Why AI Model Cost Cuts Trigger a New Wave of Nvidia Demand

DeepSeek’s emergence dramatically lowered the training cost of large language models, leading many to expect a drop in demand for Nvidia GPUs, which briefly caused Nvidia’s stock to fall by up to 16%.

However, the reduced cost also lowered the barrier for companies to train their own models, attracting many new "players" and ultimately increasing overall compute demand, which benefits Nvidia and other GPU providers such as Huawei.

As a result, Nvidia’s share price has recovered to high levels, though many factors influence price movements.

Jevons Paradox

The 19th‑century economist William Stanley Jevons observed that when the efficiency of a resource (e.g., coal) improves, its overall demand can rise instead of fall because lower costs enable more widespread use.

Similarly, more efficient lighting (LED) or higher fuel efficiency can lead to increased total consumption—a phenomenon known as the rebound effect.

Similar Phenomena

LED lighting saves energy per unit but often results in longer illumination times or more fixtures, raising total electricity use.

Improved vehicle fuel efficiency can encourage longer trips or larger vehicles, offsetting some fuel savings.

Systemic Thinking

The modern AI boom mirrors Jevons paradox: DeepSeek and similar technologies lower training costs, inviting more companies to develop large models, which in turn boosts GPU demand.

Looking Ahead Multiple Steps

Three steps ahead, we may see new competitive dynamics and industry restructuring, with software and data resources becoming as crucial as hardware.

Step Three: New Competitive Landscape and Industry Re‑structuring

Lower technical barriers expand the market but intensify competition; Nvidia may face rivals and new business models, including cloud platforms offering flexible compute.

Smaller firms entering LLM training may lack data or expertise, creating opportunities for infrastructure providers, partners, and consultants.

Step Four: Technological Ethics and Societal Shifts

Widespread LLM adoption across healthcare, education, law, and media raises ethical, privacy, security, and intellectual‑property concerns, potentially widening the digital divide.

Step Five: Beyond Traditional Business Models

Long‑term, we might see "Compute as a Service" or "Data as a Service" models, shifting revenue from hardware sales to on‑demand computing and data provision, fostering new AI innovation ecosystems.

The Jevons paradox reminds us that technological progress brings long‑term, far‑reaching impacts; strategic thinking must consider multiple future steps to capture opportunities and mitigate risks.