供应链分析
产业链结构、上下游与瓶颈环节研究 · 共 1122 条 · 第 15 / 23 页
-
澄清超大规模云厂商订单无取消风险,强调美国本土供应链优势及光子学瓶颈。
超大规模云服务商提前一年下满所有产能的订单,并不意味着有取消风险?我不确定你从哪得出这个结论。 什么20%的利润率上限?他们提到的是40%。 此外,它们拥有“美国制造”的供应链,这很罕见,$LITE 和 $COHR 的组装都在亚洲进行。 但如果人们因多年前 $AMZN 取消订单而患有创伤后应激障碍(PTSD),那也能理解。但这次应该不同,因为光子学(Photonics)是已知的瓶颈,且需求超过供应。
英文原文
Three to four hyperscaler orders for all the capacity they can make a year out doesn’t mean risk of cancellation? Not sure where you’re getting that from. What 20% margin ceiling? They cited 40%. Also they’re Made in America supply chains which is rare, $LITE and $COHR do assembly in Asia. But makes sense of people have ptsd from $AMZN order cancellation many years ago. But this time should be different since photonics is a known bottleneck and demand outstrips supply
-
光子学板块股票正经历剧烈重新定价。
@LIWEILIU476419 所有光子学(Photonics)板块股票正在经历剧烈的重新定价。 https://t.co/BBMsC9yQI8
英文原文
@LIWEILIU476419 Wild repricing event for all photonics sector stocks. https://t.co/BBMsC9yQI8
-
伊朗战争或消除以色列及关联股的生存折价,需进一步研究。
我也没料到这一点。也许伊朗战争的后果消除了相关股票如 $NBIS 以及以色列股票如 $ETOR 至 $VLN 的“生存折价因子”(existential discount factor)? 或者情况是多方面的,因为国防在以色列股市中占很大权重。 鉴于这对我来说是新领域,仍在做更多研究。
英文原文
I was not expecting this as well. Maybe the implications of the Iran war removes Israel’s existential discount factor on correlated stocks like $NBIS and Israeli ones like $ETOR to $VLN? Or its multifaceted since defense makes up a large weighting of the Israel stock exchange. Still doing more research since this is new to me.
-
三星与海力士大幅上调DRAM价格,NAND涨500%,内存需求强劲。
据 Digitimes 今日报道,三星电子(Samsung Electronics)和 SK 海力士(SK Hynix)大幅上调 DRAM 价格。 部分消息源自本周《首尔经济日报》的报道,其中指出: “客户可能需要接受比之前合同价格高出一倍以上的价格,才能获得 DRAM 配额”。 DRAM 价格的急剧上涨伴随着 NAND 价格的上涨,今日“群联(Phison)因 NAND 价格上涨 500% 而转向预付款模式”。 内存需求存在一个无情的黑洞。
英文原文
Samsung Electronics and SK Hynix reportedly sharp DRAM price increases from Digitimes today. Part of source was the Seoul Economic Daily report this week that stated: “Customers may need to accept prices more than double their previous contract rates to receive DRAM allocations”. The sharp DRAM price hikes come with NAND price hikes as “Phison shifts to prepayments as NAND prices hike 500%” today. There is a relentless black hole on memory demand.
-
Soitec对CPO至关重要,虽早但不可或缺
@beauty_oe 我觉得索伊特克(Soitec)现在投资还稍微有点早。所以我原本打算明年再投资。不过,对于预计在2028年左右普及的共封装光学(CPO)来说,它似乎将成为一个巨大的瓶颈(不可或缺的存在)!
英文原文
@beauty_oe ソイテック(Soitec)はまだ少し早すぎる気がします。なので、投資は来年にしようと思っていました。でも、2028年頃に普及するCPO(共同パッケージ光技術)にとっては、大きなボトルネック(不可欠な存在)になりそうですね!
-
Soitec是CPO关键瓶颈,预计2028年随CPO拐点迎来估值重估。
Soitec ($SLOIF) 是 $AXTI 在硅光子学领域的垄断性镜像标的。 基本上,它与共封装光学(CPO) 直接绑定,其拐点预计在 2027 年底至 2028 年。 对于 800G/1.6T 和 CPO 而言,没有 $AXTI 和 Soitec 很难实现: AXT 提供磷化铟(InP) Soitec 提供光子学绝缘体上硅(Photonics-SOI) 衬底。具有功能性垄断地位。 (Shin Etsu 在其授权下运营)->(英特尔、格芯、Tower) 但预计当 CPO 在 2028 年初触及拐点时,它们的估值将大幅重估。 然而,与 $IQE 类似,大量传统业务拖累影响着其股价。 对于 $AXTI,供应挤压正在发生。 对于 Soitec,可能早了 6 个月或一年,但不知道市场何时开始抢跑。 但这只是将其作为一个未来的瓶颈环节提出。
英文原文
Soitec ( $SLOIF ) is the monopoly silicon photonics mirror to $AXTI. Basically it’s directly tethered to CPO and its inflection point late 2027-2028 For 800g/1.6T and CPO, hard to really do it without $AXTI and Soitec: AXT for InP Soitec for photonics-soi substrates. Functional monopoly (Shin Etsy operates under their license) -> (Intel, Global Foundaties, Tower) But expect them to get rerated extremely hard early 2028 when CPO hits the inflection point. However similar to $IQE, there’s tons of legacy drag affecting their stock price. For $AXTI the supply squeeze is happening now. For Soitec might be 6 months or a year too early but don’t know when markets start frontrunning it. But just putting it out there as a future bottleneck.
