AI Stack Radar

Overall I am optimistic about the AI build-out. Each layer of the stack has its own risks, so I follow them separately.

Updated 2026-10-20

Neocloud

Optimistic, watching debt

Neocloud names such as CoreWeave and IREN may face headwinds from concern about financing and debt.

I am less worried than the headlines suggest: their assets are GPUs, and their contracts are with investment-grade counterparties.

Following: CoreWeave, IREN

Reviewed 2026-10-20

LLMs

Optimistic

The frontier model layer is led by Anthropic and OpenAI. There are real security concerns, but given the benefits I believe AI progress should not be slowed down.

I am sceptical about AI consciousness. My suspicion is that during training we create an illusion of it, and that adding rules around it may, ironically, reinforce the impression.

Following: Anthropic, OpenAI

Reviewed 2026-10-20

Connectivity & chips

Optimistic

I follow the connectivity and custom-silicon names closely, alongside Nvidia.

Following: CRDO, AVGO, Marvell Technologies, Nvidia

Reviewed 2026-10-20

Applications & agents

Watching

Many agent solutions look good, for example Muse from Meta and the Grok bot.

This layer needs more time before I form a firm view, but there is plenty of good news that the benefits are diffusing into the enterprise.

Following: Meta, xAI (Grok)

Reviewed 2026-10-20

Personal views for discussion only. This is not investment advice or a recommendation to buy or sell any security. Markets change; views are dated and may change.