This week, GSK's $10.6B acquisition of Nuvalent set the new ceiling for late-cycle oncology M&A, J&J locked up a degrader antibody conjugate platform for $1B, the FDA accepted Roche's sub-Q bispecific + ADC combination for relapsed B-cell lymphoma, and BIO 2026 opened in San Diego with the conversation finally shifting from "AI is interesting" to "AI is operating." We walk through what this week tells us about where capital is flowing, why the modality succession stack matters more than therapeutic-area framing, what the FDA-EMA Guiding Principles mean for sponsors filing in 2027, and where agentic AI is most likely to disrupt biotech workflows next. The honest open debate: 200+ AI-discovered drugs in the clinic, zero approvals — when does that change, and what happens to valuations when it does?

Chapter beats

1. Opening signal: GSK / Nuvalent and the new shape of oncology M&A (Section 3.1)

2. The Lunsumio sBLA and the bispecific-ADC pressure on CAR-T (Section 3.2)

3. BIO 2026 and the AI-pilot-to-production shift (Section 3.3)

4. Where capital is flowing — the asset/platform bifurcation (Section 4)

5. Regulatory read-through: Guiding Principles and FDA's own agentic AI (Section 5)

6. Forward look: the Modality Succession Stack and what's mispriced (Section 7 + Framework)