-
从半导体瓶颈向运行瓶颈的板块轮动逻辑合理。
@zephyr_z9 从半导体瓶颈(SeMI bottlenecks)板块轮动至运行瓶颈(Running bottlenecks)板块是合理的。
英文原文
@zephyr_z9 Rotation from semi bottlenecks into running bottlenecks makes sense
-
NOR Flash涨价潮确认,GB200单机架成本激增,推荐宏碁与华邦。
NOR Flash 会是继 HBM/DRAM 之后的下一个内存供应瓶颈吗? 近期,MCU(微控制器)、NOR Flash、KGD(裸晶)封装存储以及功率器件的价格上涨了 10% 至 80%。 涉及涨价的公司名单包括: - 华润微 - 中芯国际 - 北京君正 - 戈麦斯微电子 - 盈佳芯 - 兆易创新 - 欧姆龙 - ADI - 英飞凌 所有这些公司都提高了价格。更具体地说,中国的兆易创新将 MCU 和 NOR Flash 的价格提高了 15%–50%,普冉半导体也提高了 NOR Flash 的价格。 这证实了此前关于 $NVDA GB200 NVL72 系统的预测,即每个机架的 NOR 内容成本已超过 600 美元,并可能开始上调至 900 美元(增加 50%)。 投资者通常喜欢听到涨价的声音,因此对于相关敞口: - 宏碁半导体 (YTD: +153%) - 华邦电子 (YTD: +34.91%) 是该板块的两个最佳标的。 因为它们各自拥有 33-35% 的市场份额,尽管 TAM(总可寻址市场)较小。英飞凌等其他公司对于纯主题投资来说只是杯水车薪。
英文原文
Is NOR Flash the next memory supply squeeze from HBM/DRAM? Recently 10% to 80% hikes were made to: MCUs, NOR Flash, KGD package storage, and power devices recently. The list of companies spanning: - China Resources Microelectronics - SMIC Semiconductor - Belling - Goke Microelectronics - Yingjixin - Maxscend - Omron - ADI - Infineon have all increased prices. More specifically, China's Cmsemicon raised MCU and NOR Flash prices by 15%–50% and Puya Semiconductor raised NOR Flash prices. This helped confirm earlier projections regarding $NVDA GB200 NVL72 systems that NOR content per rack already exceeds $600 and could start hiking rates to $900 (+50% increase). Investors usually like to hear the sound of price hikes so for exposure: - Macronix (YTD: +153%) - Winbond (YTD: +34.91%) are the two best exposures for the segment. Since they both hold 33-35% of market share each, though a smaller TAM. Infineon and others look to a drop in the bucket for pure play exposure.
-
列举AI半导体供应链各环节代表公司及代码
@jahintanvir_ 大概包括这些: 内存 - $MU, 三星, SK海力士, 闪迪 晶圆代工, 封装 - $INTC, $TSM, $AMKR 光子学(Photonics) - $LITE, $COHR, $AAOI 专用集成电路(ASICs) - $AVGO, $MRVL 连接 - $ALAB, $CRDO 电力/电网 - $XLU 数据中心 - $NBIS, $CIFR, $WULF
英文原文
@jahintanvir_ Probably the likes of these: Memory - $MU, Samsung, SK Hynix, Sandisk Foundries, Packaging- $INTC, $TSM, $AMKR Photonics - $LITE, $COHR, $AAOI ASICs - $AVGO, $MRVL Connectivity - $ALAB, $CRDO Power/Grid- $XLU Datacenters - $NBIS, $CIFR, $WULF
-
AI光模块供不应求,AAOI具美国本土供应链优势。
- 即使所有人都生产 800G/1.6T 产品,供应依然不足。需求远超供应。 这并非三星/SK海力士/美光(Micron)的三巨头格局,但 $AAOI 可能是唯一来自美国本土供应链的企业,因为 $COHR 在马来西亚组装,而 $LITE 则在泰国($FN)及亚洲其他地区生产。 - 在他们的电话会议中,他们详细说明了超大规模云服务商(Hyperscalers)希望购买其能生产的所有产品。这更多是关于产能爬坡(Production Ramp),并进一步印证了第一点:全行业需求大于供应。 我认为我们不会看到产品泛滥,因为需求将在可预见的未来消化掉各家厂商的所有产能。此外,中国厂商面临类似存储领域 cmxt/YMTC 的出口管制风险的可能性也不为零。
英文原文
- even if everyone produces 800g/1.6T there’s not enough. Demand far outstrips supply. It’s not quite the trio Samsung/sk Hynix/micron but $AAOI is probably the only one from made in America supply chains since $COHR does assembly in Malaysia and $LITE -> $FN Thailand and other places in Asia. - in their call they detailed hyperscalers wanted to buy anything they can make. It’s moreso production ramp and just drives the first point home that demand > supply across the board. Don’t think we’ll see flooding bc demand will eat up anything the players can make for the foreseeable future. And nonzero chance there’s export controls like cmxt/YMTC from memory on Chinese players
-
光模块供需缺口持续,2028年迎增长高峰,相关股正重定价。
我觉得这很合理。超大规模云服务商对经过认证的800G或1.6T光模块(Optical Transceivers)的需求严重超出供应,就像内存(Memory)一样,这是意料之中的。 而且现在才2026年,最大的增长加速期预计发生在2028年,这也是$LITE、$COHR等公司目前正处于重定价(Repricing)周期的主要原因。
英文原文
Seems reasonable to me. It's expected hyperscaler demand for qualified 800G or 1.6T optical transceivers severely outstrips supply like memory. And it's only 2026, biggest growth ramp is expected to happen 2028, which is the largest part of why $LITE, $COHR, and others are going through a repricing cycle right now
-
AI光模块产能瓶颈持续,云厂商订单不减,风险收益比极佳。
是的,很多人对管理层过去的问题有创伤后应激障碍(PTSD)。我只是觉得,如果他们能实现甚至65%的指引,风险收益比就极其不对称。此外,我认为像$AMZN这样的超大规模云服务商这次不会削减订单,因为800G/1.6T光模块确实是瓶颈,且产能缓解遥遥无期。
英文原文
Yep lot of people have PTSD from management in the past. I just felt the risk-reward was extremely asymmetrical if they deliver even 65%% of their guidance. Also I do feel like hyperscalers like $AMZN won't pull back orders this time as 800g/1.6T it's genuinely a bottleneck with no end to capacity relief in sight.
