5 big analyst AI moves: Palo Alto and Reddit get fresh Buy calls, AMD upgraded

Analysts turned constructive on several AI-linked tech names, led by AMD’s upgrade to Buy with its price target raised to $375 from $220 on a “structural increase in CPU demand.” Intel’s strong Q1 print and 22% YoY Data Center and AI growth reinforced the agentic-AI CPU trade, while Arista was added to Evercore’s Tactical Outperform list ahead of earnings on expectations for a beat-and-raise. Palo Alto and Reddit also received Buy initiations, reflecting upbeat views on AI-driven cybersecurity demand and monetization opportunities.

Analysis

The common thread is not “AI spending” but a re-acceleration in the non-GPU layer of the stack: server CPUs, networking, and security controls are all benefiting as agentic workloads increase orchestration, memory access, identity checks, and east-west traffic. That matters because it broadens the AI capex wallet beyond a single winner; the second-order effect is that the supply chain shifts from pure accelerator scarcity toward a more balanced mix where CPU vendors, high-speed networking vendors, and platform security suites all gain pricing power. In that regime, the biggest underappreciated winner is likely Intel/AMD CPU attach rather than the obvious GPU names, because each agentic deployment needs more general-purpose compute per inference/workflow than pretraining did.

Intel’s surprise strength is the cleanest tell that the market may be early in a multi-quarter CPU upswing, not just a one-off quarter. If management and customers are already committing into 2027, then the earnings power inflection is less about unit share and more about mix, pricing, and packaging constraints — which can keep margins improving even before foundry becomes a true profit engine. The risk is that consensus extrapolates this too aggressively: if AI deployment pauses or customers optimize workloads more efficiently, the CPU read-through could normalize quickly and the multiple expansion in semis would be vulnerable.

Palo Alto and Reddit are both “AI beneficiaries,” but for different reasons: PANW is monetizing the security surface area created by AI adoption, while RDDT is a data asset with pricing power if licensing is renegotiated from a position of scarcity. The market seems to be pricing these as cyclical software stories, but the more durable angle is data/control-point ownership: the companies that sit closest to identity, telemetry, and human-generated text should command better economics as AI agents proliferate. The contrarian risk is that the market may be overestimating how quickly AI expands TAM and underestimating procurement friction, especially if enterprise buyers consolidate spend more slowly than bulls expect.

Key Decisions for Investors

• Go long AMD into the May 5 print via stock or call spreads; the setup is a 1-2 quarter estimates upgrade cycle if CPU attach really is inflecting, but size modestly because the move is already partially de-risked by the Intel read-through.
• Add to INTC on pullbacks over the next 2-6 weeks; best risk/reward is as a momentum/estimate-revision trade, with the risk that the market has already discounted too much of the 2027 upside into the stock.
• Pair long PANW / short high-multiple software names with weaker data moats for a 3-6 month hold; if AI is raising the security budget rather than shrinking it, platform vendors should outperform the broader software complex.
• Speculative long RDDT into earnings and the next licensing negotiation window; the trade works if investors underwrite higher data-license pricing in 2027, but cut quickly if traffic softness persists beyond one quarter.
• Relative-value long ANET vs. broader AI infrastructure basket if cluster speeds continue migrating to 1.6T; the trade benefits from networking becoming a larger percentage of AI capex as compute architectures become more distributed.