"Every great technological transition looked obvious in retrospect and impossible at the time. We are very early."
The internet in 1995 — that's where we are with AI in 2026
Transformative technology always undershoots near-term and overshoots long-term predictions
Physical world transformation is the next wave: robots, manufacturing, medicine
The economic surplus from AI will dwarf the internet era
Framework
Marc's AI Phases
$1T
AI investment 2025-2030
100×
productivity in knowledge work
50%
of GDP eventually AI-driven
Phase 1 (now): Knowledge work augmentation — ongoing
Phase 2 (2027+): Physical world automation — beginning
Phase 3 (2030+): Economic restructuring — coming
The bottleneck is now energy and compute, not models or data
The a16z thesisEvery industry will be disrupted. The question is not if, but whether you're the disruptor or the disrupted.
What's Actually Coming
The Physical World Wave
Robotics: AI-native robots are 5-7 years from mass deployment
Medicine: Drug discovery, diagnosis, surgical assistance — all transforming now
Manufacturing: Dark factories and AI-native production are real this decade
Energy: AI-optimized grids, nuclear renaissance driven by AI compute demand
The bull case
AI enables human flourishing at scale — cheaper medicine, better education, more creative work for humans.
The honest risk
Transition costs are real. Economic displacement requires new institutions, not just new models.
Playbook
Position for the AI Era
Build at the infrastructure layer — picks-and-shovels always win in gold rushes
Invest in AI companies building in physical world applications
The talent edge: people who can bridge AI capabilities to domain expertise
Regulatory moats will matter — incumbents will use them; work around them
The Netscape lessonThe browser seemed like a toy. The companies that took it seriously won the next 30 years. AI is that moment.
Contrarian
AI Pessimist Myths Marc Disputes
✗AI will destroy jobsINSTEAD →✓ Technology always creates more jobs than it destroys — over time. The transition is real and hard.
✗Regulation will slow AIINSTEAD →✓ Regulation will shape AI, not stop it. The countries that regulate intelligently will lead.
✗The AI bubble is burstingINSTEAD →✓ The investment is rational. AI is already producing measurable economic returns.
✗Open source will neutralize the labsINSTEAD →✓ Both can win. Open source democratizes capability; closed models differentiate on safety and reliability.