-
推荐关注Oxford Instruments,其为COHR/AAOI提供刻蚀沉积设备。
@Investmnt_Eagle 可以研究一下 Oxford Instruments,他们是 $COHR 和 $AAOI 在刻蚀/沉积工艺方面的知名供应商。这与 $VECO/Axtron 这类公司所做的略有不同。许多更上游的公司(如 Vital)是私有的,或者我已经提到过了。
英文原文
@Investmnt_Eagle Could look into Oxford instruments, they're a known $COHR and $AAOI supplier for etch/deposition process. Slightly different than what $VECO/Axtron type are doing. Lot of the more upstream names like Vital are private or I've already mentioned.
-
AAOI财报揭示上游设备与材料瓶颈,IQE等上游厂商受益显著。
来自 $AAOI 财报的二阶效应: -> 对 MBE/MOCVD 反应堆的需求极度旺盛。 它们用于制造外延晶圆(epiwafers)。 这为“制造”环节释放了两个信号: - $IQE 坐拥一座绝对的“金矿” - 资本开支周期将惠及类似 $VECO ($18.5亿) 和 Aixtron ($AIXXF $37亿) 这样的设备商。 引用:3-4 家超大规模云服务商,可能是 $GOOGL、$AMZN、$MSFT,对光收发器(transceivers)的需求已触顶。 -> 中际旭创(Innolight)、易普拓(Eptolink)、$COHR 等厂商可能也会经历一轮资本开支周期以满足需求。 这主要利好两家外延晶圆代工代工厂: -> $IQE ($1.75亿) ~100+ 台反应堆 -> 晶品半导体(Landmark) (~$36.8亿) ~27-30 台反应堆。 所以这大概就是为什么 $LITE 大量使用 $IQE,而 Landmark 营收屡创新高。$AAOI 则一直在建设其德州工厂的产能。 总结: 1. Aixtron 和 $VECO 受益于今年类似迷你版 $ASML 的资本开支周期。当 Aixtron 在2月指引中表示预计2026年“整体市场环境持续疲软”时,也许他们在低调隐藏实力(sandbagging)或订单尚未下达。话虽如此,我认为最佳敞口可能并非设备卖家。 2. $IQE 坐拥满足 $LITE 及其他超大规模云服务商对 InP 外延晶圆需求的产能金矿。他们只需成功重组,估值就能更像其台湾同行($36.5亿市值)。 3. $AAOI 的财报显示收发器制造商及其组装所需的组件/材料需求惊人,市场肯定尚未定价。 4. $AXTI 一如既往地成为所有环节的瓶颈,正如我去年所提到的。 名字越响亮(如 $LITE、$COHR),交易可能越拥挤/已充分定价。 但当你向上游追溯时,还有大量未被市场定价的隐藏宝石,它们将从巨大的光子学扩张中受益。 这是投资者的新淘金热,但本质是瓶颈狩猎。
英文原文
Second order effects from $AAOI earnings: -> Extreme demand for MBE/MOCVD reactors. They're needed for epiwafers. This signals two things for "manufacturing": - $IQE is sitting on an absolute gold mine - capex cycle to ( $VECO $1.85B, Aixtron $AIXXF $3.7B) type players. Quote: 3-4 hyperscalers, likely $GOOGL, $AMZN, $MSFT. demand maxed out for transceivers. -> Innolight, Eptolink, $COHR, and others will probably go through a capex cycle to meet demand too. This largely benefits both outsourced foundries for epiwafers: -> $IQE ($175M) ~100+ reactors -> Landmark (~$3.68B) ~27-30 reactors. So that's probably the reason why $LITE has been heavily using $IQE, and Landmark has been hitting record revenues. $AAOI's has been building capacity from their texas fab. TLDR: 1. Aixtron, $VECO benefits from mini $ASML style capex cycle this year. When Aixtron said 2026 "continued softness in overall market environment expected" from feb guidance, maybe they're sandbagging or orders havent come in yet. That being said, best exposure imo is probably not the machine sellers. 2. $IQE is sitting on a gold mine of capacity to meet demand across $LITE and others hyperscalers for inp epiwafers. They just need to restructure succesfully, to be rerated more like their $3.65B friend in Taiwan. 3. $AAOI earnings is just insane demand for both transceiver makers and the components/materials required to assemble them, definitely not priced in yet. 4. $AXTI as always happens to be the bottleneck of everything as I mentioned last year. The more well known the name is ( $LITE, $COHR ) the likely more crowded/priced-in. But when you go more upstream, there's a ton of hidden gems markets haven't priced in yet that benefit from the massive photonics ramp. It's the new gold rush for investors, but bottleneck hunting.
-
机构利用情报优势提前布局地缘政治风险,带动能源及军工股上涨。
你意识到当情报/新闻在过去几周说“战争迫在眉睫”时,机构可能会采取不同的头寸布局吗? 如果连散户都能提前对冲,那么鉴于伊朗周边的军事集结,我确信机构肯定早已预见。你可以从 $CVX 和部分国防承包商(Defense Contractors)的上涨中看到这一点。 他们拥有更多情报,如卫星图像和政府内部的消息。
英文原文
You do realize when intel/news says “War is Imminent”, for the past few weeks… institutions probably position differently? If even retail is are able to hedge ahead of time, then I’m sure institutions definitely saw it coming given military buildup around Iran. And you’re able to see that with $CVX and bit defense contractors all rising. They have more intel like satellite imagery and word of mouth info with the gov.
