INTC 估值透支 ⟪LINK17⟫
Intel 靠先进封装打入代工，但当前价格已透支 2028 预期。
- EMIB-T 成代工敲门砖: 多头最初因 EMIB-T 包装技术在 $50 看多 → 视其为吸引客户尝试代工的楔子 → 最终拉动晶圆级代工需求[1]Terminal Junkie · 05-09"Random Thoughts $intc : I’ve been fighting with myself on Intel for a while. I started liking the stock around *$50/shar…"。
- 估值容错率变薄: 原目标价 $118，现价 $126 迫使市场计入 2028 年 EPS。假设代工 2025-2028 达 mid-30s CAGR，目标价可看 $165.54，但被指“一个月计入了 20 年的进展”[1]Terminal Junkie · 05-09"Random Thoughts $intc : I’ve been fighting with myself on Intel for a while. I started liking the stock around *$50/shar…"[2]Ben Pouladian · 05-09"I agree it’s priced in 20 years of progress in one month"。
Anthropic 算力瓶颈 ⟪LINK18⟫
Anthropic 无法实现 10x ARR 增长，核心约束是拿不到足够算力。
- 12GW 推理电力缺口: 播客观点认为算力短缺或持续至 2027，制约其明年 10x ARR 增长[3]Chris · 05-09"well i think he and the others on pod ultimately concluded it woudnt be possible for Anthropic to do to another 10X in A…"。群内推演：假设 5 亿 DAU，按 Blackwell 效率计算，需 6-7 GW 净推理电力，算上冗余需 12 GW 容量，目前远未达到 50%[4]Arun Dhar · 05-10"Yeah. T+1yr state imagine 500m daily active users. Assume 50% consume simply API call or simple query of 1000 tokens x 1…"。
- xAI 收入来源存疑: 关于 xAI 在 Anthropic 交易中获得 $5-6b 收入的传闻，群内指出所有报道均无可靠信源[5]Arun Dhar · 05-11"I am not sure these "cluster was too heterogenous" takes re: colossus 1 failure to build SOTA are correct. I mean the fa…"。
Claude Code 迭代 ⟪LINK19⟫
Claude Code 的 Agent Teams 功能被视为改变游戏规则的创新。
- 替代 Cursor: 开发者已删除 Cursor → 完全依赖 Claude Code 的 Auto Mode 和多智能体协作（如专门的代码审查 agent） = 每天节省上百次点击[6]H-Cap · 05-09"of course. Just deleted Cursor. Was using both (Cursor as second set of eyes)"[7]H-Cap · 05-10"this is auto mode. i dont have to approve anything except "merge to main" in my case. saves me hundred clicks a day. the…"。
- 5 步渐进式 Prompt: 群内分享了 5 步渐进式 Agent 部署 Prompt：查阅 repo → 解释现有功能 → 询问痛点 → 提议单点添加 → 等待确认，非技术人员也能借此建立渐进式信任[8]H-Cap · 05-10"you can get started by pasting this prompt in your claude code environment: I'm new to Claude Code and want to set up an…"。