概念词典 · 技术
ASIC 专用集成电路
为特定算法/任务定制的芯片，AI 时代主要指各大科技公司绕开 NVIDIA GPU 自研的训练/推理加速器。
别名： ASIC／ 专用集成电路／ 专用芯片
是什么
Application-Specific Integrated Circuit (ASIC) 是为特定算法或任务定制的芯片，区别于通用 CPU 与 GPU。在 AI 语境下，ASIC 几乎等同于 大厂自研 AI 加速器，目的是绕开 NVIDIA CUDA 生态、降低单算力成本。
代表 ASIC 项目：
| 项目 | 谁的 | 设计代工 |
|---|---|---|
| TPU | $AVGO Broadcom | |
| MTIA | Meta | $AVGO Broadcom（部分） |
| Trainium / Inferentia | Amazon | $MRVL Marvell |
| Maia | Microsoft | $AVGO |
为什么 ASIC 是博通 / Marvell 的现金牛
设计一颗高端 ASIC 需要：
- 高端 SerDes / HBM PHY IP
- 7nm/5nm/3nm 工艺经验
- 与 TSMC 的产能关系
这些都是 $AVGO 和 $MRVL 在传统通信芯片时代积累的。所以 Google / Meta / Amazon / Microsoft 自研 AI 芯片实际是 “外包给 Broadcom 设计”，“自研”叙事在财务上呈现为 Broadcom 的 ASIC 业务收入井喷。
Serenity 的关键判断
他对 ASIC 板块的核心论点：
ASIC 时代到来的速度取决于 TPU 商用化进度
如果 TPU 仅限 Google 内部使用，AVGO 的 ASIC 业务有天花板；如果 TPU 像 GPU 那样卖给所有人，整个 ASIC 板块定价权迁移，进入 NVIDIA 之外的”第二极”。
代表标的：
- $AVGO Broadcom：ASIC 龙头，受益于 Google + Meta + Microsoft 三家订单
- $MRVL Marvell：Amazon Trainium 主力设计代工
- $ALAB Astera Labs：ASIC 配套 retimer + smart cable
提到此概念的推文 (115)
按时间倒序 · 滚到底自动加载更多
-
博主看好光子学供应链，认为激光器光纤将重现内存超级周期。
花旗银行TMT团队：“激光器和光纤正被定位为‘下一个HBM（高带宽内存）’。我从2025年就开始谈论像 $LITE / $AAOI 这样的激光器公司，以及向连续波(CW)架构的转变，这领先于当前的行业共识。我完全同意Lightmatter和花旗的观点。仅供参考：激光器方面：$LITE / $SIVE / $AAOI / $COHR / $MTSI / 古河电工 / 住友电工是较知名的标的。光纤方面：$GLW / 古河电工 / 古河光缆 / 住友电工是较知名的敞口（显然供应链中还有更多）。*披露，个人对该主题有持仓。补充评论称：‘紧张的供应状况将支持这些资产的稀缺性溢价。’我们已经看到激光器价格上涨，正如今天在深圳CIOE 2026上看到的 $SIVE 渠道检查以及 $LITE 的财报评论所示。因此，我认为2027-2028年的光子学超级周期将非常类似于2025-2026年的内存周期。尤其是随着高盛修正后的估计显示，每GPU-ASIC的光学内容将显著增加……
英文原文
Citibank TMT: "lasers and optical fibers are being positioned as the "next HBM." I've been talking about lasers like $LITE / $AAOI since 2025... and architectural shifts toward CW, ahead of the current industry consensus. And fully agree with Lightmatter/Citi. Just for reference: For lasers: $LITE / $SIVE / $AAOI / $COHR / $MTSI / Furukawa / Sumitomo are your better known ones. For fiber: $GLW / Furukawa / Fujikura / Sumitomo are your more well-known exposure (obviously a lot more in the supply chain) *disclosure, personal exposure to theme With added commentary that: "Tight supply conditions will support a scarcity premium for these assets." We're already seeing price hikes with lasers as seen with $SIVE channel checks at CIOE 2026 Shenzhen today + $LITE commentary from earnings. So I think the 2027–2028 photonics supercycle will look a lot like memory did in 2025–2026. Especially as optical content/GPU-ASIC goes up significantly per Goldman Sachs revised estimates...
-
分析英伟达投资偏好及半导体、机器人供应链关键瓶颈与趋势。
@Bejewele_din_Fr - 英伟达($NVDA) 喜欢投资专用集成电路(ASIC) 公司 - 内存/多层陶瓷电容器(MLCC)/ABF载板/绝缘体上硅(SOI) 因长期可见性而令人兴奋(brrr) - 传统DRAM 因中国内存向高端转移而令人兴奋(brrr) - 中欧就企业问题展开博弈 - 共封装光学(CPO) 时间线保持不变 - 机器人领域的行星滚柱丝杠很酷
英文原文
@Bejewele_din_Fr - Nvidia likes investing in ASIC companies - Memory/MLCC/ABF/SOI brrr from long term visibility - legacy dram to brrr from Chinese memory moving to higher end - China EU fighting over companies - cpo timeline in tact - planetary roller screws in robotics are cool
-
英伟达投资联发科巩固推理地位，或促使AMD与博通结盟。
没错，$NVDA 投资 35 亿美元进入联发科(Mediatek)。 这感觉像是英伟达(Nvidia)正在扮演下一代专用集成电路(ASIC)赢家如 $MRVL / 联发科(Mediatek)的“造王者”，其二级效应是侵蚀 $AVGO 的地位。 这一策略 playbook 类似于他们此前针对像 $NBIS 和 $CRWV 这样的新云厂商(Neoclouds)（相对于超大规模云项目）所采取的做法。 帮助培育下一代领导者，然后使它们在财务和战略上与英伟达保持一致。 我认为这很精彩，但此类举动可能会促使 $AMD、$AVGO 和某些超大规模云厂商(hyperscalers)走得更近。 无论如何，此举巩固了英伟达在推理(inference)领域的战略地位（此前超大规模云厂商的 ASIC 曾是长期的看空论点）。
英文原文
Yep, with $NVDA investing $3.5B into Mediatek. It does feel like Nvidia is playing kingmaker with the next generation of ASIC winners like $MRVL / Mediatek, with a second order effect of eroding $AVGO position. Playbook feels similar to what they did with Neoclouds like $NBIS and $CRWV (vs. hyperscaler cloud programs) Help create the next leaders, then make them both financially and strategically aligned with Nvidia. I do think it’s brilliant, but moves like this might push $AMD, $AVGO, and certain hyperscalers closer together. Regardless this move cements Nvidia’s strategic position in inference (where hyperscaler ASICs were a former long-term bear case).
-
推测Alchip或QCOM将跟随MRVL和联发科趋势，ASIC市场整体扩大。
@balllvest 鉴于 $MRVL 和联发科(Mediatek) 的趋势走向，我猜测 Alchip（披露：我持有该股）或 $QCOM 可能是下一个。不过我认为应用专用集成电路(ASIC) 的市场份额对所有这些公司来说都会扩大。
英文原文
@balllvest With how the trends are going with $MRVL and Mediatek. My guess is Alchip (disclosure, I own this) or $QCOM might be next. I think the ASIC pie gets bigger for all of them though.
