The rise of low-cost Chinese artificial intelligence models may paradoxically drive greater demand for semiconductors, benefiting chip giants like Nvidia (NVDA) and Micron (MU).
The central argument draws on the Jevons Paradox, a 19th-century economic theory named for economist William Stanley Jevons that challenges conventional assumptions about efficiency and consumption.
The paradox holds that when a resource becomes cheaper to use per unit, previously uneconomical applications become viable, entirely new categories of use emerge, and adoption spreads broadly.
This pattern played out historically with coal, with telecommunications bandwidth in the 1990s and 2000s, and with cloud computing as cheaper infrastructure spurred explosive usage growth.
Applied to AI, Chinese models from DeepSeek, Z.ai, and Alibaba’s Qwen are gaining traction by narrowing the performance gap with leading U.S. systems while costing far less to operate.
Rather than threatening chip demand, some analysts argue these cheaper AI models could meaningfully expand it by unlocking workloads and users that were previously priced out of the market.
The rise of models like DeepSeek and Kimi in developer usage appears to be largely additive, bringing new workloads online rather than simply substituting away from paid frontier APIs.
Cheaper, capable models simultaneously expand addressable demand for inference compute and force closed frontier labs to differentiate into harder reasoning, longer context, and agentic capability.
Those next-generation capabilities require more training compute and more inference capacity, not less, potentially sustaining elevated demand for Nvidia and Micron hardware.
Still, the bear case carries real weight, as Nvidia’s data center revenue has scaled to levels that assume compute remains the binding constraint in AI development.
Micron’s high-bandwidth memory pricing power similarly rests on the assumption of insatiable memory demand driven by dense model architectures powering the AI industry.
If MoE and MLA adoption continues to compress per-query memory requirements, volume assumptions underpinning these elevated earnings could become increasingly fragile.
Nvidia slipped 2% on recent news, extending a rough week for semiconductors as traders reassessed how much compute the AI race actually requires going forward.
The selloff echoed the early 2025 “DeepSeek moment,” when a cheap Chinese model briefly erased chip valuations across the sector before buyers returned and recovered losses.
The structural prior, according to Jevons Paradox proponents, favors rising total compute consumption over time, though near-term demand saturation remains a legitimate risk worth watching.
