Dylan Patel of SemiAnalysis: The $11M Bill, the Memory Shortage, and Why CPO Is Two Years Late
ATTRIBUTED/UNVERIFIED: Anthropic FCF-positive, $50B+ ARR, 70%+ gross margin; memory multi-year structural shortage (capacity +20-30%/yr vs demand doubling); CPO real ramp 2029 vs Street 2027 (Rubin/Feynman all-copper; Amphenol named beneficiary). Disclosed conflict: Patel/SemiAnalysis is a reported Anthropic enterprise customer.
Dylan Patel of SemiAnalysis: The $11M Bill, the Memory Shortage, and Why CPO Is Two Years Late
PROVENANCE NOTE — read before citing. This is a podcast interview, not a filing or dataset. Every claim below is Patel's stated view, not an independently verified fact. Patel/SemiAnalysis is a disclosed active Anthropic enterprise customer (reported ~$7M/year on Claude Code) — a material conflict running through every Anthropic-specific claim in this episode. Fetched 2026-07-14 via the article's header/preview; the full transcript is paywalled, so the claims below are the complete set surfaced by that preview, not necessarily the complete episode. Every downstream use MUST carry "Patel claims/estimates" framing — do not present these as verified facts.
Claims (attributed to Patel; unverified)
Anthropic financials
- "Anthropic is free cash flow positive." — Patel claims, unverified.
- Anthropic is "past $50B ARR with 70%+ gross margins." — Patel claims, unverified. Patel is described in the piece as "a large customer with monthly-close specifics," implying direct commercial visibility — but this is Patel's own disclosed vantage point, not third-party confirmation.
Memory shortage
- "Memory is not a cycle top. It is a multi-year structural shortage." — Patel claims/estimates.
- Growth-rate mismatch: memory capacity grows 20-30%/year while AI demand doubles (~100%/year) — Patel claims/estimates, framed as the structural driver of the shortage thesis.
CPO (co-packaged optics) ramp timing
- Patel's call: real CPO ramp lands 2029, vs. Street consensus of 2027 — Patel claims/estimates, a two-year delay call against the sell-side.
- Reasoning given: "Rubin and Feynman are all-copper" — i.e. NVIDIA's next two GPU generations stick with copper interconnects rather than adopting CPO, pushing the real transition out.
- Named beneficiary trade: Amphenol (copper interconnect supplier) — Patel's stated trade idea, not investment advice, unverified.
Disclosed Conflicts
SemiAnalysis (Patel's firm) is an active Anthropic enterprise customer (~$7M/year on Claude Code per reporting) — read the Anthropic-specific claims (FCF+, ARR, margin) with that in mind.
Source: Dylan Patel of SemiAnalysis: The $11M Bill, the Memory Shortage, and Why CPO Is Two Years Late, Podcast Alpha, published 2026-07-10. Fetched via header/preview (full transcript paywalled) — treat as partial-primary.