-
英伟达投资联发科，整合ASIC生态并确立标准以对抗博通。
今天，$NVDA 宣布向联发科投资 35 亿美元。看来英伟达正通过“结网”融入更广泛的 ASIC（专用集成电路）生态系统，从 $MRVL 到联发科。并让未来的生态系统与英伟达标准保持一致。（也许也是为了抵御 $AVGO）
英文原文
Today, $NVDA has announced a $3.5B investment into Mediatek. It appears Nvidia is "networking" itself into the broader ASIC ecosystem, from $MRVL to MediaTek. And have future ecosystems aligned with Nvidia standards. (maybe to fend off $AVGO as well) https://t.co/okdGKS51U3
-
总结SIVE基本面，批评其沟通策略，看好其作为CPO激光供应商的高增长潜力。
以下是我对 $SIVE 来自 ER 的观点与基本面的最终总结： 2027年可插拔光模块（1.6T 爬坡）： - $JBL（1.6T LRO，已确认）2027年上半年订单，下半年爬坡 -> -> 可能有多家超大规模云厂商客户 - 3家可插拔制造商（评估中） - 3家可插拔制造商（供应评估中） 2027年下半年至2028年+的参考设计： - $GFS（格芯）SCALE（已确认） -> -> 可能包括 $AMD 和其他超大规模云厂商作为共封装光学(CPO)客户 2027年外部光源/光引擎： - $POET（已确认） -> -> Lumilens / 超大规模云厂商终端用户 - O-Net（已确认） -> -> 超大规模云厂商+亚洲地区的巨大ODM/OEM - 可能包括 $AEVA（映射） -> -> 向光引擎供应 -> 超大规模云厂商/激光雷达 - SemiNex（已确认） 2028年光I/O： - Ayar Labs（已确认） -> -> 可能包括 $AMD、ALchip 和超大规模云厂商ASIC项目终端用户 + $NVDA NVLink - $MRVL Celestial（2023-2024年披露的潜在客户） - Lightmatter（2023-2024年披露的潜在客户） - Lightelligence（2023-2024年披露的潜在客户） 随后可能在2028年为 $AAPL 下一代可穿戴设备更新及其他来自 TFLN 与 Lightium 的项目供货。 关于产能分配： - Win Semi - 1家其他晶圆厂（拥有巨大分配量） 在磷化铟(InP)连续波(CW)DFB激光器短缺期间。 这些都是持续的发展/资格认证，因为 Sivers 正瞄准2027-2028年的下一代硅光(SiPH)+CW以及1.6T和CPO。 尤其是高盛模型显示CPO机会将从目前的近乎零增长到2028年约910亿美元的市场规模(TAM)。 然而我的批评是： => 主要焦点未向西方受众传达2027-2028年光学爬坡的经济规模。 => 应停止围绕纳斯达克上市的模糊法律帖子，并设定明确方向+更快执行。 => 过多关注于辩护较小的当前收入/过去十二个月(TTM)收入/管道转化（allspace, Tachyon），相对于未来的资格/合作伙伴规模/产能/潜力。 $SIVE 需要将自己定位为一家面向未来的、全球高增长的光学公司，为超大规模云厂商项目供应激光器。 不要让叙事被向后看的指标主导。 正如摩根士丹利所言……“关键[CPO]参与者包括……$LITE、$COHR 和 Sivers 激光供应”。 我个人是 $SIVE 的愉快股东，以高贝塔值暴露于2027-2028年的1.6T/CPO光学变革。
英文原文
Just my final recap on thoughts + fundamentals of $SIVE from ER: Pluggables (1.6T Ramp) for 2027: - $JBL (1.6T LRO, confirmed) Orders H1 2027, Ramp H2 2027 -> -> Likely multiple hyperscaler customers - 3 pluggable makers (evaluation) - 3 pluggable makers (supply assessment) Reference Design for likely H2 2027–28+: - $GFS (Globalfoundries) SCALE (confirmed)) -> -> Likely $AMD and other hyperscalers as CPO customers External Light Sources / Optical Engines for 2027: - $POET (confirmed) -> -> Lumilens / hyperscaler end users - O-Net (confirmed) -> -> Massive ODM/OEM for Hyperscalers + Asia - Likely $AEVA (mapping) -> -> supplying to OE -> hyperscalers / lidar - SemiNex (confirmed) Optical I/O for 2028: - Ayar Labs (confirmed) -> -> Likely $AMD, ALchip, and hyperscaler ASIC program end users + $NVDA NVLink - $MRVL Celestial (potential customer from 2023-2024 disclosures) - Lightmatter (potential customer from 2023-2024 disclosures) - Lightelligence (potential customer from 2023-2024 disclosures) Then potentially $AAPL for 2028 for next-generation wearable updates and other programs from TFLN with Lightium. In terms of capacity allocations: - Win Semi - 1 other foundry (with tremendous allocations) during an InP CW DFB laser shortage. These are all on-going developments/qualifications since Sivers is targeting next-generation SiPH + CW with 1.6T and CPO for 2027-2028. Especially as Goldman Sachs models the CPO opportunity going from effectively near-zero today to ~$91B TAM by 2028. However my criticisms were: => Main focus was not on communicating economic scale of 2027-2028 optical ramps to Western audiences. => Legal vagueposts around NASDAQ listing should be dropped, and clear direction should be set + executed faster on. =>Too much focus was put on defending smaller current revenue/TTM revenue/pipeline conversion (allspace, Tachyon.) relative to future qualifications/partner size/capacity/potential. $SIVE needs to position themselves as a forward looking, global hypergrowth optical company supplying lasers to hyperscaler programs. And not let the narrative get dominated by backward looking metrics. And as Morgan Stanley put it... "Key [CPO] participants include … $LITE, $COHR, and Sivers laser supply". I'm personally a happy $SIVE shareholder for high-beta exposure to the next 2027-2028 optical shift with 1.6T/CPO.
-
英伟达财报超预期，指引强劲，AI需求无放缓迹象，超额收益转向供应链。
$NVDA 财报要点(TLDR)： 营收：962.2亿美元 vs. 约921.7亿美元（超预期） 调整后每股收益(Adj. EPS)：2.22美元 vs. 2.09-2.10美元（超预期） 数据中心(DC)：890亿美元 vs. 约863亿美元（超预期） 调整后毛利率：75% vs. 75% 超大规模客户(Hyperscaler)营收：487.1亿美元，上季度为430.5亿美元（定制ASIC的增长并未真正阻止该营收加速） 关于指引： 营收：1080亿美元 vs. 约1042亿美元 调整后毛利率：74% vs. 75%（这几乎是唯一的疲软点） 过去4个季度的营收增长确实令人难以置信。 -> 681亿美元 -> 816亿美元 -> 962亿美元（当前） -> 下季度指引：1080亿美元，尽管假设中国数据中心计算营收为零。 有趣的是，英伟达表示其承诺从上一季度的1190亿美元跳升至2790亿美元，主要与未来几年的内存采购有关。所以内存需求会持续飙升。 要点：AI热潮持续不断。英伟达显然处于领先地位，且没有明显的需求放缓迹象。 鉴于英伟达已是5万亿美元+市值的公司，我认为现在的超额收益(alpha)主要来自英伟达的架构/容量决策如何影响供应链的其他环节（例如共封装光学(CPO)、内存、800V电源）。 而不是简单地寻找英伟达自身的定价错误。 无论如何，所有有趣的事情将在几分钟后的财报电话会议中发生。
英文原文
$NVDA earnings TLDR: Revenue: $96.22B vs. ~$92.17B (beat) Adj. EPS: $2.22 vs. $2.09-$2.10 (beat) DC: $89.0B vs. ~$86.3B (beat) Adj. gross margin: 75% vs. 75% Hyperscaler Revenue: $48.71B from $43.05B last quarter (custom ASIC growth hasn't really prevented this revenue from accelerating) For guidance: Revenue: $108B vs. ~$104.2B Adj. gross margin: 74% vs. 75% (kinda the only soft spot) Revenue ramp has been genuinely absurd over last 4 quarters. -> $68.1B -> $81.6B -> $96.2B (we are here) -> next quarter guidance: $108B, despite assuming zero China DC compute revenue. Fun thing to note is Nvidia says its commitments jumped from $119B last quarter to $279B, primarily related to procurement of memory (for next few years).. So memory goes brrr. TLDR: AI keeps on going brrr. Nvidia is clearly leading the charge and no obvious signs of demand slowing. Given Nvidia is already a $5T+ company, I think most of the alpha comes from how Nvidia's architecture/capacity decisions impact elsewhere in the supply chain (eg. CPO, memory, 800V) nowadays. Rather than simply finding mispricing in Nvidia itself. Regardless, all the fun stuff happens in the earnings call in a few min.