-
探讨光子学及半导体供应链中除ASML外的其他关键供应商。
谢谢!我知道有些粉丝推荐了 Aixtorn,但还有其他有趣的供应商,比如 Oxford Instruments。我已经对 $IQE 和 epiwafer(外延晶圆)代工晶圆厂做了更多尽职调查(DD),但光子学领域也有像 $ASML 这样的公司。令人惊讶的是 $VECO 也在那里,尽管我之前是因为 HBM(高带宽内存)相关业务而了解他们。
英文原文
Thanks! I know a some followers recommended Aixtorn, but there's other interesting suppliers like Oxford Instruments out there. I already did more DD into $IQE and the epiwafer merchant foundries, but there's the $ASML's of the photonics world too. Surprisingly $VECO was up there too although I knew them for HBM related stuff.
-
光子学产能爬坡影响巨大,市场机会未被充分定价。
@Frenchie_ 市场机会太多没有被定价!而且光子学(Photonics)产能爬坡的影响实在太大了。不想错过。
英文原文
@Frenchie_ There's way too much market opportunity not priced in! And the implications are just way too big for photonics ramp up. Don't want to miss out.
-
AAOI财报预示光子学新范式,资金或轮动至上游设备及材料供应商。
$AAOI 预计到 2027 年光收发器需求激增 10 倍,基于 $55 亿的市场规模 (MC),实现 $43 亿的年经常性收入 (ARR)…… 这规模巨大。 $LITE 和 $COHR 可能已经拥挤,但。 我们可能会看到资金向上游轮动,例如: $ASML 类型的供应商: - Aixtron (ETR: AIXA):在输入型金属有机化学气相沉积 (MOCVD) 领域占据约 75% 份额 - $VECO - MOCVD 和分子束外延 (MBE) 系统,可能是第二选择。 - Oxford Instruments:$AAOI 和 $COHR 的等离子刻蚀和沉积系统供应商 独立外延晶圆代工厂 (光子学领域的代工厂): - $IQE:最大的外包化合物半导体外延晶圆制造商。基本上是光子学世界的 $TSM,但有很多传统业务的拖累。 - LandMark Optoelectronics:最直接、纯粹的光收发器单元增长标的 - IntelliEPI (TPE: 2462):使用 MBE 而非 MOCVD 制造高性能磷化铟 (InP) 外延晶圆 原材料衬底供应商 (基础材料): - $AXTI/ Sumitomo:大家现在都知道了 收发器产能扩张的预测令人震惊,而这仅仅是 2027 年的情况。 到 2028 年可能会呈指数级增长。 $AAOI 创纪录的财报标志着光子学新范式的开始,获得相关敞口是一个很好的主意。
英文原文
$AAOI's 10x projected rev surge in optical transceiver demand by 2027 for $4.3B ARR off a $5.5B MC... Is enormous. $LITE and $COHR might be crowded, but. We'll likely see rotation upstream like: $ASML-type suppliers: - Aixtron (ETR: AIXA): ~75% share for inp MOCVD - $VECO - MOCVD and MBE (Molecular Beam Epitaxy) systems, probably second. - Oxford Instruments: Supplier to $AAOI and $COHR for plasma etch and deposition systems Merchant Epiwafer Fabs (Foundaries for photonics): - $IQE: largest outsourced compound semiconductor epiwafer manufacturer. Basically $TSM of photonics world but lot of legacy drag. - LandMark Optoelectronics: Most direct, pure-play for optical transceiver unit growth - IntelliEPI (TPE: 2462): MBE rather than MOCVD to create high-performance InP epiwafers Raw Substrate Suppliers (base materials): - $AXTI/ Sumitomo : everyone knows by now Transceiver ramp projections is staggering, and this is only 2027. It's likely exponentially increasing into 2028. $AAOI's record-breaking earnings signals the start of a new paradigm for photonics, it's a great idea to get exposure.
-
AI供应链受云厂商资本开支驱动,基本面改善非随机反弹。
@KyleWilson77344 它跟随 $GOOGL 超大规模云服务商的资本支出流向韩国存储、日本瓶颈/材料,以及台湾晶圆代工厂如 $TSM 或 ASIC 设计商如联发科。这主要是运营利润、极端增长以及韩国/日本治理结构的变化,而不仅仅是一次随机的反弹。
英文原文
@KyleWilson77344 Its following $GOOGL hyperscaler capex flows into Korean memory, Japanese bottlenecks/materials, and Taiwan fabs like $TSM or ASIC designers like Mediatek. It’s mainly operating profit, extreme growth, and governance changes in Korea/Japan, not just a random rally.
-
分析IQE潜在收购方受限及保持独立倾向。
如果让我猜测,可能是像 $COHR 和 $LITE 这样的公司表现出极大的兴趣。但 $IQE 可能希望保持独立运营,仅出售其拖累业务,而非整个 InP 外延晶圆(InP epiwafer)产能,这使得事情变得有些棘手。此外,由于 $IQE 是一家政府承包商,潜在收购方范围缩小,可能仅限于美英公司。
英文原文
If I had to guess, probably extreme interest by companies like $COHR and $LITE. But $IQE probably wants to stay standalone and sell off their drag only, rather than their entire inp epiwafer capacity, so this makes things a little more challenging. Also the acquisition pool narrows since $IQE is a government contractor, so likely only US-UK companies are potential acquirers.
-
分享对IQE光子学外延晶圆代工瓶颈的研究思路,提示高风险。
谢谢!刚刚发布了我对 $IQE 的个人思考过程。 无论你是否参与交易,光子学领域的 Epifoundries(外延晶圆代工厂)都是一个非常值得了解的供应链瓶颈环节。 但当然,这是极高风险、二元结果的投资,不建议任何人参与交易。 我只是发布沿途的发现+思考过程,希望其他人觉得有趣。
英文原文
Thanks! Just publishing my personal thought process on $IQE. Epiwafer foundries are a fascinating supply chain chokepoint to learn about in photonics, regardless if you take the trade or not. But of course, extremely high-risk, binary outcome, not recommending anyone to take the trade. I'm just publishing my own findings along the way + thought process and hope others find it interesting.
