Andreessen Horowitz’s inaugural “State of Markets” report for 2026 highlights a tech-driven supercycle where top-tier winners are pulling away from the pack at an unprecedented scale. David George (a16z General Partner) notes that revenue growth for top-quartile companies has exploded, reinforcing the power law nature of venture capital where a tiny fraction of firms generates the vast majority of returns.

The report addresses "bubble" concerns by analyzing infrastructure utilization. Notably, Google reports that seven-year-old TPUs (specialized chips for AI) are still running at 100% capacity. This suggests that the hunger for compute is a sustained structural shift rather than a temporary spike. This high demand remains constant even as hardware ages, proving the depth of the current AI integration cycle.

We are also seeing Jevons Paradox in action: as the cost of AI tokens (basic units of data processed by models) drops, total consumption is skyrocketing. Instead of saving costs, companies use increased efficiency to embed AI more deeply into products through indirect monetization. For this "trillion-dollar" vision to hold, the focus must now shift toward translating these high-utilization Large language model (LLM) services into tangible earnings growth.

Andreessen Horowitz’s inaugural “State of Markets” report for 2026 highlights a tech-driven supercycle where top-tier winners are pulling away from the pack at an unprecedented scale. David George (a16z General Partner) notes that revenue growth for top-quartile companies has exploded, reinforcing the power law nature of venture capital where a tiny fraction of firms generates the vast majority of returns.

The report addresses "bubble" concerns by analyzing infrastructure utilization. Notably, Google reports that seven-year-old TPUs (specialized chips for AI) are still running at 100% capacity. This suggests that the hunger for compute is a sustained structural shift rather than a temporary spike. This high demand remains constant even as hardware ages, proving the depth of the current AI integration cycle.

We are also seeing Jevons Paradox in action: as the cost of AI tokens (basic units of data processed by models) drops, total consumption is skyrocketing. Instead of saving costs, companies use increased efficiency to embed AI more deeply into products through indirect monetization. For this "trillion-dollar" vision to hold, the focus must now shift toward translating these high-utilization Large language model (LLM) services into tangible earnings growth.