Barclays reveals the top AI software stocks to buy now - Do you own any of them?

Barclays named Oracle, DigitalOcean, Salesforce, and Snowflake as its top U.S. software picks, citing AI infrastructure, cloud demand, and improving growth profiles. Oracle stands out with backlog swelling from $98 billion in fiscal 2024 to over $550 billion in fiscal 2026, implying revenue growth accelerating from about 9% to over 30% by fiscal 2027. DigitalOcean also posted a Q1 2026 beat with revenue of $258 million and EPS of $0.44, while recent analyst target changes and partnerships reinforce the sector’s AI-driven narrative.

Analysis

The near-term winner is not just the named software leaders, but the broader AI infrastructure stack that sits behind them: GPU clouds, network gear, data-center power, and implementation services all get a demand validation signal when large software platforms show multi-year backlog visibility. Oracle’s backlog step-up implies capacity is now the bottleneck, so the second-order winners are suppliers that can relieve that constraint over the next 6-18 months; any delay in rack buildout, networking, or power interconnects could push revenue recognition right, but it will not likely cancel the spend.

The market is likely underestimating how much of this is a capacity and execution story rather than pure product adoption. For Oracle and DigitalOcean, the key risk is not demand saturation but timing slippage: if contracted AI capacity ramps slower than expected, the revenue inflection gets pushed into the next fiscal year and the stocks can de-rate even with intact fundamentals. For Snowflake and Salesforce, the opportunity is more optionality than current monetization; if enterprise AI use cases stay analytics- and workflow-centric, these names can compound for years, but if AI remains a cost-cutting feature embedded in broader suites, margin upside may be more muted than bulls expect.

The contrarian angle is that this set-up favors a barbell: the market is crowded into the obvious AI infra winners, while the less obvious beneficiaries are the enablers of deployment at scale — data management, orchestration, and enterprise workflow tools. The miss is assuming all AI winners need to be model builders or cloud hosts; in practice, the first durable monetization often accrues to vendors that reduce implementation friction and improve data quality, which argues for relative strength in the picks with sticky enterprise penetration rather than the purest headline growth rates.

Short term, the main reversal catalyst would be any evidence that AI capex is front-loaded and not translating into faster bookings or utilization; that would hit the highest-multiple names first over a 1-3 month horizon. Over 6-12 months, if capacity expansion catches up and backlog converts cleanly, the setup should broaden beyond ORCL into DOCN and the data layer, while CRM remains a later-cycle beneficiary as agents move from pilots to workflow automation.

Key Decisions for Investors

• Overweight ORCL vs. long-duration software peers for the next 3-6 months; the backlog visibility lowers fundamental risk, but keep stop discipline because the stock is now trading against execution risk on capacity ramp.
• Initiate a long DOCN / short higher-beta unprofitable AI infrastructure basket pair over 1-2 quarters; DOCN has direct capacity leverage and should benefit if small-customer AI demand remains supply-constrained.
• Buy SNOW on pullbacks for a 6-12 month horizon; use it as a relative winner if enterprise AI spending shifts from model hype to data governance and analytics, with upside tied to higher attachment of AI/data engineering modules.
• Structure a CRM call spread 6-9 months out; the setup is less about current growth and more about agent-driven workflow monetization, which could re-rate the stock if enterprise adoption moves from pilot to production.
• Avoid chasing the crowded first-order AI infra complex at elevated multiples; wait for evidence of utilization conversion before adding size, since the main downside is a timing miss rather than a thesis break.