Get your free personalized podcast brief
We scan new podcasts and send you the top 5 insights daily.
As AI moves from simple chatbots to complex agents, its demand for computation grows exponentially. This is because advanced techniques like chain-of-thought reasoning and using AI to generate more efficient code fundamentally involve using massive amounts of inference to achieve better results, creating a feedback loop.
Related Insights
Unlike previous technologies defined by human demand, AI creates its own. AI agents require compute to perform tasks, and those tasks generate more data and complexity, in turn demanding more compute. This self-reinforcing cycle means the total addressable market for intelligence is effectively infinite.
Today's AI computing demand from millions of human users is just the beginning. The real explosion in demand will come from billions of AI "agents" working 24/7. This will double the workforce and create a relentless, round-the-clock need for inference computing, dwarfing current infrastructure requirements.
Leaders at frontier labs like OpenAI and Anthropic indicate that RSI—AI models that self-improve—is closer than anticipated. The arrival of RSI would trigger unprecedented demand for compute, as models consume vast resources to develop and improve themselves autonomously.
The shift from simple chatbots (one user request, one API call) to agentic AI systems will decouple inference requests from direct user actions. A single user request could trigger hundreds or thousands of automated model calls, leading to an exponential increase in compute demand and cost.
Ben Thompson argues the shift from simple chatbots to AI agents creates an exponential, non-speculative demand for compute. Agents automate complex, multi-step tasks, driving constant usage that justifies the massive capex investments by hyperscalers. This suggests the current spending is based on real demand, not bubble-fueled speculation.
The massive growth in AI token consumption isn't a sign of waste but of ambition. While the cost per "unit of intelligence" is decreasing, companies are immediately applying that efficiency to solve exponentially harder problems. Our appetite for more capable AI is growing faster than the cost is falling, leading to sustained, exponential spending.
The shift from simple query-based AI to agentic AI, where AI calls itself recursively to solve complex tasks, increases compute demand by orders of magnitude. Most people, especially non-coders, fail to grasp this exponential shift, leading them to consistently underestimate the scale and duration of the AI infrastructure build-out.
AI's computational needs are not just from initial training. They compound exponentially due to post-training (reinforcement learning) and inference (multi-step reasoning), creating a much larger demand profile than previously understood and driving a billion-X increase in compute.
The next wave of AI adoption involves 'agentic' workflows, where AI performs complex tasks autonomously. This shift from simple queries to agentic use is expected to increase token consumption by approximately 10x per task. This will drive a massive explosion in compute demand across all knowledge-work industries, not just coding.
The transition from chatbots to autonomous 'agentic' AI represents a fundamental step-change. These agents, which execute complex tasks independently, have already increased the demand for computational power by 1000x, creating a massive, ongoing need for new infrastructure and hardware.
