Should You Forget Nvidia and Buy 2 Artificial Intelligence (AI) Stocks Instead?

Alphabet shows large AI-driven traction: Gemini exceeds 750M monthly users, Google Search AI queries are ~3x longer, Google Cloud backlog rose 55% sequential to $240bn and cloud revenue jumped 48% YoY; analysts flag a potential $900bn TPU market and GOOG trades at ~9x sales versus Nvidia at ~20x. Snowflake reports 13,328 customers (+21% YoY), AI-enabled customers >9,100 (more than double year-ago), remaining performance obligations up 42% YoY to $9.77bn, product revenue +30% to $1.23bn, and guidance for a 2.5pp operating margin improvement to 12.5%, supporting upside vs Nvidia. Investors may favor GOOG and SNOW for AI exposure and faster apparent upside given lower sales multiples and accelerating enterprise demand for AI services.

Analysis

The article highlights an important bifurcation in AI value capture: silicon OEMs (training/inference hardware) versus platform/software firms that fuse models with proprietary data. Expect increasing vertical specialization where cloud customers pay premium to vendors who can both host models and meaningfully reduce time-to-value by connecting to first-party datasets; that dynamic favors Alphabet and Snowflake but also creates margin pressure on pure-play infrastructure providers as they become commoditized execution layers.

Second-order supply-chain and competitive effects: if Alphabet commercializes TPUs at scale it will compress ASPs for training GPUs and force Nvidia to cede some hyperscaler wallet share, altering pricing power across the ecosystem within 6–18 months. Conversely, Snowflake’s serverless-GPU rental capability is a direct product-level encroachment on public-cloud compute revenue and could accelerate multi-year re-contracting toward vendor-neutral, data-centric stacks.

Key tail risks are binary: (1) rapid commoditization of model training (price cuts from excess fab capacity or new GaN/ARM entrants) that collapses hardware-based moats within 12 months; (2) privacy/regulatory limits on cross-customer data blending that would slow Snowflake’s RPO conversion beyond the next 2–4 quarters. Watch short-term indicators—GPU spot pricing, hyperscaler cloud-GAAP revenue mix, and Snowflake’s RPO-to-revenue conversion rate—because they will front-run valuation re-ratings.

Practical read: rotate from pure hardware beta into mixed-platform exposures that own both model distribution (search/ads + cloud) and dataset ownership. The asymmetric payoff is in firms that monetize recurring, contractual access to first-party data and can meaningfully reduce customers’ build time for AI apps.

Key Decisions for Investors

• Go long GOOG (3–12 month horizon): buy shares or a 12–18 month call spread (e.g., buy 18-month ITM calls, sell slightly OTM calls) size 2–4% of portfolio. Rationale: capture TPU monetization, search/ads AI upsell; target +30–50% upside if TPUs or cloud backlog growth re-rate valuation; stop-loss at -18% or if Google Cloud RPO growth decelerates two quarters in a row.
• Go long SNOW (6–18 month horizon): buy shares or LEAP calls, financed partially by selling near-term calls to improve carry. Rationale: accelerating RPO + serverless GPU demand; target 40–70% upside if RPO converts to ARR and operating margin improves as guided; hedge with a 25–30% trailing stop or reduce if AI customer count growth slows below 10% YoY.
• Relative-value pair (12–24 months): long GOOG / short NVDA via call spreads (buy GOOG 18m call spread, sell NVDA 18m call spread) size 1–3% net exposure. Rationale: express rotation from hardware multiple to platform/data multiple; wins if TPUs/Cloud monetization outpace further GPU-driven re-rating. Risk: NVDA fundamental beat will hurt the short leg—limit position to 1–2% notional and set a 20% adverse move cut.
• Event triggers & monitoring: enter or add on two clear signals—(a) Alphabet announces multi-hyperscaler TPU commercial contracts or pricing plans within 6–12 months; (b) Snowflake posts another quarter with RPO growth >30% and acceleration in RPO-to-product revenue conversion. Trim positions if GPU spot pricing collapses >30% QoQ or if major privacy regulation limits cross-customer model training.