-
AI供应链多环节瓶颈显现，内存、基板、组件及先进封装需求强劲，行业加速扩张。
一些我感兴趣的要点摘要： - $SNDK 投资者日显示，预计至2030年调整后毛利率达80%，运营利润率约75%，调整后自由现金流利润率约50%。 长期协议(LTA)已占2028年产能的2/3。最低合同收入达930亿美元（当前市值约2390亿美元）... 当你的收入/目标预计在未来4年持续至2030年时，很难将其视为周期性股票。 - $CRWV 签署合同，将使用6年前的 $NVDA A100 GPU 直至2029年。 对于 Nebius/Iren 等新型云服务商(Neoclouds)而言，这是利好，因为它反驳了如 Burry 的折旧看空论点。 - 据 UBS 估计，传统 DRAM 毛利率（如美光）到2027年将达到前所未有的95%，超越 HBM 毛利率。 如果预测准确，像南亚(Nanya)/华邦(Winbond)这样的传统/标准 DRAM 玩家应该会“起飞”(go brrrr)。 - 据报道，Anthropic 实现了14倍以上的同比增长，Q2环比收入增长约2.4倍至>115亿美元，预计2028年收入达190-2000亿美元。 你的前沿实验室保持疯狂的增长速度，如果不这样反而令人担忧。 - $NVDA 据报道正洽谈向 SB Energy（软银子公司）投资30亿美元，与 Apollo、BlackRock、Blackstone、Brookfield、Goldman 等共同推动 >5000 亿美元的算力融资。$NVDA Feynman 预计2028年下半年转向 TSMC A16 + SoIC + 定制 HBM + 共封装光学(CPO)。 每天都有更多英伟达新闻。 - $MSFT Maia 300 讨论了 $TSM 2027年 >30万颗的产能，并计划扩展至100万+。最快9月发布。 $MRVL 对此消息可能更高兴。关于当前情况： - 可能是微软当前 ASIC 扩产的受益者(GUC)。 - 对于亚马逊阵营，像 Alchip（我持有股份）可能在2026年下半年随 $AMZN ASIC 项目扩产... 所以关注超大规模云厂商 ASIC 扩产时间表及其受益者可能是个好主意。 - $TSM 先进封装副总裁表示“行业未来几年可能不仅面临内存短缺，还将面临 ABF 基板供应紧张”... 强调内存和 ABF 基板在未来几年都是瓶颈。 甚至上游 ABF 基板设备供应商也很高兴，例如使用真空层压设备的 Eternal Precision 表示订单激增，工厂满负荷运行，价格上涨20%+。 - $AMAT 预计2026年先进封装收入增长 >70%（此前预期 >50%），并表示客户讨论已延伸至2030年。 （不太熟悉这家公司，但发现其从2025 Q4 $68亿 -> $70.1亿 -> 79.1亿 -> $91.2亿 -> 2026 Q4 $102.5亿的预测增长率很有趣） - Google 在 OCP APAC 表示传统 48V 已无余量。$NVDA 详细说明了2026年下半年兼容 MGX 的 800VDC 机架（时间线，Delta / Lite-On 受益）。 - 除 $SNDK 外，南亚 LTA 覆盖50%产能。CXMT 上月签署多年 DRAM 协议，因此整个内存行业似乎都在效仿三大巨头的剧本。 - 探针卡(Probe cards)仍是瓶颈，MPI(6223)表示其探针卡产能因需求超过供应而保持满负荷。 本周早些时候已覆盖 CW 激光瓶颈（$AAOI, $SIVE, $LITE），那是另一个有趣的点。 - 据 Nichidenbo 称，部分 MLCC/组件交货期已达36周。 你的三星电机、太阳诱电、村田等玩家听到这个消息应该很高兴。 TLDR: AI 供应链疯狂加速(go brrr)。
英文原文
Just some TLDRs of stuff I found interesting: - $SNDK 80% adjusted gross margins projections through 2030, ~75% operating margins and ~50% adjusted FCF margins from investor day. LTAs already 2/3rd of 2028 output. Minimum contracted revenue reaches $93B (MC is currently ~$239B)... Hard to be a cyclical stock when your revenue/targets are expected to continue 4Y later into 2030. - $CRWV signs contracts for 6Y old $NVDA A100 GPUs through 2029. For Neoclouds like Nebius/Iren, this is positive, since it's a counterargument for eg. Burry depreciation short thesis - conventional DRAM gross margins eg. Micron is estimated to reach an unprecedented 95% by 2027, surpassing HBM GMs per UBS read through for legacy/standard dram players like Nanya/Winbond should go brrrr if projections are correct. - Anthropic reportedly achieved 14x+ YoY growth and roughly 2.4x sequential revenue growth q2 to >$11.5B,. estimating growth to $190–200B in 2028r evenue numbers. Your frontier labs keep growing at stupidly fast paces, it would be worrisome if they didnt. - $NVDA reportedly in talks to invest $3B in SB Energy (Softbank subsidiary), creates a >$500B compute financing push with Apollo, BlackRock, Blackstone, Brookfield, Goldman, and others. $NVDA Feynman reportedly moves to TSMC A16 + SoIC + custom HBM + CPO in H2 2028 Just more nvidia news every day - $MSFT Maia 300 discussed $TSM capacity for >300k units in 2027, with expansion to 1m+. Unveils as soon as September. Likely $MRVL should be more happy from this news. For what's happening right now: - maybe GUC for Microsoft current ASIC ramp. - For the Amazon party, stuff like Alchip (I do own shares), likely is ramping now with $AMZN ASIC program H2 2026... So might be a good idea to look at hyperscaler ASIC ramp timelines + their beneficiaries. - $TSM VP of Advanced Packaging stated "the industry is likely to face not only memory shortages but also tight ABF substrate supply over the next few years"... Emphasis on few years for memory + ABF substrates for bottlenecks. Even upstream abf substrate equipment providers are happy, eg. Eternal Precision which uses vacuum lamination equipment stated orders surged, their plants have been running at full capacity, and 20%+ price hikes. - $AMAT expects advanced packaging revenue to grow >70% in 2026, versus prior >50%, and said customer discussions now extend all the way to 2030 (not too familiar with this company, but found their growth rate from 2025 Q4 $6.8B ->$7.01B -> 7.91B -> $9.12B -> $10.25B Q4 2026 projections pretty interesting) - Google said at OCP APAC said conventional 48V is running out of headroom. $NVDA detailed an 800VDC MGX-compatible rack H2 2026 (timeline, Delta / Lite-On beneficaries) - Aside from $SNDK, Nanya LTAs cover 50% of capacity. CXMT signed multi-year DRAM agreements last month, so entire memory industry seems to be following same playbook as ur big 3. - Probe cards remain a bottleneck, MPI(6223) said their probe card capacity remains fully utilized because demand exceeds supply. Already covered the CW laser bottleneck with $AAOI, $SIVE, and $LITE earlier this week, but that's another fun one. - some MLCC/component lead times have hit 36 weeks per Nichidenbo. Your Samsung Electro-Mechanics, Taiyo Yuden, Murata, players should be very happy to hear this. TLDR: AI supply chains go brrr.
