Gemini cooked, GCP cooking: DeepMind shake-up & cloud financialization | SemiAnalysis note

Snapshot

Date

2026-08-07

Gemini ARR（估）

~$12B（2Q26）

GCP 近季增速

约 82%（含 TPU 系统）

2027 外售叙事

IaaS/TaaS~$73B + TPU 系统~$120B

As-of 2026-09-25 (weekly refresh; equities aligned to §A. Missing series are N/A/null. Not investment advice.)

Structured research note on a SemiAnalysis piece (summary + investable mapping)—not a reprint. Defer to the original for detail.

Thesis

2026-08-05 DeepMind leadership overhaul; SA sees frontier odds collapsing as compute tilts to GCP/external TPU sales. Gemini ARR ~$12B in 2Q26 vs ~$200B YE2027 GCP external AI+TPU systems narrative. See GOOGL.

Analysis

Leadership signal

Hassabis exits day-to-day; Jeff Dean et al. leave for Discovery Loop; Koray takes over. SA reads this as not “excited about Gemini 4,” and argues DeepMind is no longer frontier (opinionated research stance).

Model & usage

Gemini 3 Pro as peak then gap widens; 1P API token growth ~60% → ~38% QoQ. Vertex stronger via third-party models.

Compute politics

Large TPU volumes sold/leased to Anthropic etc.—platform logic over hoarding for Gemini.

Cloud financialization

$1.2B TPU system sales in 2Q26 ($35B/GW); >$150B systems backlog could drive mid-100s% GCP growth narrative vs ~64% sell-side; mid/high-30s EBIT; ~+$3 EPS 2027 (model).

Implications

Outlook

Watch Gemini 4 evals, TPU mega-deals/backlog disclosure, DeepMind stability, systems-sales mix inside GCP growth; open models catching up notes closed labs retain product RL loops, AgentX v3 quantifies agentic serving stack gaps.

Risks

References

1. 原文（SemiAnalysis）: Gemini is Cooked but GCP is Cooking
2. 对照 TPUv7 商用化: Google TPUv7…
3. 对照 SpaceX offtake: SpaceX 10GW…
4. Open models catching up: Are Open Models Catching Up? · AgentX v3

Not investment advice. Copyright remains with SemiAnalysis / authors; this is an index note.