Is Nvidia, at Bargain Levels Right Now, a Buy? Here Are the Bull and Bear Cases.

Nvidia reported $215 billion in revenue for its last fiscal year, up 700% over five years, and forecasts 70% growth in the next fiscal year; it also projects $20 billion in CPU sales this year as it enters the market for chips supporting agentic AI. The article favors the bull case, citing strong AI demand and a forward P/E of 25x, while flagging that data centers accounted for 92% of recent-quarter revenue and that China’s preference for domestic chips may constrain sales.

Analysis

The CPU push matters less as a standalone chip opportunity than as a bid to own more of the server bill of materials: a tightly integrated CPU-GPU system could raise platform switching costs and make it harder for cloud customers to substitute merchant GPUs. That would pressure Intel and AMD if customers adopt the bundle, but could also intensify hyperscaler efforts to build in-house silicon and negotiate down supplier pricing. The key uncertainty is incrementality: reported CPU revenue could partly reflect system-level sales or a shift in what Nvidia supplies, rather than a comparable amount of new customer spending. Verify customer commitments, shipment timing, and whether CPU contribution expands total platform economics before underwriting the growth estimate.

Near term, strong infrastructure budgets support sentiment, but NVDA remains exposed to any capex pause or evidence that AI deployments are not earning acceptable returns. Over 1–3 months, guidance, orders, and margins are more informative than broad AI-spending intentions. Over 6–18 months, custom chips and China’s preference for domestic suppliers are structural constraints; CPU integration is a potential offset, not proof of durable pricing power. The contrarian risk is treating a low-looking forward multiple as protection: estimates can reset quickly if spending slows. Falsify the constructive view on material data-center growth deceleration, weaker margin/guidance trends, or CPU plans failing to convert into disclosed customer demand.

Key Decisions for Investors

• Do not add solely on the article’s valuation framing. If seeking exposure, scale into NVDA rather than chase, and reassess after the next results and guidance update.
• Treat the CPU revenue figure as an unverified catalyst: monitor disclosures for named customer adoption, shipment schedule, and evidence that CPU sales are incremental to existing platform revenue.
• Set a thesis review trigger if data-center growth decelerates materially or margins/guidance weaken; those would indicate capex digestion is affecting earnings, not just sentiment.
• Watch Intel and AMD for competitive read-through, and hyperscaler custom-silicon plans for bargaining-power risk; avoid a pair trade until adoption and share impact are measurable.