-
博主回顾过往观点遭质疑后获市场验证的经历，强调市场是最终裁判。
我想你一定是新来的。我的许多观点起初都会遭遇强烈反对，尤其是那些更具原创性的观点。$AXTI 遭受无尽谩骂，甚至导致我被 $RDDT WSB 论坛禁言。人们以为它是家骗人的中国公司，但路透社、Epiwafer 财报及机构验证了其磷化铟(InP)衬底地位。$RPI 被所有人称为“梗股”。彭博、金融时报等媒体称其无基本面。分析师称其想法愚蠢且将崩盘。但财报显示其前瞻营收增长58%。他们似乎忘了之前的反对，现在却将其视为高增长AI硬件公司。$SIVE 也被叫作“梗股”，我遭到瑞典媒体大量攻击。人们不懂技术细节，莫名进行人身攻击。但如你所知，这可能是我最成功的观点，已获富达研究、摩根大通机构买入及 $JBL 至 $GFS 的正式合作验证。同样的仇恨也发生在：$AAOI $30时，管理层被称“骗局”；$LITE $300时，光子学被称“泡沫”；$RKLB $20时，被视为低营收泡沫公司，我因发帖被WSB临时禁言；$HOOD $20时，人们想起其冻结GME交易；$IQE $12时，被视为英国无光子学合作的垃圾公司；$SOI $44时，欧洲分析师称其“高估”；$NBIS $75时，$IREN 阵营称我散布俄罗斯宣传且无护城河；$INTC $115时，人们认为其无法与 $TSM 竞争；$MRVL $85时，人们认为其正失去ASIC份额给博通；$AEHR $35时，人们误读财报认为其无营收；$EWY $115时，人们哭诉KOSPI是泡沫且LNG/氦气/石油将颠覆存储贸易。例子不胜枚举……我常看评论，所以记得很多仇恨（我猜是嫉妒、博眼球或缺乏技术深度）。但现在，人们可以看到每个论点被反复验证。市场成功淹没了旧噪音。好消息是，市场是判断对错的最终仲裁者，而非X上的愤怒评论。
英文原文
I think u must be new here. Many of my ideas get intense backlash at the start, especially the more original they are. $AXTI - endless hate to the point I got banned from the $RDDT WSB forum. They thought it was some scam Chinese company, but Reuters, Epiwafer company earnings, and institutions validated their InP substrate position many months later. $RPI - everyone called it some meme stock. Literally Bloomberg, Financial Times, and others went out and called it a meme stock with no fundamentals. Analysts went and said the idea was stupid and said it was going to crash “as a fact”. Earnings came out? Blew away any projection with 58% fwd revenue growth. But they seem to have forgot all the backlash they threw out at me, while they’re citing it as a high growth ai hardware company no. $SIVE? Everyone called it a “meme stock”. I get bunch of hate from Swedish media all the way up. Bunch of people didn’t understand the technical nuances, and they keep throwing personal attacks for some strange reason. But as you know it’s probably my most successful idea and it’s validated so far by institutional buying from Fidelity Research, JP Morgan as well as formally announced partnerships from $JBL to $GFS . Same hate with: - $AAOI at $30, when everyone called management a “scam” or “shady” - $LITE at $300 when everyone called photonics a “bubble” - $RKLB at $20 when everyone thought it was a low revenue launch company that was a bubble. I actually got temp banned from WSB from posting about Rocketlab since moderators didn’t like the stock. - $HOOD at $20 when everyone recalled them freezing GME buys/sells - $IQE at $12 when people thought it was just some random crap $100m company over in the UK with “no actual photonics partnerships” - $SOI at $44 when European bank analysts thought it was “overvalued”and my thesis wasn’t anything new - $NBIS at $75 when everyone in the $IREN camp said I was spreading Russian propaganda and that they had no moat - $INTC at $115 when everyone thought they couldn’t compete with $TSM - $MRVL at $85 when everyone thought they were losing ASIC share to Broadcom. - $AEHR at $35 when everyone misread their earnings and thought they had no revenue - $EWY at $115 when everyone was crying KOSPI was a bubble and LNG/helium/oil would disrupt the memory trade Can go on and on… I do read a lot of the comments, which is why I remember a lot of the hate (probably either jealously, impression farming, or lacking the technical depth is my guess). But I think at this point, people can just see each thesis validated over and over again. And the success in markets drown out the old noise. Good thing is markets are the final arbiter of what’s right or wrong, not the angry comments or posts on X.
-
回顾2025年对光模块及半导体供应链龙头的早期估值判断。
有趣地回顾一下2025年的一些随机想法。 当时， $AAOI 市值为20亿美元， $LITE 市值为260亿美元， $AXTI 市值为5亿美元。 现在： - AAOI 市值为153.7亿美元 - Lumentum 市值为744.7亿美元 - AXT 市值为72.4亿美元 当时由于处于早期主题阶段，掌握的信息较少。 当然，专用集成电路(ASIC)方面存在细微差别。
英文原文
Fun throwback to random ideas back in 2025. Back then, $AAOI was $2B MC, $LITE was a $26B MC, $AXTI was $500M Now: - AAOI is $15.37B - Lumentum is $74.47B - AXT is $7.24B Was working off less information back then, given it was an early theme. Obviously nuances with ASIC
-
通过供应商关系映射推测AI芯片供应链瓶颈。
不客气，关于 $IQE。 -> $LITE 是 $GOOGL 的已知供应商。 -> $IQE 是 $LITE 的已知供应商。 -> $AXTI 是 $IQE 的已知供应商。 Google TPU + 超大规模云服务商 ASIC 的物料清单(BOM) 极度保密。 但你可以通过关系映射大致推测，尽管没人会明说 $AXTI 或 $IQE 供货给...
