Tech • AI • Robotics
My View on AI and Computing Power (Next 5 Years)
7/10TL;DR
Rising demand for computing power, driven by artificial intelligence, is positioning data centers and GPUs as strategic assets in a rapidly transforming global economy.
KEY POINTS
A thermodynamic view of technological evolution
The development of complex systems, from stars to human societies, can be understood through their ability to process energy efficiently. This framework links increasing complexity to higher energy flow per unit of mass, effectively treating information production as a mechanism to locally reduce entropy. Modern computing systems, including GPUs, fit into this continuum as highly efficient “diffusive systems” that transform energy into structured information.
Computing power as a measure of intelligence
Intelligence is framed as the capacity of a system to sustainably increase energy dissipation per unit of mass. Under this lens, advances in computing—especially in AI—represent a measurable rise in “intelligence density.” This perspective aligns with long-term trends suggesting that increasingly compact systems deliver exponentially greater computational output.
Early predictions of AI now materializing
In 1988, roboticist Hans Moravec estimated that human-level machine intelligence would require roughly 10 trillion operations per second, projecting its arrival around 2020–2030. Despite crude assumptions, similar timelines were later supported by Ray Kurzweil. By 2026, AI systems already surpass human performance in several domains, while autonomous vehicles and delivery robots operate in real-world environments.
AI reshaping industries unevenly
The economic impact of AI varies sharply between use cases. Many organizations deploy AI for marginal productivity gains, such as text generation or search assistance. However, a smaller group leverages it for transformative efficiency. In sectors like automated accounting, AI systems can fully process complex invoices, eliminating manual work and dramatically increasing scalability.
GPU scarcity and surging demand
Companies with high-value AI applications consume large volumes of GPUs, sometimes described as “absorbing them like candy.” Each additional unit of computing power directly increases output and market share. Industry demand significantly exceeds supply, with some operators reporting up to 20 times more demand than available capacity, indicating a structural shortage.
From user adoption to autonomous demand
Unlike previous technologies, AI adoption may become self-driven. Systems increasingly design, optimize, and even partially write their successors. This creates a feedback loop where machines themselves become primary consumers of computing resources, potentially shifting demand away from human-driven usage.
Economic value shifting toward compute ownership
Computing power is emerging as a core unit of economic production, potentially rivaling or surpassing human labor. In advanced AI workflows, developers increasingly supervise automated systems rather than directly producing code. This transition suggests that owning infrastructure—data centers, GPUs, and storage—could become a key determinant of future wealth.
Uncertain pricing dynamics and commoditization
While increased supply typically drives prices down, current trends show some GPU prices rising, even for older models. This may reflect differentiated use cases across industries. Instead of compute becoming cheaper, broader economic margins could compress as AI-driven firms outcompete traditional businesses.
CONCLUSION
As artificial intelligence accelerates, control over computing infrastructure is becoming a central strategic asset, with far-reaching implications for industry structure, labor, and economic value creation.
Explain this