Why AMD, AVGO are outperforming Nvidia after Big Tech earnings

The article highlights a subtle shift in semiconductor sentiment as Nvidia briefly cooled after a big week of Magnificent Seven earnings. Investors are digesting large AI-related capex plans from Microsoft, Alphabet, Meta, and Amazon, which keeps the sector supported but introduces some near-term hesitation around Nvidia's stock performance.

Analysis

The key shift is not that AI capex is slowing, but that the market is moving from a pure demand story to a capacity-allocation story. When hyperscalers all telegraph outsized spend at once, the first beneficiaries are usually not the obvious CPU/GPU names on the headline, but the bottlenecks: advanced packaging, HBM memory, substrates, photonics, and the equipment vendors that can translate budget into delivered wafers. That creates a second-order relative-value setup where NVDA can consolidate even as the broader AI supply chain keeps outperforming.

Near term, the risk is positioning rather than fundamentals. NVDA is the most crowded expression of the AI trade, so any whiff of “spend is already priced in” can trigger de-grossing over days to weeks, even if the medium-term setup remains intact. The more interesting issue is whether hyperscalers are crowding each other into a capex arms race that compresses ROI and eventually forces a pullback in 2H25/2026; that would matter less for NVDA’s current backlog and more for the second derivative of orders in semi equipment and memory.

Contrarianly, the market may be underestimating how much of this spend is defensive rather than optional. If each platform is protecting model quality and inference latency, capex can stay elevated longer than consensus expects, but the marginal return shifts away from training chips toward inference infrastructure and networking. That argues for a broader basket approach: the trade is not “own NVDA or fade AI,” but “rotate from the obvious winner into the picks-and-shovels and latency-reduction beneficiaries.”

Key Decisions for Investors

• Trim NVDA into strength over the next 1-2 weeks; use any post-earnings-style digestion to reduce outright long exposure and redeploy into less crowded AI infrastructure beneficiaries. Risk/reward favors taking chips off the table because the name is now trading more on positioning than on near-term fundamentals.
• Go long AMAT/LRCX/KLAC-style semi equipment proxies versus short NVDA as a relative-value expression for 1-3 months. The thesis is that hyperscaler spend expands the tool cycle before it meaningfully re-accelerates GPU multiples; downside is a sharp NVDA re-rating if guidance proves even more explosive.
• Initiate a basket long on memory/advanced packaging exposure (e.g., MU plus OSAT/packaging names) for 3-6 months. These segments should capture the second-order demand lift from AI buildouts with better earnings leverage than the mega-cap capex beneficiaries already in consensus.
• For tactical hedging, buy NVDA downside spreads 4-8 weeks out rather than outright puts. This protects against a positioning-driven de-grossing event while limiting theta burn if the AI spend narrative reasserts itself quickly.
• Pair long MSFT/GOOGL/META/AMZN capital intensity beneficiaries with a short in lower-quality semiconductor cyclicals that lack AI content. The goal is to isolate the spend impulse while avoiding names that only benefit if broader electronics demand improves.