英文原文
Ur welcome with $IQE
-
长期更偏好英伟达上游瓶颈而非NVDA本身。
回复 @darkseidzz：嗯，长期看，相比做多 $NVDA，我更喜欢你所有的上游瓶颈环节，因为这些会被重估得最多（英伟达已经是世界最大公司）。 我很确定，超大云厂商 ASIC 最终会分流 $NVDA 的需求，比如 $GOOGL 的 TPU、$AMZN 的 Trainium 项目。
英文原文
@darkseidzz hmm, i prefer all your upstream chokepoints over $NVDA long term since those will be re-rated the most (nvidia already largest company in the world) pretty sure hyperscaler ASICs would eventually siphon off $NVDA demand like $GOOGL TPU, $AMZN trainium programs. wouldn't be too
-
称SIVE不是垄断，而是在ASIC/Merchant CPO架构路线具技术主导。
回复 @SidkMena：这不是垄断。只是它在 ASIC / merchant CPO 路由架构上拥有技术主导地位。 如果其他可插拔玩家也用 $SIVE 创造新的 1.6T 架构，我也不会惊讶。
英文原文
@SidkMena It’s not a monopoly. Just technical dominance over ASIC/Merchant CPO route architectures . And wouldn’t be surprised if other pluggable players created novel 1.6T architectures using $SIVE too.
-
称SIVE是最近最明显、仍在重估的CPO激光瓶颈。
回复 @_king142：$SIVE 可能是最近最明显、仍在重估中的激光器瓶颈，因为它似乎出现在每个 $NVDA NVLink CPO 生态伙伴以及 merchant/ASIC CPO 供应链里。 也许下一个是 CPO 中的无源光组件供应链。
英文原文
@_king142 $SIVE was probably the most recent visible laser chokepoint that’s still being rerated since seem to be in every $NVDA NVLink CPO ecosystem partner + merchant/asic cpo supply chains. Maybe passive optical components supply chains in CPO next.
-
没预期英伟达今年入股Marvell，但看好其ASIC/连接收入机会。
回复 @WEB3_furture：好文章！我其实没想到 $NVDA 今年会与 $MRVL 合作并入股。 但无论如何，我本来就预计 Marvell 会因为即将到来的 ASIC/连接收入机会而成为一个有吸引力的想法。
英文原文
@WEB3_furture Great article! I actually didn't expect $NVDA to partner and take a stake in $MRVL this year. But I expected Marvell to be a compelling idea anyway from their upcoming ASIC/connectivity revenue opportunites.
-
说明讨论的是ASIC/merchant NVLink CPO生态伙伴而非英伟达自身。
回复 @siflower：那是 $NVDA 直接相关。我说的是 ASIC/merchant NVLink CPO 生态伙伴。 在这些伙伴里，$SIVE 看起来像是所有人的供应商。
英文原文
@siflower That's $NVDA directly. I'm talking about the ASIC/merchant NVLink CPO ecosystem partners. Where $SIVE looks like they're the supplier to everyone.
-
关注ASIC报告和AlChip通过并购扩展TAM的机会。
我发现一份1月的 ASIC 研究报告挺有意思。 我不太常提 AlChip，但如果他们想像 Marvell 那样做并购，会有很多 TAM 扩张机会。
英文原文
ASIC research report from Jan that I found interesting. I don’t really mention AlChip much but there’s a lot of tam expansion opportunities if they want to M&A like Marvell. https://t.co/oaSMBgO72C
-
回顾AAOI/AXTI早期被FUD，如今类似CPO转折正在SIVE/Foci上重演。
当我在 $AAOI 约20-30美元做多时： 我以为 $AMZN 和 $MSFT 正在为其 ASIC 项目认证特定光收发器。 结果它更可互换、更可大规模生产。 无论如何，很高兴我关于 $LITE、$AAOI、Innolight、$COHR、$AXTI 的 thesis 表现这么好。 记住：当 $AAOI 在20-30美元时，X 上所有人都说“管理层诈骗”…… 或者在 $AXTI 12美元时说“诈骗公司”。 FUD 非常极端。 现在感觉又像似曾相识……做多下一轮 CPO 架构转变，比如 $SIVE 或 Foci？然后收到同样评论。 我们会看看我的 CPO 多头是否会像 Shunsin 对 Innolight、或 $SIVE 对 $LITE 那样兑现。 无论如何，那些名字现在都已经上涨数百到数千个百分点。
英文原文
Back when $AAOI was ~$20-30 when I went long: I thought $AMZN and $MSFT were qualifying specific optical transceivers for their ASIC programs. Turns out it’s more interchangeable/mass producible. Regardless, glad my thesis on $LITE, $AAOI, Innolight, $COHR, $AXTI played out so well. Keep in mind: everyone on X was saying “scam management” with $AAOI back at $20-30… Or “scam company” with $AXTI at $12. FUD was pretty extreme. Feels like dejavu again… going long on the next CPO architectural shift like $SIVE or Foci? And getting the same comments. We’ll see if my CPO longs play out the same way like Shunsin does with Innolight or $SIVE does with $LITE. Either way all of those are now up hundreds to thousands of percent.
-
称SIVE是CPO kingmaker，围绕Lightmatter/Celestial/Ayar/O-Net/Jabil等供应激光。
$SIVE 字面上是 CPO 的 kingmaker： 从早期的 Lightmatter、Celestial、Ayar、Lightelligence，到现在这些估值50-100亿美元以上的领导者。 再到 Marvell、Alchip、GUC 这样的 ASIC 生态。 O-Net 与 Sivers 为 CPO 量产 ELS。 Jabil 与 Sivers 为可插拔光收发器量产可插拔产品。 所有这些很可能都围绕 Sivers 激光，并以它为设计核心。 市场才刚刚开始看到这家激光公司有多重要。
英文原文
$SIVE is literally the kingmaker for CPO: From ~ Lightmatter, Celestial, Ayar, Lightelligence early on. Now the leaders with massive $5-10B+ valuations. To the ASIC ecosystems like Marvell, Alchip, GUC. With O-Net mass producing ELS with Sivers for CPO. And Jabil mass producing pluggables with Sivers for pluggable optical transceivers. All likely surround and are designed around Sivers lasers. Markets are only starting to see the precipice of how important this laser company is.
-
梳理玻璃基板光刻胶、ABF瓶颈和关键矿物供应链。
一些随机笔记： 关于玻璃基板：“YC Chem 据称成为业内首家供应玻璃基板光刻胶的公司”（KR 112290）。 - 目标年底量产，通常量产 + 首发对股价有利。 “目前正与三家以上公司洽谈玻璃基板材料供应”，所以还有更多认证。 “Samyang NC Chem 也在开发玻璃基板光刻胶材料”（482630）。 - 正在送样，明年量产，是另一个玻璃基板名字。 “$NVDA 可能通过长期协议、预付款或战略合作提前锁定高端基板产能，目标是避免 CoWoS 和 HBM 过去出现的供应限制重演。”这与 $TSM COUPE 基板有关。 “Commercial Times 强调，AI GPU、ASIC 和 CPO 需求可能在2027年前触发另一轮高端 ABF 基板短缺。” 所以还有更多基板瓶颈。 另外还有：“中国同意解决美国对稀土和关键矿物短缺的担忧，特别是钇、钪、钕和铟。” 如果你想找多头想法，可以研究钇、钪、钕和铟供应链。 现在一切都在回调，但如果你对更多随机想法感兴趣，这是值得关注的东西。
英文原文
Just some random notes: For glass substrates: "YC Chem has reportedly become the first in the industry to supply photoresists for glass substrates" (KR 112290) - targeting EOY mass production, usually mass production + first to market is good for stock prices. "is currently in discussions with more than three companies regarding the supply of glass substrate materials" so more qualifications. "Samyang NC Chem is also developing photoresist materials for glass substrates" (482630) - Sampling, mass production next year another glass substrate name " $NVDA may seek to secure high-end substrate capacity in advance through long-term agreements, prepayments, or strategic partnerships, aiming to avoid a repeat of the supply constraints previously seen in CoWoS and HBM." This is in relation to $TSM COUPE on Substrate. "As Commercial Times highlights, demand from AI GPUs, ASICs, and CPO could trigger another supply shortage of high-end ABF substrates by 2027." So more substrate bottleneck stuff. Then there's: "China's agreement to address US concerns over rare earth and critical mineral shortages specifically yttrium, scandium, neodymium and indium" So good to look into yttrium, scandium, neodymium and indium supply chains if you want long ideas. Everything is kinda correcting right now, but just something to keep a lookout for if ur interested in more random ideas.
