Cook 公开妥协: Tim Cook 在 WSJ 采访中明确表示 Apple 准备动用现金储备来增加存储供应，称“我们愿意使用资产负债表来成为解决方案的一部分”[1]TMT Breakout · 06-17"…nds ready to use its cash reserves to boost memory supply. “We’re willing to use our balance sheet to help be a part of the solution,” he said. “Obviously, more capacity is needed.”"。
Apple 压价失败: 有分析指出，Jensen 频繁造访韩国是有原因的，Apple 极大概率被存储厂商拒绝了低价要求[2]Tae Kim · 06-18"…eason why Jensen spending so much time and effort in Korea. Apple prob tried to push its weight around and the memory guys prob told Cook, go pound sand."。如果连 Apple 的购买力都无法击败芯片厂商，说明定价权完全在上游[3]Tae Kim · 06-18"If Apple buying power can beat chips, noone can."。
TPU v9 份额稳固: JPM 的 Harlan 确认 AVGO 仍在 TPU v9 项目中且没有延迟，直接反驳了此前 SA 关于 v9 只有联发科 (MDTK) 的传言[4]PandaBMU · 06-17"Harlan from JPM saying AVGO is in TPU v9 and there is no delay...SA was saying v9 is only MDTK"。群友将此比作去年 Marvell 与 Alchip 在 Trainium 3 上的分歧[5]Mei · 06-17"It is like Marvell vs Alchip Trainium 3 situation last year (and the year before). Opposite views across the Pacific oc…"。
FY28 营收看高: Wolfe 报告指出，Apollo/Blackstone 的 XPV 融资工具可能支持 Anthropic 的采购，使 AI XPU 出货量在 FY28 达到 15GW，支撑约 $250-300B 的 FY28 XPU 营收机会[6]TMT Breakout · 06-17"Couple typos in the AVGO wolfe recap: should read $250-$300B (Not "M") of FY28 XPU revs (not FY26)...Corrected below: Fo…"。JPM 甚至给出了 28 年 20-22GW 的预期[7]PandaBMU · 06-17"He has 20-22GW for AVGO in 28' with OAI at 7GW and ANT at 7GW vs 1.5GW OAI in 27' and 3.5G…"。
AMD CPU 份额变化: 访谈显示 CoreWeave 需求依然强劲。历史上的 CPU 主要用于控制平面（多为 AMD 和 INTC），但未来部署将发生变化，AMD CPU 的评论引发关注[8]Josh S · 06-17"yeah no it's always been AMD cpus historically and some INTC cpus"。
前 CEO 代理信: 前 CEO Toby Neugebauer 发布代理信，指责管理层未能推进来自五个主要战略方（包括超大厂和芯片公司）的入站兴趣，以及投行带来的三个额外交易机会[9]AGdoom22 · 06-17"Proxy letter out today from Toby Neugebauer ex-CEO as well (GPT summary) • Neugebauer claims Fermi had inbound interest …"。
OpenAI 10GW 传闻: Citizens 研报指出，The Information 报道 OpenAI 正就在俄亥俄州租赁 10GW 数据中心进行谈判（由 NVDA 提供资金），这大大增加了 Fermi 签署首个 11GW 综合体租约的可能性[10]Adam Fee · 06-17"Was from June 9th or 10th: "Citizens views The Information's report that OpenAI is in talks talks to lease a 10-gigawatt…"。
光模块情绪转弱 空·光通信
产能过剩担忧压倒基本面，买方寻找理由降仓
InP 产能过剩恐惧: Mizuho 的 Klein 指出，光模块板块交易情绪恶化，买方担忧全球新增过多 InP 产能，可能导致明年激光器供应过剩和价格崩盘[11]TMT Breakout · 06-17"…nt TSEM, MTSI and lastly SMTC. I hear more and more talk of FEARS DRIVEN BY TOO MUCH InP CAPACITY BEING ADDED GLOBALLY, raising risk of excess laser supply and pricing collapsing at some point next yr. Hence many think buyside ests have n…"。
预期见顶: 市场认为当前买方对 CY28 的预期已经比卖方高出 25-35%，且 CPO 采用慢于预期，毛利率可能在今年或明年见顶[11]TMT Breakout · 06-17"… pricing collapsing at some point next yr. Hence many think buyside ests have no real room for further upside as they are already 25-35% above sell side for CY28 and CPO update and adoption way slower vs expected. Add in …"。
算力争夺失败？: 群内推测 Noam 离职的可能原因包括：厌倦了为新架构争取算力分配，或与传统 DeepMind 团队在通往 AGI 的路线（LLM vs 其他）上存在分歧[12]Anoncapital · 06-18"There's no way to spin this in a positive way. 3 possible explanations: 1. Noam got tired of fighting for compute alloca…"。
