For much of the AI boom, investors believed one thing: expensive AI meant booming demand for Nvidia Corp (NVDA) chips. Now the opposite may be becoming true.
As OpenAI cuts prices, Chinese challengers like DeepSeek and Kimi introduce lower-cost models, and enterprises gain access to cheaper AI than ever before, the economics of artificial intelligence are changing rapidly. While that may sound like bad news for companies building AI models, it could ultimately strengthen the investment case for Nvidia.
The AI Price War Is Driving Costs Lower
AI usage has become dramatically cheaper over the past few months.
SoFi Technologies, Inc. (SOFI) Chief Market Strategist Liz Thomas noted on X that average AI token costs have fallen from $2.07 per million tokens in late May to $1.02, citing OpenAI's price reductions and the growing availability of lower-cost open-source models from companies including Kimi and DeepSeek.
The broader trend is becoming increasingly difficult to ignore.
OpenAI recently reduced prices for some of its frontier models by as much as 80%, while Anthropic has also introduced lower-cost offerings as competition intensifies. Chinese AI companies, including DeepSeek and Moonshot AI, have further accelerated the industry's shift toward cheaper inference.
The competition is no longer just about building the smartest model. It's increasingly about building the most affordable one.
Cheaper AI Could Mean More Demand for Nvidia
At first glance, falling prices might appear negative for the AI ecosystem. Lower prices usually imply lower revenue per transaction. But technology markets often behave differently.
Economists call it the Jevons Paradox—the idea that making a resource cheaper and more efficient often increases total consumption rather than reducing it.
Early evidence suggests AI may already be following that pattern. After OpenAI reduced prices for some of its models, usage surged sharply. Business Insider reported that usage of GPT-5.6 Luna increased roughly fourteenfold, while Terra usage rose fivefold, with revenue also increasing despite the lower prices.
For Nvidia, that's an important distinction.
The company doesn't earn money from the price customers pay per token. It benefits when AI developers, hyperscalers and enterprises deploy more computing infrastructure to serve growing demand.
If cheaper AI encourages businesses to automate more workflows, launch more AI agents and process more inference requests, the total amount of computing required could continue rising—even if each individual AI query costs less.
The AI Winners May Shift, but Nvidia Still Stands to Benefit
The AI price war is undoubtedly putting pressure on model developers.
OpenAI, Anthropic and others must balance lower pricing with the enormous cost of building and operating frontier AI models. At the same time, open-source alternatives are forcing proprietary model providers to compete more aggressively on both performance and economics.
Hardware companies occupy a different position in that ecosystem.
As long as total AI workloads continue expanding, demand for GPUs, networking equipment and AI infrastructure can grow even if software becomes increasingly commoditized. The Wall Street Journal recently argued that the rise of cheaper open-weight AI models is unlikely to reduce demand for the industry's "picks and shovels," because broader adoption ultimately requires more computing capacity.
That doesn't mean Nvidia is insulated from every competitive threat. Efficiency gains, custom AI chips and evolving model architectures remain important variables.
But falling AI prices alone are not necessarily bearish.
What Nvidia Investors Should Watch Next
The more important metric may no longer be the price of AI, but its usage. If lower costs encourage enterprises to embed AI into more products, automate more workflows and serve millions of additional users, infrastructure demand could continue climbing even as token prices fall.
For Nvidia investors, the next phase of the AI boom may be driven less by increasingly expensive models—and more by making AI affordable enough to be used almost everywhere.
