ARK Weekly Roundup: Grail NHS Trial Failure, FDA One-Pivotal-Trial Reform, Chinese Humanoid Robots

[!warning] Conflict of Interest ARK Invest is an active fund manager promoting its own positions. ARK holds Grail and genomics positions. Their analysis of the NHS trial failure is colored by their investment thesis. Read with appropriate skepticism.

Key Themes

• Grail NHS trial missed primary endpoint: Largest RCT for MCED test (142,000 participants) did not achieve statistically significant reduction in Stage III+IV cancer diagnoses. Stage IV declined meaningfully but Stage III increased. Stock dropped ~50%. Grail extending follow-up window. Core challenge: sensitivity of only 17% for Stage I and 40% for Stage II cancers. ARK frames this as "neither refuting nor validating" MCED -- a notably measured take given their position.
• FDA shifts to one-pivotal-trial default: Commissioner Makary and CBER Director Prasad formalized that one pivotal trial plus confirmatory evidence replaces two-trial standard. Could save up to $350M per drug program. Already 66% of 2024 NME approvals used single pivotal trial. Vision for "continuous trials" with cloud-accessible data replacing sequential phases.
• Chinese humanoid robots: ~98% of global humanoid robot sales. ARK estimates general-purpose humanoid is ~200,000x harder than robotaxi. Cumulative compute investment for human-level proficiency ~$55B, crossover point potentially in 2028.

RDCO Relevance

Grail setback is important signal for multiomics investment thesis -- sensitivity limitations are real. FDA reform could meaningfully reduce biotech development costs and timelines. Humanoid robotics compute requirements data point useful for AI infrastructure projections.