As the subsidies for driving the adoption of AI tools reduces, I see the following combination unfolding:
- Open-source software, [[AI agents]], skills, harnesses becoming default options as good or better than closed source - ref: [[Open source is often misunderstood]] | [[Open Source Hyperscaler MoC]]
- On-device inferencing exploding with improvements in edge hardware, memory and storage capabilities
- Private AI rigs in offices and homes becoming as normal as a boiler and an AC
- Private clouds becoming easier to spin up on co-located hardware systems as compute becomes a commodity asset - ref: [[Inference is Eating AI Compute]]
This will be complemented by differentiation becoming a combination of:
- Niche application specific intelligences built with domain expert oversight and verification in RL environments
- Trust based access to customers, partners and distribution
- Speed of iteration, high agency, common sense, time in the market and business model
As this watt-bit value equation unfolds, companies such as Evos and Navon are best positioned to win.
![[Screenshot 2026-05-25 at 23.35.03.png]]
Relevant links:
- [[Personal Investment Roundup#AI & Computation Stack]]
- [[The Watt-Bit Spread and the Future of AI Power Markets]]
- [[7 Powers]]
- [[AI era Defensibility]]