AI Model Costs Plunge 95% in Three Years
AI model costs have fallen about 95% in three years, matching a decline in personal computer prices that took roughly 15 years, according to an a16z analysis using Goldman Sachs data. Its large language model pricing index fell from 100 in March 2023 to about 5 by September 2026, as competition, more efficient proprietary models and cheaper open-source alternatives pushed prices down.
The Silicon Data LLM Token Expenditure Index also hit a record low of $0.97 per million tokens in August 2026, down 29% from July and more than 50% below its May peak. Lower AI model costs benefit service users, but excess computing capacity could pressure margins and raise questions about infrastructure spending by major technology companies. The report has no direct cryptocurrency price catalyst, though cheaper AI compute may affect crypto-AI projects and broader technology-sector sentiment.
Neutral
The report does not identify a direct effect on cryptocurrency prices, token flows or blockchain activity, so the most appropriate market classification is neutral. The sharp decline in AI model costs could benefit crypto-AI businesses that rely on inference or model services by lowering operating expenses. However, cheaper AI alone does not guarantee more demand for blockchain-based services or greater token value.
In the short term, traders may treat the figures as a broader technology-sector signal rather than a crypto catalyst. Any reaction in crypto would likely depend on accompanying moves in technology shares, risk appetite and liquidity. A related concern is potential overcapacity: if AI infrastructure supply grows faster than demand, margin pressure could weigh on major technology companies and affect sentiment across growth assets. Similar episodes of rapid cost reductions in technology have often helped expand usage, but have also intensified competition and squeezed providers’ margins.
Over the longer term, sustained declines in AI costs could support new applications, including some crypto-AI projects, if adoption translates into real users and revenue. Traders should watch whether falling costs lead to stronger usage and sustainable business models, as well as broader indicators such as tech-sector performance, funding conditions and crypto market liquidity. Without evidence of those effects, the report alone is unlikely to change the market trend.