-
用 napkin math 算 Nextronics 的 CPO 价值
所以这是我在做多 Nextronics（8147）时做的便签纸计算。 他们是 $NVDA 的 CPO 供应商，供应 CPO connectors 和 cage thermal modules。 我按 2028 年的前瞻市盈率大概算了 2 倍，这也是我认为它有望在 2028 年重估到约 20 亿美元以上市值、具有很强风险回报的原因。 只看它的 CPO 暴露： -> CPO connector 大概 15 到 25 美元 -> ELS thermal cages 估计约 50 美元 每个 switch 18 个单元：18x50 = 900 美元 CPO Connectors：每个 switch 72 个 optical engines，72 x 15 = 1080 美元 （如果 $NVDA 扩到 Spectrum-X switch，这个数会到 1920 美元） Total Nextronics Content：大约 1980 美元（保守算成 2000 美元） 这意味着 BOM 只占 rack 的 0.08%，大概只占 switch 的 1.5%。 机构眼里这太小了，所以大概率被忽略。 但对 Nextronics 这家约 2 亿美元的公司来说，这很重要吗？ 当然，绝对很重要。 我算的时候把 Nextronics 在 Nvidia connector/cage 市场的份额打了 50% 折扣，因为存在多供应商。 并且我用的是 GS 的预测，再假设 $AVGO、$MRVL、ASIC CPO 生态的规模只有 $NVDA 的 30%。 净利润率：22.4%（GM 38% 时）到 24.0%（GM 40% 时）。 但如果按其他预测，只看 rack shipment： 2026：CPO revenue 约 1010 万，净利润（22.4%）约 226 万 + 基础业务 1250 万 = 1470 万（540k 单元是 connectors，40k 单元是 cage，且已除以 50%） 2027：CPO revenue 约 1.72 亿，净利润（22.4%）约 3853 万 = 5103 万（约 800 万单元 connectors，103 万单元 cage） 2028 scale up 扩张：CPO revenue 约 4.5 亿，净利润约 1.0093 亿，基础业务约 1130 万（约 4000 万单元 connectors，298 万单元 cages，例如 Nvidia ELS 量是 1990 万） 所以前瞻市盈率大概是 2026 年 15.4 倍，2027 年 4.45 倍，2028 年 2 倍。 当然，放量后综合毛利可能会下降，也可能会有别的玩家把份额稀释到 25% 左右，预测也可能比 GS 更高或更低。 但不管怎样，哪怕我偏差 50%，这仍然极具不对称性。 2028 年通常是 CPO 玩家大重估的一年，2026 年其实还很早。 希望我算得对，但 20x 的前瞻市盈率会对应 22.6 亿美元市值。 即便我们把： -> 市占率降到 15%。 -> 净利润率压到 14%。 -> connector ASP 降到 10 美元。 按 20x 倍数，股价仍然能有大约 4.5 倍回报，市值超过 10 亿美元。 我们走着看吧。（NFA，只是投机性财务建模）
英文原文
So here's the napkin math I did on Nextronics (8147) when I went long. They're the $NVDA CPO supplier for CPO connectors and cage thermal modules. And I modeled around 2 FWD p/e for 2028, which is why I think risk-reward is very compelling for a potential 10x rerating to ~$2B+ MC in 2028. Just for their CPO exposure: -> CPO connector runs roughly $15 to $25 -> ELS thermal cages, maybe ~$50 from est. 18 units per switch: 18x50 = ~$900 CPO Connectors: 72 Optical Engines per switch 72 x $15 = $1,080 (If $NVDA scales their Spectrum-X switch, it goes to $1,920 for CPO connectors). Total Nextronics Content: ~$1,980 (rounded to $2k for calculations) in conservative case. Implied BOM % of rack: 0.08%. Maybe ~1.5% of switch. This looks microscopic to institutions so it probably is ignored. Is it material to Nextronics, a ~$200m company? Yes, absolute massive. For calculations: Applying 50% haircut to Nextronics' share of the Nvidia connector market/cage market because of multi-source. And I’m using GS projections, and assuming $AVGO, $MRVL, ASIC CPO ecosystem is 30% size of $NVDA. Net Income Margin: 22.4% (at 38% GM)- 24.0% (at 40% GM). But going off other projections from just, a rack shipments: 2026: CPO revenue ~10.1M, net income (22.4%) ~2.26M + $12.5M base = $14.7M (540k units for connectors, cage, 40K units, already divided by 50%) 2027: CPO revenue: ~$172M, net income (22.4%): ~$38.53M = $51.03M (~8M units for connectors, ~1.03M units for cage) 2028 scale up expansion: CPO revenue: $450m, net income: $100.93M, ~$11.3M base (~40M units for connectors, 2.98M unit for cages, eg. Nvidia ELS volume is 19.9M) So implied fwd p/e 15.4x for 2026, 4.45x for 2027, 2x for 2028. Of course at scale, blended margins might go down, there might be other players bringing market share down to like 25%, etc. and projections might be more or less than GS. But regardless seems highly asymmetrical even if I'm off by a whole 50%. 2028 is usually the massive re-rating for CPO players, 2026 is still really early. Hope my math is right, but 20x fwd p/e multiple would be $2.26B MC. Even if we drop: -> market share to just 15%. -> compress their net income margin down to 14%. -> connector ASP to $10. At a 20x multiple, the stock would still achieve a ~4.5x return to a $1B+ market cap. We'll see if this is right or not. (NFA, just speculative financial modeling)
-
SIVE 业务很分散
哦，$SIVE 肯定没问题，因为它们已经高度多元化了。 它们大概基本上服务所有人： -> $AAPL 硅光 -> $JBL（可能最大） -> $AMD CPO 项目配合 $GFS， -> Ayar -> Alchip/GUC 和其他 ASIC 设计商 -> O-Net ELS 还有其他客户。虽然这会在实质上让 2028-2029 年后期的预测稍微往后推一些（比如 Q4 5 亿美元、2029 年 10 亿美元）。 但如果它们的光源最终以某种方式进了 Marvell，我也不会惊讶。
英文原文
Oh $SIVE is definitely fine, since they're vastly diversified. They likely basically serve everyone: -> $AAPL Silicon Photonics -> $JBL (probably largest) -> $AMD CPO program with $GFS, -> Ayar -> Alchip/GUC and other ASIC designers -> O-Net ELS amongst others. Though this does materially slow later 2028-2029 projections a bit (eg. $500m Q4 Celestial, $1B 2029). But wouldn't be surprised if their light source ends up in Marvell one way or another.
