Nvidia Reports Earnings This Month, and I'm Not Buying Shares. 1 AI Stock to Buy Now Instead.

Nvidia reported fiscal Q4 revenue of $68.1 billion, up 73% year over year, and guided fiscal Q1 revenue to about $78 billion, implying roughly 75% growth. However, the article argues that Broadcom-led custom chip programs and Amazon's $20 billion annualized custom silicon business are creating credible alternatives to Nvidia, which could pressure its pricing power and valuation. Amazon's AWS revenue rose 28% to $37.6 billion, while its chips business is said to be growing at a triple-digit pace and could reach a $50 billion annual run-rate if sold externally.

Analysis

The market is underestimating how custom silicon changes the bargaining power in AI infrastructure. Nvidia can still win the near-term training cycle, but once hyperscalers and frontier labs have credible in-house or semi-in-house alternatives, the fight shifts from unit demand to pricing, attach rates, and utilization discipline. That is the real second-order risk: even if GPU shipments stay strong for another few quarters, the margin stack can compress before revenue growth rolls over.

The cleaner beneficiary is the supplier of the picks-and-shovels for this customization wave. Broadcom gains because every custom-program win deepens switching costs and locks in multi-year design-in revenue, while also positioning it as the toll collector on the industry’s desire to diversify away from Nvidia. Alphabet and Meta are more subtle beneficiaries: every dollar they shift from merchant accelerators to bespoke silicon should improve long-run inference economics and lower cost per token, which is a competitive advantage that only shows up later in cloud share and ad-model efficiency.

Amazon is the most interesting asymmetry because the chip story is not standalone; it is a margin and ecosystem flywheel inside AWS. If management can keep custom silicon sold out while maintaining cloud growth, the market may be too slow to re-rate AWS as a structurally higher-quality compute platform rather than just a distribution channel for third-party chips. The main risk is capex discipline: if AI workloads decelerate even modestly over the next 6-12 months, Amazon’s spending intensity becomes the biggest earnings lever in the group, not the revenue driver.

Consensus seems too focused on whether Nvidia ‘wins’ the next quarter and too little on how the trade changes when customers become producers. That transition usually creates a two-stage outcome: first, multiple expansion in the alternative supplier; second, a valuation reset in the incumbent once growth becomes less monopolistic. The market may also be underpricing how fast custom silicon can move from internal efficiency tool to external commercial business, especially if AWS or Broadcom can prove they are selling capacity, not just substituting it.

Key Decisions for Investors

• Long AMZN vs short NVDA over the next 3-6 months: Amazon has more durable second-order upside from custom silicon monetization and AWS margin support, while Nvidia is exposed to eventual pricing pressure as alternatives scale.
• Add AVGO on weakness for a 6-12 month horizon: use pullbacks to build a position, as custom design wins create recurring revenue visibility and higher strategic value than a standard semicap exposure.
• Buy a NVDA downside put spread into the earnings window or 1-2 months out: the thesis is not a near-term miss, but a re-rating event if guidance acknowledges slower margin expansion or more competition in 2027-2028 demand.
• Pair trade GOOGL or META long vs NVDA short: both have internal incentives to lower AI compute cost and should benefit if custom silicon improves inference economics; this is a cleaner hedge against the merchant-chip cycle.
• Watch AMZN capex and AWS margin prints closely over the next 2 quarters; if capex stays elevated while AWS margins hold, add to the trade, but if capex rises faster than sell-through, fade the bullish custom-chip narrative.