-
光子学拥挤,转向小众瓶颈环节。
@yianisz 恭喜 $AAOI,去年没几个人提到它。最近光子学领域拥挤度很高,所以我一直在转向市场尚不了解的更小众的光子学瓶颈环节。
英文原文
@yianisz Congrats on $AAOI, not many people were mentioning it last year. There’s been a lot of crowding in photonics recently so I’ve been rotating into more niche photonics bottlenecks that markets don’t really know yet too.
-
定义光子学“四大天王”并指出IQE适合高风险组合,看好板块整体表现。
$AAOI 是另一家在最近两个月股价翻倍的 Photonics(光子学)公司! 感觉从 $AXTI 到 $AAOI 的每一家光子学公司最近都在翻倍或翻三倍? 我将这些光子学敞口标的称为“四大天王”: 1. $LITE(TPU 的高 BOM(物料清单)占比) 2. $AXTI(万物材料) 3. $COHR(供应链的重要组成部分) 4. $AAOI(美国制造) 其他如 $IQE,我最近也稍微加了一些仓位,主要用于风险偏好较高的组合(它是 $LITE 和其他公司的 Epiwafer(外延晶圆)代工厂),但其财务表现并不出色。 话虽如此,很高兴两个月后,这些名字中的许多都开始起飞了。
英文原文
$AAOI is another photonics name that has just doubled in the last two months! Feels like every photonics company from $AXTI and $AAOI has been doubling or tripling recently? I’ve dubbed this the “Elite Four” for photonics exposure: 1. $LITE (High BOM of the TPU) 2. $AXTI (Materials for Eveything) 3. $COHR (huge part of supply chain) 4. $AAOI (made in America) Others like $IQE I’ve personally added recently slightly more for risk-on portfolios (epiwafer foundry for $LITE and others), but don’t have a great financial profile. That being said glad 2 months later, a lot of these names are taking off
-
对内存超级周期预测的怀疑反证了其疯狂程度。
@o15CR 是的,历史上他们确实非常务实。 第一反应是对这些预测持极度怀疑态度,但这恰恰说明了如果预测成真,内存超级周期(memory supercycle)有多么疯狂。
英文原文
@o15CR Yeah historically they were really grounded. First reaction was to be extremely skeptical of these projections, but that just goes to show how insane the memory supercycle is if it turns out true.
-
解析散热瓶颈及$AEHR通过高温压力测试提升良率的逻辑。
其中的细微差别在于有两点。一是散热管理瓶颈(液冷、铜线寿命延长等)。 $AEHR 处于半导体良率的制造环节。利用热量进行压力测试,以发现可能导致早期失效的潜在缺陷单元。 由于芯片在极端高温下运行,导致早期寿命失效的缺陷风险很高。因此,他们需要在极端高温下测试每一颗芯片。
英文原文
So the nuance is that there's two. One is thermal management bottlenecks (liquid cooling, copper life extension, etc). $AEHR sits in the manufacturing side for semi yield. Heat is used to stress test to see potential defective units for early failure. Since chips run at extreme temperatures, a defect causing early life failure is high. So they need to test every chip at extreme heat.
-
对比三星与SK海力士2027年盈利预测及极低估值。
以下是摩根士丹利(MS) / 麦格理(Macquarie) 对 2026-2027 年的预测盈利。三星(Samsung) 2026年:约 1820 亿美元 vs. 约 2108 亿美元;2027年:约 2351 亿美元 vs. 约 3337 亿美元。SK 海力士(SK Hynix) 2026年:约 1329 亿美元 vs. 约 1905 亿美元;2027年:约 1670 亿美元 vs. 约 3128 亿美元。因此,三星 2027 年市盈率(P/E) 约为 3.7 倍,SK 海力士约为 2.2 倍。
英文原文
These are projected earnings for 2026-2027 from MS / Macquarie. Samsung 2026: ~$182.0B USD vs. ~$210.8B USD 2027: ~$235.1B USD vs. ~$333.7B USD SK Hynix: 2026: ~$132.9B USD vs. ~$190.5B USD 2027: ~$167.0B USD vs. ~$312.8B USD So roughly 2027 3.7x P/E for Samsung and 2.2x P/E SK Hynix.
-
英伟达财报虽超预期但股价跌,长期看AI加速,应关注云厂商资本开支。
$NVDA 远超预期,预计营收780亿美元,高于预期的720亿美元。 该股在财报发布后下跌4.82%。 这是一个有争议的观点,但在我看来,英伟达的财报是一个部分滞后指标。 最需要关注的是超大规模云服务商(Hyperscaler)的资本支出(Capex)预测以及 $TSM 的预测。 市场倾向于认为 $SNDK 等存储芯片、来自 $TSM 的半导体、电力/电网等其他板块,将主要依据上述两项指标涨跌,因为它们代表了整个人工智能(AI)行业。特别是随着从 $GOOGL 到 $AMZN 的超大规模云服务商纷纷加速构建其专用集成电路(ASIC)(除英伟达外)。 英伟达大幅超预期仅仅是已实现的确认,表明人工智能交易可能会加速。 短期流动性和做市商(MM)头寸影响周度价格,但长期来看,情况似乎将进一步加速。
英文原文
$NVDA blows away expectations, with $78B in revenue projected vs. $72B. The stock dropped 4.82% on the earnings. Controversial opinion but Nvidia earnings is a partially lagging indicator to me. The #1 thing to look out for is hyperscaler capex projections and $TSM projections. Markets like $SNDK in memory, semis from $TSM, power/grid and others are likely to rise/fall based on those two in specific as they represent the entire AI sector as a whole. Especially as hyperscalers from $GOOGL to $AMZN rush out to build their on ASICs (apart from Nvidia) Nvidia beating by a wide margin is just realized confirmation that the AI trade is likely to accelerate. Short term liquidity and MM positioning affect weekly prices, but long term, things look to accelerate even further.