短视的资源倾斜: 有观点批评 Google 降低 DeepMind 算力优先级的做法是“近期的财务贪婪，非常短视的举动”[13]Tae Kim · 06-18"near term financial greed. very myopic move. Satya doing the rigth thing (to be fair, it might not work…"。
数据偏鹰: Fed 连续四次按兵不动，18 位官员中有 9 位预计今年至少加息一次，PCE通胀预计到 2028 年才能回到 2% 目标[14]AGdoom22 · 06-17"SUMMARY OF FED DECISION (6/17/2026): Fed leaves rates unchanged for the 4th straight meeting 9 out of 18 officials expec…"。
任务小组战术: Warsh 频繁提及“价格稳定”并引入“任务小组”(task forces)，被视为一种政治缓冲和拖延战术。有分析认为这是为了避免上任首日引发市场恐慌而建立信誉的手段[15]Isaac · 06-17"As a general rule in politics, usually when a politician starts talking about a task force, they're punting. Here I thin…"。
市场反应背离: 尽管会议基调被多位群友认为非常鹰派（类似 2022 年），但市场（特别是 QQQ）在发布会后强劲拉升。群内总结：“芯片股永远赢”[16]Tae Kim · 06-17"remember the most important rule that overrules all rules: chip stocks always win"。
板块轮动: 有观点指出，近期 AMZN、META 等超大厂的疲软主要由资金轮动驱动。Mag 7 被抛售以资助金融和工业板块（如 GS, JPM, GE）的上涨[17]ashutze · 06-17"nobody wants to hear this, but this hyperscaler weakness is mainly driven by sector rotation; Mag 7 sold and financials and industrials being bought (GS, JPM, CAT, HON all leading SNP500 higher; AMZN, MSFT, G…"。
开源利空论被否: 有人猜测开源模型是否拖累了超大厂，但多位群友反驳：云厂商在托管开源模型上能获得更好的利润率，开源对 AWS 等反而有边际帮助[18]TMT Breakout · 06-17"Don't see why neg for hypers - they serve and get better margin dollars on open source models. Really not sure why hypers down today."，OSS 不应被视为利空[19]RI · 06-17"Ya OSS shouldn't be a -ve for the hyperscalers. It’s not like the mkt has gave them credit for tokenomics …"。
🟥▼与 2022 年 FOMC 类比: 加息周期重演风险 · 加息周期/美联储政策 Bessents Burner: I mean im getting 1:1 flashbacks to this point in 2022 guys, its very much so hawkish and unless brent stays pinned at 60 or 70 we will enter a hiking regime
🟥▼利率波动率回归: 少写期权意味着 Rate Vol 回来 · 利率波动率/风险偏好 Arun Dhar: one thing we can ALL agree on is that Warsh is willing to write less options and so Rate Vol is back. On the margin, stocks don't like rate vol.
🟦▼前 CEO 代理信: 管理层错失 5 大战略方 + 3 笔交易机会 · 管理层失误/并购机会 AGdoom22: Proxy letter: Neugebauer claims Fermi had inbound interest from five major strategic parties (including hyperscalers, chip companies, and tech investors) that management failed to pursue. An
🟦▲LFL 提价: 罕见调高定价 · 提价/定价信号 Mei: TEL to increase LFL pricing by SD, which is rare in this sector. There is also chatter that other SPE companies eliminating bulk discounts and asking for more expedite fees.
🟦▲定价信号: 行业普遍取消批量折扣、加收加急费 · 定价信号/供需紧缺 Other SPE companies eliminating bulk discounts and asking for more expedite fees. Pricing is the signal the market is consistently looking for uncovered bottlenecks.
🟥▼市场开始定价其为 AI capex 的下一个股权融资候选者 · 股权融资/AI资本支出 Methodica: AMZN has traded poorly since June 1, roughly coinciding with GOOG's equity offering. It feels like the market may be starting to price AMZN as a potential next candidate to fund the next le
🟥▲亲赴韩国确保内存供应: 间接证据 · 管理层表态/内存供应 Tae Kim [+1:3]: there's a reason why Jensen spending so much time and effort in Korea. Apple prob tried to push its weight around and the memory guys prob told Cook, go pound sand.
🟥▼因子交易主导: 价差荒谬至极 · 价差/因子交易 Ted: $SOXX vs. $MAGS spread truly absurd. Maverick: The factor trading has become nothing less than rediculous at this point. The nerds have taken over.