Wall Street Analyst Warns of AI Bubble and Urges Investors to buy SaaS Stocks. 5 Stocks to Buy if He Is Right.

Bill Gurley warned of a potential AI infrastructure bubble and recommended rotating into beaten-down SaaS stocks; the article highlights ServiceNow, Salesforce, Workday, UiPath and Adobe. The names are trading down roughly 25%–35% YTD (NOW ~25%, CRM >25%, WDAY >35%, PATH >25%, ADBE >25%) but show durable revenue/ARR growth (ServiceNow >20%, Salesforce >10% CAGR to 2030, Workday mid‑teens, Adobe low double‑digits) and valuations roughly forward P/S ~3.5–7.5x and forward P/E ~11–28x. Bottom line: a cautious buy-the-dip case for high-quality SaaS given systems-of-record defensibility and early AI traction, but risk remains from AI infrastructure froth.

Analysis

The shift from model-first AI to agentic orchestration favors software vendors that own the control plane and the system-of-record — they can reprice value from raw compute to workflow automation and capture a larger share of incremental customer spend with much lower CAC than model vendors. That dynamic amplifies switching costs: once agents, bots and rules are embedded into end-to-end processes, the marginal cost of moving to a cheaper model provider is high because orchestration, provenance and governance are the scarce assets.

A meaningful second-order effect is capital reallocation across the stack: if investor enthusiasm for bespoke AI infrastructure fades, hyperscalers and model providers will see slower topline multiple expansion while SaaS vendors with embedded agents will see faster monetization of existing cohorts. Expect shorter-term margin volatility as vendors transition to consumption/credit models (higher revenue volatility but higher LT monetization per seat), which creates a multi-quarter earnings dispersion opportunity between high-FCF, entrenched workflow vendors and asset-light infra plays.

Tail risks are concentrated and time-staggered: an overnight negative model benchmark or a marquee hallucination could crater demand for agent deployments within days, while slower adoption or pricing pushback from large enterprises plays out over quarters. The reversal trigger that hurts the SaaS rotation is not AI failing technically, but enterprise pushback on economics (e.g., customers preferring cheaper inference-as-a-service bundles) — that would collapse near-term multiple expansion even if long-term moats remain intact.

Contrarian framing: the market is underpricing the optionality from orchestration-led monetization for vendors with deep data/control planes, while simultaneously over-allocating to infrastructure facing substitute risk from orchestration and rules-based automation. That implies a barbell approach: buy durable workflow owners for asymmetric upside and employ disciplined, short-dated protection against headline risk that can wipe out near-term multiples.

Key Decisions for Investors

• Buy ADBE 12–18 month LEAP calls (or equivalent long stock exposure) sized 2–4% of equity risk budget. Entry on a <=5% pullback or immediately for thematic exposure; target 30–60% upside if consumption monetization accelerates within 12 months. Risk: premium decay and macro slowdown — put 25% of position as a stop or hedge with 6–9 month 5% OTM index puts.
• Pair trade: Long NOW / Short PATH (equal dollar, net-zero delta), 6–12 month horizon. Rationale: capture orchestration vs RPA execution mix; expect 15–30% relative outperformance if orchestration wins adoption. Risk control: close if pair relative moves >12% adverse or on next two quarterly prints that miss revenue/ARR cadence.
• Short INFA (small position, 3–6 months) or buy put spread to limit gamma — thesis: integration/monetization risk and limited incremental moat vs acquirer. Target 15–25% downside; cap max loss at 12% of notional using defined-risk spreads.
• Buy short-dated (1–3 month) market protection: S&P/tech 3–5% OTM puts sized to cover 8–12% portfolio drawdown risk. Purpose: hedge headline-driven de-risking of multiple expansion in the event of an AI infra shock or a high-profile hallucination that compresses sector multiples.