-
ALMU量子点尚早,建议关注超大规模数据中心供应链瓶颈。
@Racetoamilly 恕我直言,$ALMU 和量子点(Quantum Dot)技术可能还要几年才能落地。也许关注超大规模数据中心(Hyperscaler)供应链中现有的瓶颈环节会更好。
英文原文
@Racetoamilly IMO $ALMU and quantum dot is prob a few years out. Probably better to look at existing bottlenecks used in hyperscaler supply chains
-
AXTI因光电子供应链瓶颈受追捧,需挖掘多跳连接以发现未定价机会。
如果你听了匿名者的话,$AXTI 现在是不是已经‘上天’了? 在 $GOOGL 创下资本支出记录后,市场正疯狂寻求光电子(Photonics)供应链的敞口。 如果你问 AI:“AXT 或 IQE 是否属于 Google TPU 供应链”: - 它会说“不”,因为存在被掩盖的多跳连接。 但如果你映射出从: $AXTI ($21亿) -> $IQE ($1.5亿) -> $LITE -> 超大规模客户 ASICs… 的流向,你可能会发现市场尚未定价的早期机会。
英文原文
$AXTI has now reached the Moon if you listened anon? Markets are scrambling to get exposure to photonics supply chains following $GOOGL record capex numbers. If you ask an AI: “Is AXT or IQE part of Google TPU supply chains”: - It will say no due to obscured multi-hop connections. But if you map the flow from: $AXTI ($2.1B) -> $IQE ($150M) -> $LITE -> Hyperscaler ASICs… You might discover something early that markets haven’t priced in.
-
光子学或成未来两年极端瓶颈,市场正提前抢跑。
@boogienator2988 光子学(Photonics)在未来两年可能会像内存一样成为极端的瓶颈。市场目前只是在抢跑。 https://t.co/IpjQ3bKP3D
英文原文
@boogienator2988 Photonics probably extreme bottleneck next two years like memory. Markets are just frontrunning now. https://t.co/IpjQ3bKP3D
-
分析AXTI双寡头地位及IQE作为外延晶圆龙头的估值潜力。
是的,分析得很到位!$AXTI 基本上在两个不同的瓶颈环节(上游原料加工和磷化铟(InP)衬底)处于双寡头地位。 $IQE 确实非常关键,但大多数人并没有将 $LITE 与其他超大规模云服务商(Hyperscalers)进行映射关联。 它可能是西方最大的外延晶圆(Epiwafer)工厂,其市值($38.5亿)与 IQE($1.5亿)的巨大差异,主要源于资产负债表状况以及纯题材(Pure Play)暴露度的不同。 如果他们能转向纯题材并消除拖累因素,估值重估(Rerate)的空间可能会相当大。
英文原文
Yep, that’s a good analysis! $AXTI is basically a duopoly in two different bottlenecks (upstream feedstock processing and InP substrates). $IQE is really critical but most people don’t really do the mapping to $LITE to other hyperscalers. It’s probably the largest western epiwafer fab, and landmark is $3.85B MC. vs IQE $150m mainly due to balance sheet and pure play exposure. If they just pivot to pure play the cut the drag could be rerated quite a bit
-
对比AI供应链中实际应用的半导体公司与科研项目类公司。
@martijnde_boer $ALMU 和 $POET 感觉更像是科研项目,而 $AXTI、$IQE、$LITE 等都在超大规模数据中心(hyperscaler)供应链中实际使用。 鉴于 $POET 资金充裕且通过 Celestial 拥有后门(backdoors),其机会更大。
英文原文
@martijnde_boer $ALMU and $POET feel more like science projects while $AXTI, $IQE, $LITE and others are all actively used in hyperscaler supply chains. $POET has a better chance given they’re well funded and have backdoors via Celestial.
-
探讨超大规模云服务商自研ASIC芯片的BOM及供应链细节。
@SharestepAI 超大规模云服务商(谷歌、亚马逊、微软、Meta)自研定制芯片(ASIC)的物料清单(BOM,即组件、供应商、成本细分)。
英文原文
@SharestepAI bill of materials (meaning components, suppliers, cost breakdown) of the custom chips (ASICs) a hyperscaler (google, amazon, microsoft, meta) makes.
