ARK Disrupt: "Anthropic's Revenue Growth Sets Records, & More" (2026-10-19)
Why this is in the vault
Hard revenue/growth-rate data points on Anthropic (ARK's lead item, drawing on The Information, Business Insider, and Bloomberg reporting) are a useful external check on the commercial health of the lab underpinning the founder's Anthropic cert and phData credibility bet.
Issue contents
Four distinct bylined items in this Monday roundup:
- Anthropic's Revenue Growth Sets Records (Brett Winton, Chief Futurist) — the lead item. Per third-party reporting cited by ARK: Anthropic's Q3 2026 revenue was $16.8B (up ~45% sequentially vs. >140% in Q2), with annualized growth decelerating from 3,500% (Q2) to 340% (Q3) — a deceleration ARK attributes partly to Anthropic's crackdown on open-weight distillation and partly to intensifying competition with OpenAI. ARK estimates combined frontier-lab (Anthropic + OpenAI + modeled Gemini/xAI) ARR at ~$150B, up 5x YoY. ARK frames Anthropic as having "10x-d" revenue TTM — among only ~4 large US companies ever to more than triple TTM revenue while adding >$30B in absolute revenue (the others: NVIDIA Apr-2024, Micron Q2-2026, and OpenAI this same quarter). Reported adjusted gross margin: 85%. Framed in the context of Anthropic's pre-IPO investor meetings and an expected S-1 filing.
- Recursion's Phenomics AI Platform Delivers Clinical Proof Points (Ovid Amadi, PhD) — Recursion Therapeutics raised $200M; Phase 1b/2 data for REC-4881 (FAP) and early DAHLIA data for REC-1245 (RBM39 degrader, oncology); a Sanofi milestone payment. ARK holds Recursion positions across its biotech funds.
- Tesla's Roadster Can Fly! (Tasha Keeney, CFA) — next-gen Roadster demo at a SpaceX site; speculation about a SpaceX-Tesla combination. ARK holds TSLA across multiple funds.
- CoreWeave's Bond Yield Raises Alarms (Jozef Soja) — CoreWeave's 2032 unsecured bonds hit 15% yield (up from 9.625% at June issuance) amid AI-demand-slowdown concerns and a Gartner forecast that >80% of neoclouds cease independent operation by 2030. ARK's house view: CoreWeave's ~$50B in debt/leases is tied directly to its ~$104B contracted backlog (Microsoft, Meta, Anthropic, OpenAI as counterparties), so the financing structure is sounder than headline yield suggests. ARK holds CRWV.
Bias disclosure
ARK Invest is a fund promoting its own book, not a neutral research desk — this is fund commentary, no formal sponsor block because there's no third-party sponsor, but the self-interest is structural. ARK holds public positions in at least two of the four names covered here (Recursion, Tesla) and in CoreWeave, and its public funds carry heavy exposure to the broader "AI infrastructure buildout" thesis that a strong Anthropic/OpenAI revenue narrative directly supports (CoreWeave's backlog and bond story depend on frontier-lab demand staying robust). Anthropic itself is private — ARK has no direct equity position to talk up — but the Anthropic growth narrative is load-bearing for the CoreWeave bull case ARK argues two items later in the same issue. The "wall of worry is wrong, AI is not a bubble" framing is also ARK's standing house view, repeated across its last several weekly issues (see Related). Treat the underlying data points (reported Q3 revenue, reported growth-rate deceleration, reported gross margin) as third-party-sourced and reasonably solid; treat the "penetration rate," "10x-d" framing, and bond-yield reassurance as ARK's self-interested interpretation layered on top.
Mapping against Ray Data Co
The concrete connection is to [[project_phdata_cert_escalator_path]] and [[project_credibility_for_phdata_sales]], not to [[project_investing_markov_capital_cycle]]. The founder banked the Anthropic "Claude Certified Architect – Foundations" cert on 2026-08-17 as a career-credibility bet inside phData, and the explicit framing (founder, 2026-09-27) is "build credibility to help phData sales... domain = agents in production." Anthropic going from a research lab to a company reportedly doing $16.8B in quarterly revenue, 10x-ing TTM, and prepping an S-1 is direct evidence that the ecosystem the founder staked career capital on is now a commercially dominant, not speculative, platform — it strengthens the credibility bet rather than merely being adjacent color.
The capital-cycle investing thesis mapping is weak and should be stated as such: [[project_investing_markov_capital_cycle]]'s lead thesis is specifically the chip-fab/memory (DRAM/NAND) capital cycle, not frontier-AI-lab revenue growth. The CoreWeave bond item (debt financed against signed contracts, not speculative capacity) is a closer analog to capital-cycle mechanics — it's a leading indicator of compute-capacity financing stress — but it still sits one layer removed from the memory-cycle thesis the Markov tracker is actually built for. This issue is evidence for the general "AI infrastructure demand is real, not a bubble" prior, not a direct capital-cycle phase signal; filing it as a capital-cycle data point would overstate the fit.
Related
[[2026-10-12-ark-invest-openai-reporting-gap-ecdsa-spacex-spectrum]], [[2026-06-11-ark-invest-ai-infrastructure-buildout]], [[project_phdata_cert_escalator_path]], [[project_credibility_for_phdata_sales]], [[project_investing_markov_capital_cycle]]