An overview of macro tech trends: a capex explosion, unprecedented chip demand growth, supply chain bottlenecks, model commoditization, AI automation, and more
Twice a year, I produce a big presentation exploring macro and strategic trends in the tech industry. — New in May 2026, ‘AI eats the world’.
Context & Ripple Effects
The prior annual overviews trace a progression from early generative-AI opportunity and scaling questions in 2023 to a 2025 focus on larger-model training and rising capital expenditure, then to 2026 concerns around platform shifts, Nvidia, and US power backlogs.
This installment broadens that arc into a connected set of constraints: demand for chips and infrastructure is rising alongside supply-chain bottlenecks, while models themselves are becoming less differentiated. Related coverage also points to a tension between AI revenue growth and thin margins after data-center and chip depreciation.
First-order effects
- Companies building or buying AI capacity face immediate pressure from higher infrastructure spending, tight chip availability, and supply-chain constraints.
- Model providers face faster commoditization pressure, shifting near-term differentiation toward distribution, product integration, and automation use cases rather than the model alone.
Second-order effects
- Infrastructure bottlenecks make access to chips, data-center capacity, and power a competitive variable for AI platforms and their customers, rather than a back-office procurement issue.
- Thin margins despite growing AI sales increase pressure on AI vendors to improve utilization and monetization; customers are likely to compare increasingly interchangeable models more aggressively.
Third-order effects
- If the pattern persists, AI competition may split between capital-intensive infrastructure control and application-layer distribution, with standalone model advantage becoming harder to sustain.
- The industry’s limiting resource shifts from model development alone toward the physical and operational capacity needed to deploy AI at scale; the durability of that shift depends on whether demand and revenue ultimately support the capex cycle.
The trend: AI is moving from a model-development race into an infrastructure-constrained deployment cycle, where capex discipline and practical automation matter as much as raw model capability.