-
AI供应链客户归属难确认,仅能通过财报营收增长侧面印证。
@Crawfordbro11 无法100%确定其准确性。未来这很可能仅体现为 $AXTI 或 $IQE 财报中营收的增长,而不会有像 Google 或 Microsoft 这样的公司明确声明“我们使用 x 供应商”。
英文原文
@Crawfordbro11 No way to 100% know if it's correct or not. It would probably just show up as a increase in revenue in the earnings for $AXTI or $IQE in the future without a company like Google or Microsoft saying "we use x supplier" https://t.co/HCsGhaPbqJ
-
通过多跳关系映射,挖掘光子学上游关键供应商AXTI和IQE的投资逻辑与风险。
-> $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 和超大规模客户 ASIC 的物料清单(BOM)极其机密。 但即使没人会说 $AXTI 或 $IQE 供应给某家超大规模客户,你也能通过关系映射大致推测出来。 你会惊讶于有多少多跳关系被忽视了。 (引用内容):我是第一个讨论 $AXTI 与光子学物料清单/供应链关系的人: $IQE 也很有趣,作为少数西方供应商之一。 基本上,如果你看 $GOOGL TPU/超大规模客户 ASIC 的光子学流向,大概如下(很可能,但未披露): 光收发器(最高 BOM): Lumentum/Cloud Light: ~ 关键 / $AXTI-> $AXTI/住友/JX -> $IQE (外延晶圆) -> $LITE / Cloud Light -> $FN (代工) -> $GOOGL TPU Merchant 光学供应链: ~ 关键 / $AXTI -> $AXTI / 住友 / JX -> $LITE / $AVGO / $COHR (EML) + $MRVL / $MTSI / Semtech -> 中际旭创/新易盛 -> $GOOGL 所以如果你想找光子学 BOM 更上游的“登月计划”型/涨价股:$AXTI 和 $IQE 是必经之路。 $AXTI 此前基本面很差,但最近的 Northland 融资轮确立了其涨势。 $IQE 目前基本面很差(净债务 2350 万英镑),但可能是供应链中最关键的部分。如果他们能出售台湾业务,仅凭其 inp 业务,股价大幅上涨也不足为奇。 有 1800 万英镑可转换票据(几乎可以忽略不计),然后是新发 1.2 亿至 1.54 亿股(~12% 至 15%),相对于当前规模也几乎可以忽略不计。 另一方面,$LITE 和中际旭创可能更成熟。 TLDR: $IQE -> 似乎对西方供应链至关重要,市值 1.3 亿英镑。净债务,如果出售台湾业务 -> 强烈重估,或者他们可能只是稀释你。 但如果台湾业务未能出售,预计会被稀释到像 Wolfspeed 那样。所以风险巨大,务必自行研究风险。 但 $AXTI 和 $IQE 对我个人来说很有趣(我持有 $IQE)。
英文原文
-> $LITE is a known $GOOGL supplier. -> $IQE is a known $LITE supplier. -> $AXTI is a known $IQE supplier. Google TPU + Hyperscaler ASIC BOM are extremely confidential. But you can kinda guess from relationship mapping even though nobody will say $AXTI or $IQE supplies to X hyperscaler? You'd be surprised how multi-hop relationships are missed.
-
梳理AXTI-IQE-LITE供应链,指出市场或遗漏光子学标的。
@Huyster11 我在 $AXTI 上赚了几百%,所以我过得很好! 只是想聊聊 $IQE,它是 $LITE 的知名供应商。而 $AXTI 是 $IQE 的知名供应商。 大家都在买入与谷歌资本支出挂钩的光子学交易,但感觉市场可能漏掉了一两个名字。
英文原文
@Huyster11 I’ve made a few hundred % on $AXTI so I’m doing fine! Just wanted to talk about $IQE which is a known $LITE supplier. And $AXTI is a known $IQE supplier. Everyone’s buying up the photonic trade tied with Google capex but feels like markets might have missed one or two names
-
解析光子学上游供应链,看好AXTI与IQE但提示IQE稀释风险。
我是最早将 $AXTI 与光子学(BOM)/供应链联系起来的人: $IQE 也很有趣,作为少数西方供应商之一。 基本上,如果你看 $GOOGL TPU/超大规模ASIC的光子学流程,大概是这样(很可能,但未披露): 光模块(最高BOM): Lumentum/Cloud Light: ~ Vital / $AXTI-> $AXTI/住友/JX -> $IQE (外延晶圆) -> $LITE / Cloud Light -> $FN (代工制造) -> $GOOGL TPU Merchant optical 供应链: ~ Vital / $AXTI -> $AXTI / 住友 / JX -> $LITE / $AVGO / $COHR (EML) + $MRVL / $MTSI / Semtech -> 中际旭创/新易盛 -> $GOOGL 所以如果你想找光子学BOM/涨价的“登月型”股票,更上游的光子学BOM:$AXTI, $IQE 是你的方向。 $AXTI 之前基本面很差,但最近的Northland融资轮确立了其涨势。 $IQE 现在基本面很差(净债务2350万英镑)但可能是供应链中最关键的部分。如果他们能出售台湾业务,我不惊讶它会仅因inp业务就大幅上涨。 有1800万英镑可转换票据(这几乎没什么),然后是1.2亿到1.54亿新股(~12%到15%),相对于当前规模也几乎没什么。 另一方面,$LITE 和中际旭创可能更成熟。 TLDR: $IQE -> 似乎对西方供应链至关重要,市值1.3亿英镑。净债务,如果出售台湾业务 -> 强烈重估,或者他们可能只是稀释你。 但如果台湾业务未能出售,预计会被稀释到像Wolfspeed那样。所以巨大、巨大的风险,请自行研究风险。 但 $AXTI 和 $IQE 对我个人来说很有趣(我持有 $IQE)。
英文原文
Was the first to talk about $AXTI in relation to photonics BOM/supply chains: $IQE is very interesting too as one of the only Western suppliers. Basically if you look at photonics flow on $GOOGL TPU/hyperscaler ASICs kinda looks like this (very likely, but undisclosed): Optical Transceivers (highest BOM): Lumentum/Cloud Light: ~ Vital / $AXTI-> $AXTI/Sumitomo/JX -> $IQE (Epi-Wafers) -> $LITE / Cloud Light -> $FN (Contract Manufacturing) -> $GOOGL TPU Merhcant optical supply chain: ~ Vital / $AXTI -> $AXTI / Sumitomo / JX -> → $LITE / $AVGO / $COHR (EML) + $MRVL / $MTSI / Semtech -> Innolight/Eoptolink -> $GOOGL So if you want moonshot-type photonics BOM / price-hikes stocks deeper upstream in the photonics BOM: $AXTI, $IQE and your way to go. $AXTI had terrible fundamentals before but the recent Northland fundraising round cemented its run. $IQE has terrible fundamentals now (Net debt £23.5 million) but is probably one of the most critical parts of the supply chain. If they manage to sell their Taiwan operations, wouldn't be surprised if it went up quite a bit just from their inp business. There's £18m convertible notes (which is basically nothing), then there's 120 to 154m new shares (~12% to 15%), which is also kinda nothing relative to current size. On the other hand, others $LITE and Innolight are probably more established. TLDR: $IQE -> seems critical to Western supply chains, $130MC. Net debt, if they sell Taiwan business -> strong re-rating or they might just dilute you anyway. But if the Taiwan business fails to be sold, probably expect to be diluted to oblivion like Wolfspeed. So huge, huge, risk ad do you own research into risks. But $AXTI and $IQE might are personally interesting to me (I do own $IQE).