-
CPU 开始像存储一样短缺
CPU “严重短缺” 现在看着 CPU 一天天变得越来越像存储，挺好笑的，像 $SNDK 那样不断涨价。 “自 3 月以来，服务器 CPU 已上涨 10% 到 20%” $INTC 和 $AMD 计划在 Q3 继续涨价。 “CPU 供应很可能在 2026 和 2027 年继续紧张”，其中“Intel 的服务器 CPU 短缺可能长达 6 个月，AMD 则为 8 到 12 周”。 $TSM 也在因这波短缺而增加 capex。 “恐慌的主要原因是 CPU 和 AI ASIC 需求同时爆发；目前库存包括 Intel、AMD 主流代际 CPU，以及 $NVDA 即将推出的 Vera CPU，全部都在 3nm 工艺上。” 看起来我们又绕回来了？
英文原文
CPU "serious supply shortage" It's still funny to see CPUs become more like memory day-by-day with the price hikes like $SNDK. "Server CPUs have jumped between 10% and 20% since March" with $INTC and $AMD planning to further price hikes Q3. "CPU supply is likely to remain tight through 2026 and 2027" with "server CPU shortages reaching up to six months for Intel and 8-12 weeks for AMD" $TSM is also increasing capex in response to this shortage. "The main reason behind the fear is the simultaneous outbreak of CPU and AI ASIC demand; inventory currently includes Intel, AMD mainstream generation CPU, and $NVDA's upcoming Vera CPU, all using 3nm process Looks like we've come full circle?
-
AMD 的 CPO 供应链
哇……如果 AMD 走 Ayar 路线，并在 $GFS 上推进，那么 Sivers 看起来就是 $AMD CPO 项目的主要激光供应商。 有意思的是：$LITE 和 $MTSI 曾经被悄悄地、按顺序从 Ayar 的网站上移除了。 （特别感谢我的一位关注者 Setian 发来私信。） 所以 $SIVE 很可能先成为 Ayar 一代产品的主要激光供应商……进而成为经由 Alchip 或其他 ASIC 设计公司流转的所有 Ayar 客户的供应商。 Ayar 上个月还为了量产融资 5 亿美元，而 $SIVE 已经被设计进去。 这个发现对 $SIVE 来说结构上非常重大，因为它削弱了 $LITE 等公司作为主要供应商的叙事。 尤其也削弱了市场对 $SIVE 在 CPO 里只占很小激光份额的说法，如果它们很可能就是主要激光供应商的话。
英文原文
Woah... Sivers looks like the primary laser supplier to $AMD's CPO program if AMD goes the Ayar route at $GFS. The interesting thing is: $LITE and $MTSI were silently removed from Ayar's website sequentially over time. (Special thanks to one of my followers Setian for the DM.) So $SIVE likely became the primary laser supplier for Ayar first-gen... and by extension for all of Ayar's customers routing through Alchip or other ASIC design firms. Ayar also raised $500M last month for VOLUME PRODUCTION, where $SIVE is designed in. This find looks structurally massive for $SIVE as it undercuts narratives about $LITE and others being primary suppliers. And especially about $SIVE having only a small % laser share for CPO if they're likely to be the primary laser supplier.
-
主题框架
CPO / 1.6T。 基板（玻璃核心、InP、SoI）。 CPU / ASIC。 变压器。
英文原文
CPO / 1.6T. Substrates (Glass Core, InP, SoI). CPUs / ASICs. Transformers.
-
GOOGL ASIC 新闻对 SIVE 不够计价
我很惊讶 $SIVE 没有因为 $GOOGL 和 Marvell 的 ASIC 新闻而涨得更高。 甚至连 $MRVL 隔夜都涨了 6.24%+。 市场大概还不知道上游激光源到底来自哪里，对吧？ https://t.co/0Hgfo0dQLo
英文原文
Surprised $SIVE isn’t up higher from the $GOOGL ASIC news with Marvell. Even $MRVL is up 6.24%+ overnight. Markets still probably don’t know where the upstream laser source comes from yet? https://t.co/0Hgfo0dQLo
-
Marvell与Google洽谈TPU开发，Sivers有望成为光子光源供应商
$MRVL正在洽谈为$GOOGL开发TPU（张量处理单元），涵盖两种不同的芯片型号。 如果谈判顺利，$SIVE很可能成为Google供应链的光源供应商。 尤其因为其定制ASIC（专用集成电路）可能推动Celestial IP在MPU（微处理器单元）架构中实现规模扩展（光子结构，Photonic Fabric）。这与$POET和Sivers两家公司的布局都相吻合。 定制ASIC需要数年时间，因此来自Google的实质性营收可能要等到2028年下半年。 供应商可能采用多元采购策略，但拥有Google作为潜在终端客户是一个超越2028年的结构性多年催化剂。 市场通常具有前瞻性。
英文原文
$MRVL in talks to develop $GOOGL TPU for two different chip models. High possibility $SIVE likely became the light source for Google's supply chains if talks go well. Especially as their custom ASIC likely pushes Celestial IP for scale up (Photonic Fabric) in the MPU's architecture. And this maps to both $POET and Sivers. Custom ASICs take years, so there's probably no material revenue from Google until H2 2028. Vendors are also probably multi-sourced, but having Google as a likely end user is a structural multi-year catalyst past 2028. Markets are typically forward looking.
-
认为 Enplas 仍有双重瓶颈，值得在中小市值里关注。
@Letterkees 我还是认为 Enplas 会在 6 个月内翻倍。 通常你不会看到服务 hyperscaler 的 <10 亿美元公司长期还停留在那个市值区间，而且它们在 photonics / ASIC-GPU 测试上握着两个不同的瓶颈。
英文原文
@Letterkees I still think Enplas doubles within 6 months. Normally you don't see <$1B companies serving to hyperscalers stay in that MC range for too long and they own two different chokepoints in photonics/ASIC-GPU testing.
-
认为 Enplas 的双瓶颈与客户结构很有意思。
Enplas（6961）在大约 9.85 亿美元市值时看起来挺有意思。 它们握着两个瓶颈： 1. MLA（微透镜阵列）的主导供应商。比如 SiPH switch、1.6T、3.2T 光子产品。 2. IC 测试 socket 的寡头供应商（AI 芯片测试） 手头现金大约 1.55 亿美元，没有债务。股东权益/资产比大约 89%（一般 50%+ 就算稳），所以我觉得下行风险很低。 而且它们推测的客户是： 1. 光子（MLA）：高度可能是 Innolight、Eoptolink、Furukawa、Intel（SiPh）。大概率还有 $COHR、$LITE。 2. GPU/ASIC（测试 socket）：大概率是 $TSM、ASE，以及这些类型用来测试 $NVDA GPU、$GOOGL TPU（Google Ironwood 很可能就是这个方向）。 “为主要 GPU 制造商以及 hyperscaler 的 ASIC 相关项目扩大量产订单。” 然后在 OFC： OFC： “我们将展示支持当今 AI 快速技术进步的 800Gbps 和 1.6Tbps 收发器以及 CPO（共封装光学）的新产品。” 本质上你得到的是一家同时服务一线半导体、晶圆厂和 hyperscaler 的公司，并且它能同时受益于 MLA 光子段带来的 1.6T，以及之后 OFC 产品带来的 CPO TAM 扩张。 这是我想分享的一笔持仓，因为它在不到 10 亿美元市值时就显得挺酷。
英文原文
Enplas (6961) at ~$985M seems kinda interesting. They hold two chokepoints: 1. Dominant supplier for MLAs (micro lens arrays). Eg. SiPH switches, 1.6T, 3.2T for photonics. 2. Oligopoly supplier for IC Test sockets (AI Chip Testing) Cash on hand: ~$155M, no debt. Equity-to-Asset Ratio: ~89% (50%+ usually is solid), so low downside risk imo. And their speculated customers: 1, Photonics (MLAs): Highly probable: Innolight, Eoptolink, Furukawa, Intel (SiPh). And prob $COHR, $LITE. 2. GPU/ASICs (Test Sockets): prob $TSM, ASE, types use these to likely test $NVDA GPUs, $GOOGL TPUs (Google Ironwood is highly probable). "expanding mass production orders for major GPU manufacturers, and for ASIC-related projects for hyperscalers." Then at OFC: OFC: “We will be showcasing our new products for 800Gbps and 1.6Tbps transceivers and CPO (Co-Packaged Optics) that support today’s rapid technological advancements in AI.” Basically you have a company that supplies T1 semis, foundries, hyperscalers, that benefits from 1.6T from MLA photonics segment + CPO TAM expansion later from OFC products. Was one my positions wanted to share my thoughts about since it seemed p cool at sub <$1B MC.
