A Famed Billionaire Tech Investor Just Sold Nvidia Shares to Buy These 2 AI Chip Stocks

The article argues Nvidia is a relative bargain, trading at a forward P/E of 17x for fiscal 2028 (ending Jan 2028), supported by its dominant AI training GPUs, CUDA moat, and expanding inference/agentic AI stack. It highlights Tiger Global’s Q2 activity—cutting Nvidia stake by ~7% while adding Cerebras and increasing Intel—as a signal of continued AI capex/inference opportunity (with Cerebras positioned as up to ~6x faster for inference vs LPUs). Intel is framed more cautiously: while it remains a server CPU market share leader, it is losing share to AMD/Arm and has foundry PC cost headwinds, and the stock is described as “no longer cheap,” suggesting investors stay sidelined.

Analysis

The more important signal is not the names themselves but the shift in how AI capex gets allocated. If inference keeps taking share from training, value migrates away from “single-chip hero” narratives toward stack owners with software, networking, and deployment control; that favors the highest-quality platform names and punishes vendors relying on one architecture or one end-market to justify their multiple.

NVDA still looks like the cleanest way to own that transition because it monetizes the full system, not just compute. The risk over the next 1-3 months is not valuation in isolation but whether hyperscalers keep pushing custom silicon and procurement diversification; if that happens, the stock can still compound, but the pace of multiple expansion likely slows. A break in AI capex cadence or a guide-down on networking attach would be the clearest falsifier.

CBRS is a different trade: it is a high-beta call option on latency-sensitive inference workloads, but it carries execution and concentration risk that public-market enthusiasm may underprice. The market may be extrapolating a few design wins into a durable platform story; the key question over 6-18 months is whether the company can move from bespoke deployments to repeatable volume without margin dilution. AMZN is a secondary beneficiary because cheaper inference expands cloud usage and improves the economics of AI services, while INTC’s near-term AI lift still looks more like cyclical share capture than a durable moat.

The contrarian view is that consensus may be overestimating how quickly inference becomes a broad, standardized market. If low-latency inference remains premium and fragmented, NVDA keeps the pricing power; if it commoditizes faster than expected, the economic winner shifts to the cloud vendors and application layer, not the chip vendors alone.

Key Decisions for Investors

• Long NVDA / short INTC for 3-6 months: favor the platform with durable software and networking attach over a stock whose AI benefit still looks cyclical; invalidate if INTC server-CPU share stabilizes and foundry losses narrow faster than expected.
• Initiate a small tactical long in CBRS only on weakness, sized as a venture-style position rather than a core holding: upside is large if inference deployments scale, but stop out if customer concentration widens or delivery/service costs rise.
• Pair long AMD / short ARM over 6-12 months if inference economics push buyers toward heterogeneous, cost-sensitive deployments; thesis breaks if ARM materially gains AI-server socket share or premium valuation is sustained by licensing expansion.
• Add AMZN as a secondary beneficiary watch item, not a headline trade: lower inference costs can lift AWS AI consumption over the next 2-4 quarters, but only if cloud pricing remains disciplined and utilization improves.