AVGO: Buy the Dip Before the Rocket Takes Off
Broadcom 预期强劲 AI 增长与 3D 堆叠芯片里程碑。
1. 关键信息
- 3D-stacked chips 1M target by 2027; TPU v7 cost per token down 70% vs v6 (GS).
- AI-accelerator CAGR ~mid-to-high 50% 2024–2029 (TSMC).
- $11B AI chip order from Anthropic H2 2026; new 2026 custom customer adds $1B; OpenAI expected 2027.
- FY26 AI revenue projected to >2×; FY27 further doubling; $15B revenue per GW for OpenAI deal.
- 10 GW total deployment target for 2027; "fully secured" demand 2026–2028.
- XPU customers: Google, Anthropic, Meta, OpenAI, Apple, plus 1 unnamed; direct-attached copper preferred.
- 32% revenue concentration in one customer; software (VMware) ~42% mix buffers AI margin.
- China exposure: 17% of revenue, but much routed to US/EU endpoints.
- Regulatory: AVGO least directly impacted vs NVDA/AMD by export controls (custom ASIC + network + diversified).
2. 羊毛/优惠信息
无
3. 最新动态
- Reuters: 1M 3D-stacked chips by 2027 confirms long-term visibility; premium pricing yields strong dollars despite margin pass-through.
- Option flow: predominantly selling puts (low premium) on AVGO; implied move ±8.79%; Feb/Mar 280–290 puts noted (#35, #39, #40, #41).
- Sentiment: mixed; some waiting for "explosive" beat (#31, #32), others buying dips (#13, #29).
- Short interest: AVGO 1.05% (least shorted mid-mega cap tech) (#21).
4. 争议或不同意见
- Bear case: customer concentration, legacy software (VM/Mainframe), acquisition slowdown, antitrust (#2, #16).
- Margin dilution concern vs belief that custom AI chips remain accretive (#9, #14).
- Execution risk on 10 GW and 100B AI revenue by 2027 (#47, #50).
- Regulatory overhang could slow hyperscaler projects but AVGO structurally insulated (#54).
5. 行动建议
- Accumulate on dips targeting $365–$480 fair value; monitor March 4 earnings for guidance confirmation.
- Use defined-risk options (e.g., put spreads) to manage volatility while awaiting catalyst.
- Balance concentration risk via position sizing; hedge with less sensitive names if needed.