The AI Surge: Jevons’ Paradox and the Future of Demand

In a world where artificial intelligence (AI) is becoming the heartbeat of modern technology, a curious paradox looms large: Jevons’ Paradox. This principle, which posits that as technological improvements increase the efficiency with which a resource is used, the overall consumption of that resource may also rise, seems particularly relevant in the current AI landscape.

The Supply Constraints of AI

Sundar Pichai, Alphabet’s CEO, noted in 2026 that we are still grappling with supply constraints in AI capabilities, hinting at both momentum and rapid adoption. According to Pichai, advancements in AI are not merely about throughput but also about meeting skyrocketing demand. It’s an exhilarating time for AI, but the supply-side challenges are significant, hinting at potential failures if the momentum stalls.

Andy Jassy, CEO of Amazon, echoed this sentiment in the same year, predicting that demand would outstrip capacity not just for 2026 but likely extending into 2027. As tech giants scramble to meet existing needs, the sheer volume of anticipated demand for 2028 is nothing short of striking. This combination of limited supply and incessant demand creates a high-stakes game for players in the AI arena.

Price Dynamics and Market Trends

As demand for AI capabilities grows, so too does the cost associated with utilizing them. Recently, Anthropic launched its Fable 5 product, marking a significant price increase to $50 per million output tokens—effectively doubling the cost of its predecessor, Opus 5. Google’s Gemini flagship similarly saw price jumps, moving from $1.50 to $12 across four generations. Such escalations raise a critical question: What will happen if the price of AI continues to double?

Interestingly, not all companies are following this trend. OpenAI made headlines by slashing the prices of its GPT-5.6 Luna by 80%, clearly signaling an aggressive market-capture strategy or perhaps a breakthrough in operational costs. This price fluctuation adds layers of complexity to the market for AI technologies.

The Consumption Cycle and Segmentation

One of the intriguing aspects of AI is the role of market segmentation. The theory suggests that while premium services may become increasingly expensive, value and mid-market tiers could absorb workloads that push high-end services out of reach. This segmentation can sustain Jevons’ principle by allowing overall GPU-hours consumed to continue increasing despite rising prices for premium AI models.

Interestingly, the premium tier is priced at 13 times that of the value tier while only providing about 20% more intelligence. The mid-market frontier, represented by products like GPT-5.6 Sol and Kimi K3, delivers 96% of the intelligence of premium models but at 40% of the cost. This effectively democratizes access to AI while keeping overall consumption on an upward trajectory.

The Future of AI Model Routing

As multiple tiers of AI capabilities become the norm, the role of routers—those strategic layers connecting buyers with the appropriate models—grows increasingly crucial. Routers will facilitate the nuanced job of routing queries to the right models based on customer needs. This functionality might exist internally within models or externally in customers’ software, indicating a shift towards more integrated solutions.

For major AI labs, maintaining a foothold in all three tiers—premium, mid-market, and value—is essential to remain competitive. Startups, on the other hand, can establish unique market positions by specializing in a niche at the frontier that larger corporations can’t economically rival. For instance, DeepSeek V4 Flash offers striking capabilities at merely $0.03 per operation, serving as a case study for maintaining competitive differentiation.

Jevons’ Paradox in Action

Ultimately, as the AI landscape evolves, Jevons’ Paradox manifests in ways that are both exciting and complex. While prices may rise and supply constraints loom, segmentation and strategic advancements can propel overall consumption higher. This dynamic interplay between capacity, demand, and innovative pricing structures offers a rich tapestry for theorizing the future trajectory of AI. The race is on, and the implications of these developments will ripple through various sectors for years to come. With the stakes so high, it will be intriguing to observe how this narrative unfolds.