Executive Summary
SemiAnalysis has increased AI token spending from tens of thousands to $7 million annually—representing 25% of total salary expense—while generating productivity gains equivalent to 5-15x workforce multiplication. Dylan Patel's firm exemplifies explosive demand dynamics where implementation costs have collapsed but economic value creation has skyrocketed. Anthropic's revenue surge from $9B to $40B ARR with 72%+ gross margins demonstrates supply constraints creating unprecedented pricing power. The critical insight: token demand is growing faster than infrastructure capacity, creating a multi-year bottleneck across memory, logic, and fabrication equipment. TSMC's capex trajectory toward $100 billion by 2028 represents a 75% increase from current levels, yet still insufficient to meet demand. Memory prices will double or triple as capacity constraints force demand destruction through pricing. The semiconductor equipment supply chain faces a 'tail whip' effect where upstream suppliers like ASML, Lam Research, and Applied Materials cannot scale fast enough to support foundry expansion. This creates a structural advantage for companies controlling critical bottlenecks in the AI infrastructure stack, while simultaneously threatening to concentrate economic benefits among entities with preferential access to frontier models.
Key Insights
what Dylan Patel said“Ideas are cheap and plentiful but execution is very easy. So really only the good ideas are the ones that can justify the spend on super cheap implementation.”
what Dylan Patel said“DRAM will double or triple from here still, because that's how much capacity is required, and they have to steal capacity from somewhere else. And the only way to steal capacity from somewhere else in a capitalist economy is demand destruction via higher pricing.”
what Dylan Patel said“Three years from now, TSMC is going to spend $100 billion on capex. Maybe two years from now, it might be 28. Sincerely, they may spend $100 billion on capex in 2028”