-
用 1.69 亿美元撬动西方 hyperscaler 供应链。
Serenity 指南： 如何只用 1.7 亿美元就让西方 hyperscaler buildout 变得更脆弱。 只要用 1.69 亿美元拿下 Nippon Chemical（4092）！ 做 InP 衬底需要：铟和高纯磷。 你以为 $AXTI 是瓶颈？NCI 才是瓶颈里的瓶颈。 NCI 实际上是高纯红磷卡点的领头羊，份额大约 26-27%（Rasa 更少，其余就是中国）。 而且它们向 $AXTI、Sumitomo、JX 供应这些材料，用来做 InP 衬底。 所以……如果你拿出 1.6 亿美元收购 NCI（再加上 Rasa 作为更小产能），你就能移除西方世界生产 6N/7N 红磷、也就是制造 InP 衬底所需的主力产能！ 没有 InP 衬底：就没有 photonics。 有趣的是：中国科技公司其实也会被 NCI 打乱。 对 $AXTI 来说，这个映射 / 依赖关系其实非常有意思： - AXT 的 Tongmei 在 STAR Market 上市时就披露过，它在结构上依赖从日本进口高纯前驱体材料 - WITS 数据显示，大约 460 美元 / 公斤的高纯磷正从日本流向中国 所以它们确实在暗中依赖 NCI。 中国虽然也有 Wylton Chemical、Qin Xi New Materials、Jinding Electronics、Chuxiong Chuanzhi、Guizhou Wylton Jinglin Electronic Materials 这类产能。 但这些都还是更小的玩家，没法在规模化 InP 衬底生产上替代 NCI 提供的高纯红磷产能。 $LITE 的 CEO 之前就说过 InP 衬底让他夜不能寐。 所以现在有了 NCI，你就能让他永久失眠？ 只要 1.69 亿美元。 所以供应链大概是这样： -> DGC 磷矿开采后运给 NCI -> NCI 把黄磷精炼成高纯红磷 -> Sumitomo / JX / AXT 把红磷和铟熔合，长成 InP 衬底 -> $COHR / $LITE 把 InP 衬底做成激光器 -> Innolight / Fabrinet 把它们封装成 800G / 1.6T 收发器 -> $NVDA / $GOOGL 用在 ASIC / GPU 集群里。 基本上，整个西方都依赖 NCI 来制造用于 photonics 的 InP 衬底。 我持有一点很小的仓位，纯属好玩。不过日本并不以随意提价著称。 所以如果你真的买下这家公司然后把价格抬高 15000%（比如政府意识到后把公司收回去），大概率会碰到监管问题，比如 FEFTA…… 也许提价 30-50% 还能压缩一下 forward P/E？但大概率不会最后变成 $AXTI 那样。 不管怎样，这家公司按大概 1.6 亿美元市值定价，本身就是一个巨大的国家安全风险。 至于基本面，它们市净率 0.54、forward P/E 11.4，所以其实也可能是被低估了。 TLDR： -> 它是下一个 $AXTI 吗？不是。 -> 它是不是西方 hyperscaler buildout + photonics 里一个未知的结构性瓶颈和关键漏洞？是。 -> 还有没有重估空间？ 只要回到 1 倍市净率，立刻就有 80-85% 的上涨空间。如果再给它更高的 forward P/E，可能更多。 总之，用 1.69 亿美元找到 hyperscaler 供应链里的一个主要失效点，挺好玩的。
英文原文
A Guide by Serenity: How to Cripple the Western Hyperscaler buildout with just $170m. Just take over Nippon Chemical (4092) with $169m! For InP substrates, you need: Indium and High Purity Phosphorus. Thought $AXTI was a bottleneck? NCI is the bottleneck of the bottleneck. NCI is actually the leader of the high purity red phosphorus chokepoint holding 26-27% of the market share (Rasa has less share, then the rest is China). And they export to $AXTI, Sumitomo, JX that need it to make InP substrates. So… if you have $160m to spend to acquire NCI (plus Rasa as smaller capacity), you can remove the leading Western world’s production of 6N/7N red phosphorus needed to make InP substrates! And without InP substrates: no photonics. Fun fact: China’s tech companies would get pretty disrupted with it too by NCI. For $AXTI, the mapping/reliance is actually pretty interesting: - AXT's Tongmei outlined its structural reliance on importing high-purity precursor materials from Japan on their STAR Market listing - WITS data showing ~$460/kg high-purity phosphorus flowing from Japan into China So they secretly do depend on NCI. China does have capacity like Wylton Chemical, Qin Xi New Materials, Jinding Electronics, and Chuxiong Chuanzhi, Guizhou Wylton Jinglin Electronic Materials as well. However, they’re all smaller players so can’t make up for high purity red phosphorus capacity provided by NCI for InP substrate production at scale. $LITE CEO already said inp substrates keeps him up at night. So now with NCI, you can give the guy permanent insomnia? For just $169M. So here's what the supply chain looks like: -> DGC phosphate rock mine and ships it to NCI -> NCI refines Yellow Phosphorus into High Purity Red Phosphorus -> Sumitomo / JX / AXT melt the Red Phosphorus with Indium to grow InP Substrates -> $COHR / $LITE fab InP substrates into Lasers -> Innolight/Fabrinet package them into 800G/1.6T transceivers -> $NVDA / $GOOGL use them for ASIC/GPU clusters. And basically, the entire West depends on NCI to make InP substrates for photonics. I hold some very small positions, just for fun. However, Japan is not well known for price hiking. So you’d probably run into regulatory problems eg. FEFTA if you bought the company and hiked prices 15000% (like government seizing back the company once they realize)… Maybe 30-50% hikes is possible to compress fwd p/e? But very likely wont end up like $AXTI. Regardless, this company is a massive, massive national security risk priced at ~$160m. As for fundamentals, they’re trading at .54 book value and a forward P/E of 11.4 so it’s probably undervalued anyway. TLDR: -> Is it the next $AXTI? No. -> Is it an unknown structural bottleneck + critical vulnerability of the Western hyperscaler buildout with photonics? Yes. -> Is there still room for re-rating? Just reverting to Book Value of 1 is immediate 80-85% upside. Maybe more if you give it multiples past 11 fwd p/e. Regardless, it’s fun to find a major point of failure in the hyperscaler supply chains for $169m.