Leading bank calls AI threat to enterprise software "illogical," upgrades five stocks for 2026

HSBC's February research argues AI will primarily create long-term value for enterprise software rather than replace it, forecasting 2026 as a shift from infrastructure build-out to monetisation via workflow-embedded AI agents and rising inference demand. The bank upgraded CrowdStrike to 'buy', maintains 'buy' ratings on Oracle, Microsoft, ServiceNow, Salesforce and Palantir, and downgraded IBM and Asana while keeping 'reduce' ratings on CoreWeave and Palo Alto Networks; it cites persistent GPU scarcity, enterprise switching costs, and the need for proven uptime as defensive moats for incumbents. For investors, the note implies a sector tilt toward established software platforms and select cybersecurity names that can monetise agent-driven AI, while hardware and niche infrastructure plays face structural headwinds.

Analysis

Market structure: HSBC’s thesis tilts value capture toward enterprise software (ORCL, MSFT, NOW, CRM, CRWD) as AI monetisation moves from infrastructure to agentic, workflow-embedded products in 2026. GPU/data‑centre scarcity implies sustained premium for compute suppliers (NVDA/AMD) and higher capex for hyperscalers, but the lion’s share of recurring revenue and margin expansion will accrue to software platforms that embed agents and capture per-seat or per-workflow pricing.

Risk assessment: Key tail risks are a rapid open‑source LLM that commoditises APIs (3–12 months), a sudden GPU oversupply that collapses infra pricing (6–18 months), or regulatory limits on training/use of private data (EU/US actions within 60–180 days). Hidden dependencies include systems integrators, implementation risk and enterprise governance — failure rates on large rollouts could delay revenue capture by 1–3 quarters; catalysts that accelerate adoption are major Fortune 100 agent deployments and quarterly beats showing inference > training spend.

Trade implications: Favor concentrated longs in software franchises with enterprise moats: establish 2–3% positions in CRWD and ORCL and 1–2% in MSFT/NOW for 6–18 month horizons; hedge beta by trimming pure‑play infra exposure (CRWV, PANW) and consider 1–2% short exposures where earnings leverage is absent (ASAN, IBM). Use 3–9 month call spreads on CRWD/ORCL to cap capital and buy 6–9 month 10–20% OTM call spreads on NVDA as a levered play on sustained GPU tightness.

Contrarian angles: Consensus may underprice the pace of open‑source model improvements and enterprise willingness to adopt non‑proprietary agents — this would compress software multiples and boost systems integrators. Conversely, market may already overvalue software winners; require concrete agent P&L line items in next 2 quarters before adding large convex long positions. Historical parallel: internet era showed multi‑year lag from infra investment to software monetisation — expect 12–36 months for durable revenue recognition.

Key Decisions for Investors

• Establish a 2–3% long position in CRWD (CrowdStrike) within 2–4 weeks; target 6–12 month horizon, take profits at +30% and set stop-loss at -12%; hedge with 0.5% portfolio put protection if portfolio tech beta >1.2.
• Buy a 2% long position in ORCL (Oracle) and a 1% long in NOW (ServiceNow) as 6–18 month core AI monetisation plays; consider 6–9 month call spreads (10–20% OTM) to leverage upside while limiting cost; trim if quarterly licence/ARR growth misses by >100bps.
• Establish 1–2% short exposures to CRWV (CoreWeave) and PANW (Palo Alto) over the next month, size each at 0.5–1% of portfolio; unwind if either reports >20% QoQ improvement in revenue-per-GPU or secures multi‑year enterprise contracts covering >$100m ARR.
• Buy 6–9 month NVDA call spreads (10–20% OTM) sized to 0.5–1% portfolio to capture continued GPU scarcity; if GPU spot supply indicators (freight/ASM/AP orders) show >30% QoQ increase, reduce NVDA allocation by 50%.
• Initiate a pair trade: long ORCL (2%) vs short ASAN (1%) for 6–12 months to capture moat-based re‑rating; exit if ORCL misses guide by >150bps or ASAN posts >25% YoY ARR growth, reassess positions then.