As Bitcoin hits price discovery it's important to pay attention to the coins showing strength.Nansen 24-hour net inflows show that AI tokens continue to run hard as majors break:i) @virtuals_io : ACP + Decentralized Governance.Once again 7-figure net inflows and top of the leaderboard as ACP gathers momentum.The first phase of governance proposals passed highlighting an engaged veVIRTUAL holder base.Expect more to come as innovative teams launch agents to disrupt their respective industries, and bolster the ...

Robinhood’s @OpenAI tokens walk a legal tightrope, says @0xethlaw in an interview with @thesamreynoldshttps://t.co/Ma4KLg3csp ...

Core Insights - Anthropic's Claude 4 AI model has achieved a significant milestone by being able to work continuously for seven hours, which is a notable improvement over existing AI models that typically operate for seconds to minutes [3][5][10] - The performance degradation observed in AI tools like ChatGPT during long conversations is attributed to limitations such as context window size and token limits, which can lead to reduced accuracy and coherence [9][10] - The future of AI working continuously for 24 hours is technically feasible but raises concerns regarding the cost, environmental impact, and the need for self-imposed limits on usage [11][12][13] AI Performance and Limitations - Claude 4 can hold 200,000 tokens, nearly double the capacity of ChatGPT's 128,000 tokens, which contributes to its ability to maintain performance over extended periods [7] - Once the token limit is reached, the context window is flushed, necessitating a restart of the conversation, which can affect the continuity of information [8] - The computational burden of managing large amounts of information is a key factor in the performance degradation of AI models during lengthy interactions [9] Future Considerations - The cost of maintaining high-performance AI systems is significant, requiring advanced cooling and electricity, which could limit the feasibility of extended work sessions [11][12] - The environmental impact of AI operations is becoming increasingly concerning, with potential implications for resource availability, such as water shortages due to cooling demands [12] - The discussion around the pursuit of a 24-hour AI workday includes ethical considerations regarding the costs and benefits of such advancements [13]

证券研究报告 科技 英伟达 Computex：开放互联生态+ 端侧 AI 部署，引领 AI 生产力变革 华泰研究 2025 年 5 月 20 日│美国 动态点评 英伟达 Computex Keynote：开放互联生态，加快端侧 AI 部署 英伟达 CEO 黄仁勋于 5 月 19 日发表 Computex 2025 主题演讲，我们认为 对行业启示为：1）新发布 NVLink Fusion 可与第三方 CPU 和 AI 芯片集成， 或吸引大型云厂商客户，意味着互联行业开放生态的到来；2）推出 DGX Spark/Station 和 RTX PRO 服务器，或表示 AI 需求已逐步转向个人开发者 到企业级客户；3）强调通过 AI 基建引领 AI Factory 驱动的工业革命。此外， 区域布局上，英伟达宣布与鸿海和中国台湾方面合作建设搭载 1 万颗 Blackwell 芯片的 AI 超级计算机，并将中国台湾总部搬至北投士林科技园区。 英伟达作为 AI Tokens 基础设施建设者，引领 AI Factory 驱动的工业革命 黄仁勋强调，AI Factory 是生产 AI Tokens 的"智能工厂"，包括从数据预处 ...

Computex 现场连线#1：英伟达、高通主旨演讲 20250519 摘要 COMPUTEX 是全球最重要的 PC 和服务器产业链活动之一，今年展会包括英伟 达、高通、红海、联发科和 NXP 等公司的主旨演讲以及整个 PC 和服务器产业 链的展示。英伟达的演讲符合预期，重点在 AI 服务器产业链的发展，包括 AI 工厂、agentic AI、机器人以及企业 AI 系统转型。高通则宣布进入服务器 CPU 市场，这对 ARM 生态系统形成闭环有利，并打破 X86 生态。此外，中国 台湾的服务器产业链非常强大，从整机到各个零部件都非常完整，但应用端较 弱。 • AI Token 被视为未来重要工具，数据中心将成为 AI 工厂，以能源为输入 产出 AI Token，未来数据中心每月或每年生产多少 AI Token 将成为重要 指标，标志着 AI 应用落地。 • 中东市场出货量预计增加，可能取代新加坡，对英伟达和 AMD 均有影响， 值得关注。英伟达芯片已交付 OEM，收入已确认，但 OEM 后续工作进度 影响整体供应链。 • 英伟达第二季度财报指导可能在 45 至 47 亿美元之间，考虑了 5.5 亿美元 HRC ...