brad gerstner

Anthropic added $11B ARR in single month (March 2025), fastest revenue ramp in tech history; unit cost of intelligence plummeting 90% YoY on inference, creating Jevons paradox where cheaper intelligence drives exponentially higher consumption; Gerstner predicts $80-100B exit run rate by year-end.

SpaceX's successful $75B raise at $1.75T with staged lockups and early index inclusion created a repeatable playbook for mega-IPOs; Anthropic ($100B+ revenue trajectory) and OpenAI ($70B+ revenue, GPT-6 imminent) are next in line, with top-tier investors (Altimeter) committed to buying at scale.

Physical limits on memory wafers, logic wafers, and powered shell constrain token production; no dark GPUs exist today — every chip is utilized. Hyperscalers (Google, Amazon, Microsoft, OpenAI, Anthropic) all report being token-constrained, creating multi-year demand visibility for semiconductor and infrastructure suppliers.

Nvidia uses frontier models to design its next-gen chips, which then enable better models; this recursive loop makes Jensen Huang a permanent, price-insensitive buyer of the absolute best models, anchoring demand for Anthropic/OpenAI at the intelligence frontier regardless of commodity token pricing.

Despite 90% token cost reductions over 2.5 years and open-source advances, Anthropic and OpenAI's share of enterprise wallet is increasing because complex, long-horizon agentic tasks (e.g., replacing a $200/hr consultant) require bulletproof intelligence where the $12 inference delta is irrelevant, preventing commoditization at the frontier.

AI-enabled coding is currently ~5% of market but will reach 95% within years; Anthropic's Claude Code launch (Feb 2025) triggered revenue explosion from $4B to $30B run rate in months, proving enterprise demand for labor augmentation at scale.

All US GDP growth comes from data center buildout driving AI productivity. Moratorium would cause recession, high unemployment, and immediate loss of global AI race to China (100 fission reactors vs 1 in US). Local community concerns (water, electricity) require tangible benefit-sharing solutions to maintain social license.

Unlike 2000's dark fiber, every GPU and token is fully utilized today; Google, Amazon, Microsoft, OpenAI, Anthropic all report token constraints limiting revenue. Memory/logic wafers, power, and shell are hard bottlenecks creating a structural supply ceiling that sustains pricing power for compute providers.