The price of using artificial intelligence is falling fast. A closely followed measure of AI token prices touched fresh lows this week, CNBC reported, the latest sign of deflating costs in an increasingly competitive market for large-language-model access.
The LLM Token Expenditure Index, a daily gauge of market prices from intelligence firm Silicon Data, fell to 97 cents on Monday. According to CNBC, that marked the index's lowest reading since its creation late last year — and the figure has more than halved from the high recorded earlier this summer. For more context on this story, see our ongoing latest AI developments.
What the Index Measures
Silicon Data's index tracks the going market rate for a large-language-model token — the unit of text that models read and generate, and the unit on which most AI pricing is based. When the index falls, running inquiries on popular chatbots like OpenAI's ChatGPT, Anthropic's Claude or Google's Gemini generally costs providers less to serve and users less to buy.
That is good news for anyone building on top of AI models. It is more complicated for the companies behind them. As CNBC noted, a sharp slide in prices can condition consumers to expect lower rates for AI access, eroding the pricing power of the labs that produce the models.
Why Token Prices Keep Falling
The recent drop is being driven in part by the rise of open-source Chinese models. In a Tuesday post cited by CNBC, Charles-Henry Monchau, chief investment officer at Syz Group, pointed to Moonshot's Kimi K3, which can fetch lower prices than alternatives from leading frontier labs. Cheap, capable open-weight models set a reference price that closed models struggle to compete above.
Frontier labs have also been cutting their own prices. OpenAI announced price cuts to two of its GPT-5.6 models in late July, and Monchau told CNBC's readers that other frontier labs have rolled out offerings with "dynamic pricing" capabilities, allowing access rates to rise and fall with demand. Falling production costs across the industry — cheaper compute, more efficient models — have added further downward pressure on the market rate for tokens.
The Squeeze on Foundation Model Labs
The economics of token deflation are stark for the companies that build the models. "Foundation model labs are the most directly exposed," Monchau wrote, according to CNBC. "Token deflation compresses the revenue line while compute commitments stay fixed."
His suggested response is already visible across the industry. "The strategic response is visible: the moat must shift away from raw model capability — where the open-weight gap is now measured in months — toward distribution, memory and context," Monchau wrote. In other words, when the raw token becomes a commodity, the durable advantages lie in owning the customer relationship, retaining user context, and embedding AI into products people already use.
The timing matters. As CNBC reported, the LLM Token Expenditure Index's slide could add profit pressure to AI leaders like Anthropic and OpenAI as they contemplate entering the public market — both companies confidentially filed for initial public offerings with regulators this summer. Public-market investors tend to scrutinize declining unit prices far more intensely than private ones.
Ripple Effects for Investors and Users
The deflationary trend also complicates the investment case for the broader AI buildout. CNBC noted that investors may need to rethink their expected returns on the hundreds of billions of dollars being poured into data centers and chips, with megacap technology companies including Nvidia and Microsoft among those that have committed billions to expanding AI capacity. If the output of all that capacity — the token — keeps getting cheaper, revenue projections built on today's prices look optimistic.
Markets were already in a cautious mood. CNBC reported that technology stocks led the broader market down on Tuesday, with the tech-heavy Nasdaq Composite sliding nearly 1% and the S&P 500 ticking down 0.4%.
For users, meanwhile, the trend is straightforwardly welcome. Developers running AI features at scale, startups building products on API access, and everyday chatbot users all benefit directly as the cost per inquiry falls. The open question is how low the floor goes — and which providers can still run a profitable business when it gets there.
What to Watch
Three signals are worth monitoring in the coming months: whether the LLM Token Expenditure Index stabilizes or keeps sliding; whether more frontier labs adopt dynamic pricing to defend revenue; and how OpenAI and Anthropic frame token economics in the run-up to their potential IPOs. For continuous coverage of AI pricing, model economics and the business of AI, follow our AI business section.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
