UBS On-Air: Market Moves Signal over Noise with Ulrike Hoffmann-Burchardi
Mar 29, 2026
A briefing on which market moves are signal and which are noise. Discussion of Google's TurboQuant news and why memory stocks sold off. Examination of AI efficiency forecasts and the Jevons paradox driving compute demand. Analysis of the Anthropic Claude Mythos leak and its impact on cybersecurity. Conversation about AI expanding attack surfaces and the investment implications for memory and cyber names.
AI Snips
Chapters
Transcript
Episode notes
Market Overreacted To TurboQuant Announcement
- Google's TurboQuant claim sparked a sharp market reaction: Micron fell 15.5% and SK Hynix and Samsung dropped double digits.
- Even storage-focused Sandisk declined in sympathy despite not being levered to compute bottlenecks.
Memory Sell-Off Was Largely Noise
- Google's TurboQuant announcement shouldn't have shocked memory stocks because the paper was published last April and Frontier AI labs likely already use it.
- UBS CIO models already assume ~20x AI inference efficiency gains by 2030, so this fits expected software-hardware evolution rather than a disruptive surprise.
Efficiency Gains Can Increase Compute Demand
- Efficiency gains can increase total compute demand via the Jevons paradox, meaning cheaper AI inference may raise overall memory and compute consumption.
- UBS expects lower per-unit compute costs to drive more AI use, boosting aggregate demand rather than reducing it.