China Telecom Forecasts Annual Token Consumption to Reach 1 Quintillion by 2026 · 0913-645
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
AI computing demand is shifting fully from model training to inference and agents. The China Telecom Research Institute predicts that China's annual Token consumption will reach 1 quadrillion by 2026 and exceed 350 quadrillion by 2030 [4][16]. Meanwhile, OpenAI has opened the Agents API that powers Codex, and Nvidia is reported to be considering investing $10 billion in Anthropic's massive IPO, as both upstream and downstream players in the industry chain enter a new expansion cycle in tandem [20][21].
🚀 Key Updates
- China's Token consumption may reach 1 quadrillion this year [4][16]: The share of inference computing is expected to reach 80% by 2029, with a compound annual growth rate of nearly 12x.
- OpenAI launches public beta of Agents API [21]: It opens the harness runtime framework that powers Codex to developers, while retaining three execution environment options.
- Nvidia plans to invest $10 billion in Anthropic's IPO [20]: The latter aims to raise up to $100 billion, with a valuation potentially reaching about $2 trillion.
- Apple Watch is redefined as the spearhead of Apple's AI hardware [1]: Through Siri AI, audio intelligence, and readiness index, it is upgraded from an accessory to a "personal intelligence hub."
- Micron significantly expands HBM production [14]: It plans to increase monthly capacity to 100,000 wafers by the end of the year, doubling from last year, challenging the landscape of SK Hynix and Samsung.
- 25 Fields Medal winners jointly warn that AI is undermining the foundations of mathematics [19]: This responds to the controversy over OpenAI's new model "solving" the Navier-Stokes problem in 88 hours.
- The optical fiber sector rises against the trend [17]: Demand is shifting from "connecting people" to "connecting computing power," with China Mobile's centralized procurement prices up 90%, and the shortage rate potentially reaching a high level by 2027.
- Cline launches a desktop client [23]: It offers a $6/month package aggregating mainstream open-source models, but does not specify the exact usage quota.
🔗 Sources
[1] Just now, Apple Watch became Apple's latest AI hardware — https://www.bestblogs.dev/article/4b12a2ea3c?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [4] Explosive growth! This year it may reach 1 quadrillion → — https://www.bestblogs.dev/article/a0bc2f9b10?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [14] Micron HBM, major expansion — https://www.bestblogs.dev/article/e35005f619?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [16] China's token consumption will reach 1 quadrillion this year — https://www.bestblogs.dev/article/945db3aedf?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [17] The optical fiber sector rises against the trend, what happened? — https://www.bestblogs.dev/article/9e3e6defd1?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [19] Top mathematicians jointly call out: AI is destroying the foundations of mathematics — https://www.bestblogs.dev/article/a465e5bd31?utm_source=rss&utm_medium=feed&utm_campaign