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

Broadcom is benefitting from an expanding TPU partnership with Google after Google released its Gemini 3 family (Nov. 18), which Bloomberg Intelligence reports was trained solely on TPUs; Gemini 3 Pro ranks at the top of LLM benchmarks and should drive further TPU demand. Broadcom shares rallied more than 4% on Nov. 19 and closed at $385 on Nov. 25, near their all-time close of ~$386 on Oct. 29, and have delivered nearly a 67% total return in 2025; Google is up roughly 70% YTD and NVIDIA about 32%. Analysts remain constructive: MarketBeat’s consensus target is $374 (≈3% downside) while analyst targets averaged $423 (≈10% upside) with Raymond James at $420; Broadcom reports earnings Dec. 11 and Marvell (a competitor) reports Dec. 2, both events to watch for signs of TPU/custom AI chip demand.

Analysis

Market structure: Google (GOOGL) and Broadcom (AVGO) are clear near-term winners—TPU adoption takes incremental AI training spend away from NVIDIA (NVDA) GPUs and boosts AVGO’s high‑ASP custom ASIC revenue. Expect AVGO gross‑margin tailwind if TPU volumes scale; NVDA faces cyclical demand pressure for training GPUs but retains ecosystem lock‑in (CUDA) that limits total share loss. On supply/demand, hyperscaler-driven bespoke silicon reduces spot GPU demand volatility but concentrates semiconductor revenue into fewer, higher‑margin deals.

Risk assessment: Tail risks include Google shifting back to GPUs or in‑house ASICs (estimated 15–25% plausibility over 12–36 months) and regulatory/antitrust scrutiny of hyperscaler–chip partnerships (~5–10% probability). Short term (days–weeks) volatility will be dominated by AVGO earnings on Dec 11 and MRVL’s Dec 2 print; long term (12–36 months) the key is TPU workload migration and Google CapEx cadence. Hidden dependency: AVGO’s AI revenue is highly concentrated to Google/Anthropic—any contract re‑pricing materially compresses forward margins.

Trade implications: Tactical: favor defined‑risk long AVGO exposure ahead of Dec 11 but hedge event risk—target 2–3% position sizing with 8% stop and 10–15% upside target to analyst mean $423 within 3–6 months. Consider buying a Jan 2026 AVGO 360/420 call spread (limited cost) or selling covered calls 4–6% OTM post‑earnings to harvest premium; pair trade long AVGO vs short NVDA (dollar‑neutral, 1–2% net) to express TPU share shift while limiting market beta.

Contrarian angles: The market may be overstating permanent share transfer—NVIDIA’s software stack and broad GPU TAM make full displacement unlikely; AVGO is near all‑time highs (high P&L risk) and revenue concentration raises idiosyncratic downside. Historical parallels (hyperscaler custom silicon wins that later re‑centralized) suggest validating multi‑quarter revenue trends before levering up; if MRVL shows strong competitive wins on Dec 2, reprice AVGO exposure quickly.

Key Decisions for Investors

• Establish a 2–3% long position in AVGO common stock before Dec 11 earnings, target 10–15% upside to ~$420 within 3–6 months; place a stop‑loss at ~8% below entry (manage position size to limit portfolio drawdown to <0.25%).
• If preferring options, buy a defined‑risk Jan 17, 2026 AVGO 360/420 call spread sized to equal ~2% portfolio exposure (max loss = premium); roll or exit if AVGO >$430 or if Dec 11 EPS/AI revenue beat exceeds +10% YoY.
• Construct a dollar‑neutral pair: long AVGO (1.5–2% of portfolio) and short NVDA (1–1.5%) to express TPU tailwinds vs GPU slowdown; cut the pair if NVDA outperforms AVGO by +6% or if AVGO reports AI revenue miss on Dec 11.
• Hedge AVGO event risk: buy 5% OTM protective puts expiring ~30–60 days after Dec 11 (or sell 4–6% OTM covered calls after earnings) to monetize elevated post‑report implied volatility; reduce AVGO exposure by 50% within 5 trading days if MRVL (Dec 2) guides stronger competitive traction than expected.