–5 years): allocate a small satellite position (0.5–1% portfolio) to capture long‑term option value from regional orchestration winners. Concentration risk and FX volatility warrant a patient time horizon; expect multi‑year path with >2x upside in base case but high dispersion.
• Long ZS (Zscaler) or another cloud‑native security vendor (12 months): buy into names that can leverage transaction and telemetry data to expand into fraud/agent protection layers. Use a 12–18 month horizon with tight stop (15–20%) — reward dependent on enterprise security re‑arming against automated agent risks.