AI analysis for the people deploying capital.

No hype. No filler. Just the signal that matters for your portfolio.

Deep dives on AI infrastructure, competitive dynamics, and value capture.

16 years in tech11 years in AIFormer Google engineerFormer CoreWeave engineerFocus: PE, underwriters, due diligence, public markets

Latest Deep Dives

All deep dives →
July 24 Is Not Priced Into AI Infrastructure: $13-$38 billion of unpriced capex exposure in 2026 hyperscaler bridges, repricing on the Section 122 expiration.
tariffsSection 122Section 232

July 24 Is Not Priced Into AI Infrastructure

The 2026 hyperscaler capex bridge carries $13 to $38 billion of unpriced tariff cost that did not appear in any Q4 earnings call. Section 122 expires on July 24, but the duties do not come off - only the legal authority collecting them rotates. The 60 to 90 day handoff window is the tradeable event.

Read full analysis →
KAIROS daemon process: wake / evaluate / sleep cycle with 15-second tick intervals. 512K lines leaked, 1.3-20x cost range, 90.6% CPU bottleneck, $19B ARR.
Anthropicinferenceneoclouds

The Anthropic Leak and the Daemon Tax: What KAIROS Means for Inference Economics

On March 31, Anthropic accidentally shipped 512,000 lines of Claude Code's unobfuscated source to npm. Inside: a daemon called KAIROS that converts every seat into persistent inference. The cost model that underpins the GPU infrastructure buildout just changed.

Read full analysis →
The Decode Tax: 591x over-provisioned, 0.17% GPU utilization, $20B Groq IP license
ASICinferenceNvidia

The Decode Tax: Who Survives When Custom Silicon Demands Hyperscaler Scale

Nvidia spent $20B on Groq to fix the part of inference where GPUs waste 99.8% of their silicon. The distribution surface question is settled. The open question is who captures margin in the repricing.

Read full analysis →