Exchange Ranking
Top Exchanges
24h Volume Ranking
Popularity Ranking
Exchange BTC Balance
Proof of Reserves
Decentralized Exchanges
Funding Rate
Funding Heatmap
Liquidation Data
Max Pain
Long/Short Ratio
Whale L/S Ratio
Binance/Okex/Huobi L/S
Bitfinex Margin L/S
ETF Tracker
Solana ETF
XRP ETF
Hong Kong ETF
Bitcoin Treasuries
Crypto Reversal
Ethereum Reserves
HyperLiquid Wallet Analysis
Hyperliquid Whale Watch
Large Transactions
On-chain Movement
Bitcoin ROI
Stablecoin Market Cap
Options Analysis
News
Articles
Economic Calendar
Features
Wallet
Contract Calculator
Security
Collections
Watchlist
FollowingZach Pandl, research director at Grayscale, predicts that AI token consumption may surge 24 times by 2030.
As cost reductions drive penetration and usage, AI token consumption is expected to grow 24 times by 2030. AI agents may consume 5 to 50 times the number of tokens when processing tasks than standard chatbot interactions, which will significantly increase the need for inferential calculations.
Cheap AI models will expand application scope
Cost reductions make artificial intelligence easier to apply to various scenarios and automated workflows. According to Anthropic, its latest Sonnet model reduces the cost per task by up to 30%. At the same time, competition from low-cost open source weighting models has also put pressure on token prices. Tokens are a unit of measurement of the text processed and generated by AI models, and developers need to pay for the usage of tokens.
As prices fall, companies may find more AI application scenarios that were previously difficult to implement due to high costs. A research report released by Grayscale,"Investment Computing Power Bottleneck", explores in depth how cheaper smart technologies can support wider adoption.
AI agents may double computing needs
OpenAI's recently launched "dots" are an all-weather agent designed to continuously work around user goals. Unlike chatbots that only answer a single question, agents can complete multiple steps in a single workflow.
For example, preparing for client meetings may require reviewing notes, analyzing portfolio changes, and drafting an agenda. According to Grayscale's analysis, the token consumption for these tasks may be 5 to 50 times that of a single chatbot interaction, depending on the complexity of the tasks. As a result, the popularity of agents may drive a significant increase in inferential computing, the computing power needed to run AI models. Goldman Sachs predicts that by 2030, token consumption may increase 24 times due to widespread adoption of agents.
Physical infrastructure remains a computing power bottleneck
Although AI models are becoming more efficient, total computing demand will still rise as usage grows faster than efficiency increases. However, providing the necessary infrastructure takes time. Power supply, data centers and GPUs remain key elements in supporting expanding workloads. Their limited supply may limit the rate at which computing power can grow with the spread of AI.
Grayscale also cited the example of CoreWeave, which recently reported signing a new contract at a higher power price. This case shows that although the cost of AI tasks is declining, pricing pressures still exist.
Pandl's analysis points out that the increase in usage caused by cheaper models and more powerful agents is enough to offset the impact of increased efficiency. In this scenario, owners of scarce physical computing infrastructure will benefit from increased demand.
