Technology thesis · Artificial Intelligence
medium conviction growthEdge AI
On-device LLMs have crossed from demo to product – 8B models run on phones, Apple Foundation Models open to developers – making power and privacy, not model quality, decide where inference runs.
Position maintained continuously · last reviewed Jun 24, 2026
The thesis
State of the art (2026)
Edge AI in mid-2026 is defined by on-device LLMs crossing the usefulness threshold. Qualcomm has demonstrated 8B-parameter Llama running on a phone NPU, and Apple used WWDC 2026 to double down on its on-device Foundation Models, opening them to developers rather than retreating to the cloud. Copilot+ PCs now ship with 45–55 TOPS NPUs as standard, though uptake is uneven. The defining new vector is SPANs XFRA programme, announced April 2026 with PulteGroup and NVIDIA, which places liquid-cooled RTX PRO 6000 Blackwell nodes in homes as distributed inference capacity from 2027. Power, privacy and latency – not raw model quality – now drive where inference runs.
Everything below is live inside CanaryIQ
The full analysis behind the verdict — the structure is real; the content unlocks when you log in.
Signal stack
Evidence stacked leading → lagging
Technology-native KPIs
Metrics that predict trajectory, tracked over time
Landscape map
Who builds what — and who depends on whom
Catalyst calendar
Dated events that will move the position
Technology roadmap
Milestones on the path to maturity
Watchlists
Companies, people and papers — each with a remove-by condition
Decision frameworks
The same call, framed for your desk
Thesis changelog
When our view changed, and why
Change our mind
2 disconfirming conditions
The rest is inside
You've read the verdict. The file is much deeper.
The full signal stack, technology-native KPIs tracked over time, the landscape of who depends on whom, the dated catalyst calendar, decision frameworks for every desk, live watchlists and the changelog of every time our call on Edge AI has changed — all live inside CanaryIQ.