AI Token Prices Hit Record Low at 97¢, Squeezing OpenAI and Anthropic as Demand Surges

AI Token Prices Plunge to Unprecedented Levels

AI token prices have crashed through a critical threshold, with the LLM Token Expenditure Index falling to 97 cents per million tokens on Monday—its lowest reading since Silicon Data launched the benchmark late last year1

1

2

2

. This represents a dramatic collapse from the peak recorded earlier this summer, when the index stood at roughly $2 per million tokens, according to UBS3

3

. The index, which tracks the usage-weighted effective price across a defined universe of large-language models and is listed on Bloomberg under ticker SDLLMTK, has more than halved in just months1

1

. The 8.6% decline over the seven days leading to Monday signals accelerating token deflation in an increasingly competitive landscape2

2

.

Open-Source Chinese Models Drive Market Disruption

The recent drop in record low token prices stems partly from the emergence of open-source Chinese models like Moonshot Kimi K3, which undercut pricing from established frontier AI companies, according to Charles-Henry Monchau, investing chief at Syz Group1

1

2

2

. OpenAI announced price cuts for two of its GPT-5.6 models in late July, while other frontier labs have rolled out dynamic pricing capabilities that allow access rates to fluctuate with demand1

1

. These developments have intensified downward pressure on the market rate for tokens. Decreasing production costs across the industry have also contributed to the index's decline, creating a perfect storm for AI providers1

1

2

2

.

Margin Pressure Mounts for OpenAI and Anthropic

Token deflation compresses the revenue line for frontier AI companies while compute commitments stay fixed, creating acute pressure on business models2

2

. Foundation model labs face the most direct exposure to this pricing collapse. The timing couldn't be more delicate for OpenAI and Anthropic, both of which confidentially filed for initial public offerings with regulators this summer1

1

2

2

. Lower token prices risk conditioning AI model users to expect cheaper access permanently, eroding pricing power for AI providers over time1

1

. The strategic response is already visible, Monchau notes: companies must shift their competitive moat away from raw model capability—where the open-weight gap has narrowed to just months—toward distribution and contextual advantages like memory and context1

1

2

2

.

Paradox of the AI Economy: Lower Prices, Higher Spending

Despite falling AI token prices, overall spending on AI computing infrastructure continues climbing, illustrating the "Jevons paradox" where greater efficiency spurs increased consumption rather than reduced spending3

3

. Token usage on OpenRouter, a platform providing access to hundreds of large language models, has increased tenfold since the start of the year3

3

. Average monthly spending among the platform's top 1% of users has surged to $7,500 from $2,500, UBS reports3

3

. Token demand rises through longer prompts and responses, multimodal apps, search and retrieval, and more intensive reasoning capabilities. AI agents that can use tools, develop plans, and execute tasks without human oversight further amplify token consumption3

3

.

Infrastructure Investment Cycle Remains Intact

The AI flywheel continues spinning: cheaper AI enables greater usage, higher usage supports demand for computing capacity, and stronger demand encourages further investment in data centers and equipment3

3

. Megacap technology companies including Nvidia and Microsoft have poured billions of dollars into plans to expand their capabilities to power AI1

1

. Steve Hou, Silicon Data's head of research, suggests the drop may signal that existing supply across frontier and lower-cost models is already sufficient to handle most tasks2

2

. For investors, UBS emphasizes that token pricing and consumption have become critical indicators of activity across the AI economy3

3

. Rather than signaling weaker spending, falling token prices may evidence that AI adoption is broadening rapidly enough to push aggregate consumption and infrastructure investment higher. Technology stocks led the broader market down on Tuesday, with the Nasdaq Composite sliding nearly 1% and the S&P 500 ticking down 0.4%1

1

2

2

. Investors may need to recalibrate their outlooks around potential return on invested capital as the AI economy evolves with these pricing dynamics1

1

.