Google's TPU success raises questions for AMD stock, says Bernstein

Bernstein warns that Google's decade‑plus TPU ASIC program (developed with Broadcom) and plans to offer TPUs externally could meaningfully weaken AMD's narrative as a second source to Nvidia and its alignment with OpenAI, introducing downside risk to AMD's long‑term targets. The note flags Broadcom as a likely beneficiary with AI revenue currently underappreciated, while Alphabet reported $49 billion of quarterly backlog growth helped by a major OpenAI cloud agreement and is actively expanding TPU supply to third parties (including reported talks with Meta). Several brokers reflect this dynamic: BNP Paribas Exane and Citizens reiterate Outperform views on Alphabet, while DA Davidson stays Neutral with a $300 target, underscoring mixed but material competitive and revenue implications across chipmakers and cloud providers.

Analysis

Contrarian angles: the market may overstate near-term TPU displacement — ecosystem lock-in (CUDA, software stack) takes multiple years, so a NVDA or AMD pullback >12–15% on TPU headlines is likely overdone and creates a buying window. Historical parallel: prior ASIC vs GPU skirmishes (e.g., crypto miners) showed coexistence with GPUs regaining share when software demand scales; if AVGO executes, it becomes a longer-term consolidation/strategic-acquisition candidate. Watch for unintended consequences: Google-TPU expansion could invite regulatory scrutiny that slows commercialization — use >10% price moves or 2–quarter missed guidance as re-entry or exit triggers.

Key Decisions for Investors

• Establish a 2–3% long position in AVGO within 2 weeks, target 12–18% upside in 6–12 months; hedge with a 6–9 month 5–7% OTM covered-call if AVGO rallies >10%; stop-loss -8%.
• Add a 2% long position in GOOGL (or buy 9–12 month calls 5–10% OTM) to capture TPU monetization and cloud upside; re-evaluate after next quarter’s cloud revenue/backlog disclosure (target +15% in 12 months).
• Reduce direct AMD equity exposure by 1–2% and buy 3–6 month puts (10% OTM) sized to cover the reduction; if AMD loses a major hyperscaler inference deal within 90 days, increase puts to 3–4% notional.
• Implement a 1–2% dollar-neutral pair: long AVGO / short NVDA (equal notional) for 3–9 months to capture ASIC vs GPU reallocation; unwind if NVDA closes >+12% on no-fundamental news.
• If NVDA or AMD corrects >12% within 30 days absent fundamental deterioration, initiate 1–2% tactical buy for mean-reversion (add more on a 20% drawdown); otherwise prioritize AVGO/GOOGL exposure.