Technology thesis · Artificial Intelligence
high conviction growthArtificial intelligence
AI is now general-purpose infrastructure with a four-way frontier race; value is concentrating in compute and power, and the real 2026 constraint is electricity and grid access, not GPUs.
Position maintained continuously · last reviewed Jun 24, 2026
The thesis
Core thesis
AI adoption is faster than any prior general-purpose technology – 53% global population adoption and 70% corporate deployment within 3 years (Stanford AI Index 2026). Enterprise spending is real: global IT spending hits 6.31tn USD in 2026, up 13.5% YoY, with AI spending alone forecast near 2.59tn USD. JPMorgan Chase reclassified AI from experimental R&D to core infrastructure. The infrastructure layer (Nvidia, hyperscalers) captures most value; the application layer remains fragmented. Regulation (EU AI Act GPAI duties from 2 Aug 2026, high-risk obligations deferred to Dec 2027) creates compliance moats favouring large providers.
State of the art (2026)
By mid-2026 the frontier is a tight four-way race with a compressed release cadence. Anthropic Claude Opus 4.8 (28 May) leads the Artificial Analysis Intelligence Index at 61.4, with Claude Fable 5 (9 June), OpenAI GPT-5.5 (April), Google Gemini 3.1 Pro (February) and xAI Grok 4.3 (late April) within a narrow band. The story has moved from raw benchmark scores to agentic, long-horizon work and to who can power the build-out: combined 2026 hyperscaler capex is guided at roughly 635-690bn USD, around three-quarters of it AI infrastructure, while Stargate commits 500bn USD over four years. Nvidia Blackwell is shipping in volume; Vera Rubin NVL72 ramps in H2. Power and grid interconnection, not GPUs, are now the gating supply problem.
The compute bottleneck determines who wins
Three supply problems control the AI timeline. (1) GPU and advanced packaging: TSMC CoWoS capacity stayed tight into 2026, though Blackwell volume and the H2 Vera Rubin ramp are easing it. (2) Power: grid interconnection, transformers and turbines now delay large US data-centre projects by 24-72 months, making electricity scarcer than silicon. (3) Talent: a few hundred frontier researchers shape the leading models, and their moves between labs signal where capability is heading. Together these create winner-take-most dynamics where early movers lock up capacity.
The agent transition
2026 marks the shift from AI as tool to AI as agent. Enterprises report agent deployments in code development, legal, financial tasks, and administrative support moving from experiments to production. This transition changes the value proposition from 'AI assists humans' to 'AI acts autonomously' — with profound implications for liability, regulation, and workforce restructuring.
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
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
3 disconfirming conditions
Comparable wave
The historical analogue on the S-curve
Common mistakes
What the market gets wrong right now
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 Artificial intelligence has changed — all live inside CanaryIQ.