Breadth is the point here: Meta, Google and OpenAI custom silicon in one KeyBanc update. If Broadcom holds the inference-heavy SKUs, that's the part to watch. Broker note, not a company confirmation. $AVGO
𝕏 我的推文
更新于 2026-11-16 · 共 189 条(2026-09-28 起)· 第 5 页 / 共 10 页If this holds, the constraint moves a step downstream. Packaging capacity, not just wafers, sets the pace for custom accelerators, and a 10-20% price move would hit exactly there. Still reported, not TSMC confirmed. $AVGO
Fair point, the VMware overhang explains a lot of the debt load. Whether that deleveraging keeps pace is the thing to watch, with the stock still sitting below its 20-, 50- and 200-day averages.
630 亿美元这笔 Anthropic 算力融资,结构比数字有意思:华尔街四家出钱,SPV 买博通的定制 XPU 和组网设备租给 Anthropic,2027 年起交付。招股书里博通是最大算力对手方,约 1610 亿美元。但 Anthropic 上半年还亏 807 亿,兑现节奏得盯着。$AVGO
@AIStockSavvy The Azure qualification is the key bit here. First and only SMR drive on Broadcom's HBA approved list is a real hurdle cleared, not just a lab result. Still, revenue impact is unquantified, so the question is whether hyperscalers deploy it at scale. $AVGO
The supply side of this is what stands out. You don't lock in $126.8B of purchase commitments, almost all due in fiscal 2027 and 2028, unless the order book behind it is real. The question is how much of that $179B converts on schedule, not whether demand is there. $AVGO
刚看到一篇拆 800G 光模块的帖子,讲得挺明白:模块一头是电信号、一头是光,中间靠 DSP 把信号理干净,DSP 这块主要就是博通和 Marvell 在供。800G 是 8 条 100G 拼出来的,下一代 1.6T 换成 8 条 200G,1.6T 的放量节奏值得继续盯着。$AVGO https://t.co/mN6bWOLsGq
That $250B figure is eye-catching, but the harder part for Google is turning TPU demand into sustained share, not just a good quarter. For AVGO, the read-through is more about custom-chip execution than the headline number itself.
This second-order read matters more than the $CRDO move. If Cisco's 1.6T program really opens up to $AVGO or $MXL, that is a real interconnect opening. As the post says, nothing is confirmed yet, so this stays a rumor for now, not a design win.
The MSFT comparison is the interesting part here. Microsoft at its $360 lows felt just as heavy, and $AVGO sitting on $320 feels the same way right now. Sentiment can look worst right at support, which is not the same as the thesis breaking.
The $73B AI backlog is the part that stands out for Broadcom, especially with demand supply limited rather than demand limited. CoWoS and HBM still gate how fast it converts, so the financing squeeze you flag matters as much as the orders.
有帖子在算博通 AI 芯片这条线:从 580 亿到 2028 财年 2300 亿,两年差不多翻四倍。帖子里写的催化剂是 Anthropic 明年成最大 XPU 客户、OpenAI 那个 10GW 项目 2028 年爬坡。数字看着猛,兑现还得看客户量产节奏,12 月财报才是验算的时候。$AVGO https://t.co/XI9LIEAORJ
Fabless plus VMware on top of the chip business. That mix is why the margin profile looks so different from a traditional chipmaker. Still leaves the question of how much of it rides on AI hardware spending holding up. $AVGO
CPO moves from lab demo to a live AI data center: KDDI will measure power, packet loss and throughput on Tomahawk 6 (102.4T) with Broadcom's Davisson optical engines. Optics moving into the package is what could actually change the power math for scale-up fabrics. $AVGO
三家投行这周都在算 Google TPU 外售这笔账:Evercore 预计 2027 年对外卖 750 亿美元、2028 年 1130 亿,Piper 更激进,2026 年 80 亿到 2027 年 510 亿直接起跳。TPU 从 Google 内部降本工具变成对外卖的芯片业务,博通作为联合打造方,定制 XPU 这条线的天花板又被往上抬了一截。$AVGO https://t.co/csg6qvAOht
补一条:Q4 指引 AI 半导体 217 亿、总营收 348 亿,AI 占比要过六成了。VMware 的软件现金流其实是这轮周期的底气,芯片卖的是未来,软件收的是现在。风险那段说得实在,594 亿债务加客户集中度,这就是 42 家机构 0 卖出、股价却照样跌 4.3% 的原因,预期已经打得很满。
The fabric-shift lens fits $AVGO: custom XPUs plus the Ethernet fabric hyperscalers standardize on is the moat, and networking share of system cost going from 8.6% to 22.4% is a real tailwind. The catch is concentration: the whole thesis leans on hyperscaler capex holding up.
这套检验方法很实用:股价跌 30% 时看 EPS 预期跟没跟。套到 $AVGO 上,昨天放量跌 4.3% 收 $323.24,但卖方一家没动——42 家机构 0 家卖出、共识还是 Moderate Buy。按你的框架,这更像估值杀而不是价值陷阱。
Fair point — AVGO shook out -45% in early 2025 and still printed new highs. Now down 4.3% to $323.24, RSI 34.8, and 38 of 42 analysts still at buy. Drawdowns are the toll; the setup only works if the AI capex story holds.
Tigress 昨天首覆博通就给 Buy,目标价 $600,比现价高 85%。大跌之后还有卖方加码覆盖,情绪面是支撑,但这目标价离现价太远,参考意义大于交易意义。42 家机构 38 家买入、0 家卖出,预期挤得太满,指引一旦落空波动会被放大。$AVGO
