
The article is bullish on AMD and Broadcom, arguing both could see 50%+ upside over the next year as AI spending shifts toward inference and custom chips. AMD is highlighted for its inference/agentic AI opportunity, with cited potential for $175 billion in revenue and about $50 in EPS by 2030, while Broadcom is said to have a path to more than $100 billion in custom chip revenue in fiscal 2027 and $180 billion in AI sales by 2028. The piece is opinion-driven rather than event-driven, so it is more likely to influence investor sentiment than trigger an immediate price move.
The key second-order trade is not just "AI demand up," but a mix shift from training-heavy capex to inference-heavy deployment, which changes the margin pool across the stack. That favors vendors with memory-dense architectures, strong CPU attach, and custom silicon capabilities, while making pure training exposure less differentiated over the next 6-18 months. If hyperscalers keep pushing for supplier diversification, pricing power migrates from the dominant incumbent toward a duopoly-plus-custom model, expanding total TAM but compressing the scarcity premium on any single name. AMD’s upside depends less on headline GPU units and more on whether it can convert inference workloads into a durable CPU + accelerator bundle. The market may be underestimating how much agentic AI increases host-side CPU demand, which could lift AMD’s data center mix even if GPU share gains are incremental rather than explosive. The real risk is not demand—it is execution and supply allocation; if ramp timing slips by even two quarters, the stock can derate quickly because expectations are already discounting a steep slope. Broadcom is the cleaner structural beneficiary because custom ASICs are a board-level cost-reduction decision, not a speculative GPU swap. The second-order effect is that every successful custom design deepens networking demand and creates a stickier software/services pull-through, so the AI franchise can compound beyond chip revenue alone. The main risk is concentration: a few hyperscaler wins can look extraordinary until a design cycle pauses, so investors should expect lumpy revenue recognition and avoid extrapolating straight-line growth. The consensus may be overconfident in the pace of monetization and underpricing competitive responses from Nvidia, which can defend share via software ecosystem and rapid platform iteration. The more interesting trade is not long-only beta to AI, but relative value between names that benefit from inference mix shift versus those still priced for training dominance. Over the next 3-9 months, this should show up first in guidance quality and gross margin inflection, not necessarily in near-term unit growth.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Request DemoOverall Sentiment
strongly positive
Sentiment Score
0.72
Ticker Sentiment