Google's Gemini 3 Sends Broadcom Soaring: TPUs Take Center Stage

Broadcom shares surged after Google released its Gemini 3 LLM, which Bloomberg Intelligence says was trained solely on Google TPUs developed with Broadcom, implying potential incremental TPU revenue (Nov. 19 >4% move). The stock closed at $397.57 (P/E 101.42, dividend yield 0.59%) and has delivered nearly a 67% total return YTD as investors price in deeper Google partnership and AI spending; Jefferies named Broadcom a top semiconductor pick versus NVIDIA. Wall Street consensus price targets range from $210 to $460 with an average/MarketBeat target ~ $377.6, and Broadcom reports earnings after the close on Dec. 11 — a key event for reassessing AI-related revenue outlook.

Analysis

Market structure: Broadcom (AVGO) is a direct beneficiary of Google (GOOGL) shifting training to TPUs — expect incremental design-win revenue and higher ASIC/adapter sales versus GPU-centric vendors. Nvidia (NVDA) is the obvious near-term loser in training spend share, but its installed base and software stack keep it competitive; market share shifts will be measured in quarters and likely concentrated in hyperscalers first. HBM/memory and advanced packaging demand should rise 5–15% for TPU-class workloads, tightening supply if foundry/packaging capacity does not scale.

Risk assessment: Tail risks include regulatory scrutiny of hyperscaler-supplier exclusivity, export controls limiting TPU deployment, and Broadcom revenue concentration on Google; a single large client re-prioritizing could knock 5–10% off AVGO revenue growth guidance. Near-term (days–weeks) the Dec 11 AVGO print and Dec 2 MRVL report are high-probability catalysts; medium-term (3–12 months) watch foundry capacity and model training cadence; long-term (years) software portability and Nvidia’s software moat are latent threats. Hidden dependencies: Broadcom’s upside depends on Google’s continued TPU-first strategy and Anthropic traction — both are execution risks.

Trade implications: Tactical long exposure to AVGO is warranted but should be size-limited and event-hedged; market-implied move into Dec earnings suggests buying on pullbacks to $360–$370 or initiating a smaller position now and hedging ~10% OTM puts. Relative-value: pair long AVGO vs short NVDA (notional ratio ~0.6) to express TPU share gains while limiting tech beta; use options (call spreads or put spreads) across the Dec–Mar window to define risk. Re-weight semis toward AI infrastructure names (AVGO, MRVL conditional on Dec 2 beat) and take profits on pure-GPU exposure if NVDA rallies >10% without fundamental beats.

Contrarian angles: The market may overstate a binary GPU-to-TPU switch — historically such shifts take multiple generations and billions of installed GPU-capex persist; Nvidia’s ecosystem and software (CUDA/NGC) remain powerful switching costs. Broadcom’s valuation already prices significant AI upside (P/E>100 in article)—a 10–15% miss in guidance would produce outsized downside. Unintended consequences include hyperscaler negotiating harder on pricing or insourcing TPU designs, compressing Broadcom margins over 12–24 months.

Key Decisions for Investors

• Establish a 1–2% portfolio long position in AVGO immediately; increase to 2–3% only on a pullback to $360–$370. Hedge the initial position by buying a 3-month put ~10% OTM (e.g., strike ~10% below entry) to cap downside to ~8–10%. Target 15–20% upside in 3–6 months; stop-loss at 8–10%.
• Implement a relative-value pair trade: long AVGO (notional X) and short NVDA at ~0.6X notional to express TPU share gains vs GPU exposure. Size this at 0.5–1% net portfolio exposure, rebalancing after Dec 11 and NVDA product/capture announcements within 30–90 days.
• Use options around earnings: if implied move into Dec 11 >4%, sell a near-term call spread (bear-call) on AVGO to collect premium (e.g., sell 30-day 410/430 call spread) or buy a cheaped-in debit call spread if expecting upside. Avoid naked directional options without defined risk.
• Initiate a conditional 1–2% position in MRVL ahead of Dec 2 earnings only if revenue/AI design-win commentary beats consensus; trim quickly on margin/guidance misses. Monitor SOXX flows and reduce commodity/GPU-exposed cyclicals if semis ETF SOXX outperforms broad market by >5% over 30 days.
• Reduce unhedged exposure to pure GPU plays (NVDA, NVDA suppliers) by 2–4% of portfolio if NVDA rallies >10% in the next 30 days without matching fundamentals; redeploy proceeds into AVGO and AI-infrastructure names with event-driven hedges.