-
澄清供应链映射简化,补充COHR等潜在供应商及谷歌多源采购。
是的,只是想补充一个免责声明:这是对可能供应链的极度简化映射。 $COHR、Eoptolink 或 Fabrinet 很可能也是。在输入端方面,JX/Sumitomo/AXTI 也是。如果你再往上游追溯一级,Vital 也许也是。 $GOOGL 肯定使用许多来源。
英文原文
Yes, just wanted to add a disclaimer: this is an extremely oversimplified mapping of the likely supply chain. $COHR, Eoptolink, or Fabrinet are likely too. Alongside JX/Sumitomo/AXTI on the inp side. Maybe Vital if you go one level higher upstream. $GOOGL definitely uses many sources.
-
对比摩根士丹利与麦格理对三星及SK海力士营收预测,前者更保守。
摩根士丹利(Morgan Stanley)对SK海力士(SK Hynix)和三星(Samsung)的预测看起来更接地气一些。 摩根士丹利 vs. 麦格理(Macquarie)关于三星的预测: 2026年:约1820亿美元 vs. 约2108亿美元 2027年:约2351亿美元 vs. 约3337亿美元 SK海力士: 约1329亿美元 vs. 约1905亿美元 约1670亿美元 vs. 约3128亿美元(哇) 麦格理在各方面都显著更乐观,他们对SK海力士2027财年的预测简直完全是另一个世界了,哈哈。
英文原文
Morgan Stanley's projections here seem a bit more grounded on Sk Hynix/Samsung. MS vs. Macquarie on Samsung 2026: ~$182.0B USD vs. ~$210.8B USD 2027: ~$235.1B USD vs. ~$333.7B USD SK Hynix: ~$132.9B USD vs. ~$190.5B USD ~$167.0B USD vs. ~$312.8B USD (wow) Macquarie is significantly more bullish across the board and their projections for SK Hynix in FY2027 are in a completely different world lol.
-
看好光子学、内存、先进封装及电网板块,因超大规模云服务商投资集中于此。
我对光子学(Photonics)、内存(Memory)、先进封装(Advanced Packaging)以及美国的电力/电网(Sector)板块个人非常看好! 超大规模云服务商(Hyperscalers)日益增长的投资全部流向这些领域。毕竟资金最终总要流向某处,因此我认为上述提到的股票群体将受益最大(成为首选标的)。
英文原文
フォトニクス、メモリ、先端パッケージング、それに米国の電力/グリッドセクター、個人的にめちゃくちゃ期待してます! ハイパースケーラーの増大する投資は全部そこに向かってますからね。資金は結局どこかに流れるわけじゃないですか?だから、上に挙がってる銘柄群が最大の恩恵を受ける(大本命になる)と思います
-
简街资本因错判贝莱德韩国指数期权波动率而面临亏损风险。
@MansaTrader 简街资本(Jane Street)很可能是贝莱德(BlackRock)韩国指数基金的对手方。他们在2028年期权链上错误定价了隐含波动率(Implied Volatility)。仅波动率风险敞口(Vega)的扩张就足以在一周内使期权价值翻倍,而隐含波动率的快速上升通常会导致做市商(Market Makers)亏损。
英文原文
@MansaTrader Jane Street is the likely counterparty of the Blackrock’s South Korea Index. They mispriced IV on the 2028 option chain. Vega expansion alone would have doubled option values in a week and rapid IV increases typically cause a loss to MMs.
-
指出电力与半导体是主要瓶颈,并赞赏对方对术语的严谨态度。
@darydayani 电力、半导体(Semiconductors)以及其他瓶颈*。 但我很欣赏这种语义上的严谨。
英文原文
@darydayani Power, Semis, and other bottlenecks*. But appreciate the semantics
-
SK海力士三星创新高,AI算力产业链迎世代财富机遇。
满眼皆是绿色(上涨)。SK海力士、三星(以及大概率 $EWY)今日均创历史新高。人工智能正在为算力(电力、瓶颈环节和半导体)的未来拥有者带来世代财富。
英文原文
It’s just an endless sea of green. SK Hynix, Samsung (and likely $EWY) are all time highs today. AI is bringing generational wealth to the future owners of compute (Power, Bottlenecks, and Semis). https://t.co/igFzY9whEB
-
类比Toto和Yamamura,分析内存与光子学供应链瓶颈。
@hanbonjovi 在内存领域,确实存在与 Toto(马桶制造商)的相似之处。但在光子学领域,则是 Yamamura(瓶子制造商)。https://t.co/mG5BReQw7
英文原文
@hanbonjovi There’s definitely parallels to Toto the toilet maker for memory. But Yamamura, the bottlemaker for photonics. https://t.co/mGg5BReQw7
-
光子学2028年放量,存储需求结构性强劲且具定价权。
关于光子学(Photonics)的好问题,你可以看到2028年起会有大规模扩张。现在是2026年,市场正在提前定价预期($LITE的利润目前不算高,但随着Google TPU等项目及其他应用的规模扩大,预计未来几年利润将激增)。 对于存储(Memory),你看到它现在正体现在资产负债表上,净利润同比增长约300%。预计这种势头将持续到2028年。 在我看来,存储需求是结构性的,他们将在2028年晚些时候之前保持很强的定价权。
英文原文
Good question for photonics you can see major scale up 2028 onward. It’s 2026 now and things are just getting priced in anticipation ( $LITE profits aren’t exactly high, but they’re expected to balloon in a few years as Google TPU programs and others scale up). For memory you see it hitting the balance sheet now as net income increases like 300% Y/Y. This is expected to last well into 2028. Memory demand is structural and they’ll maintain a lot of pricing power into late 2028 though imo.
-
分析AI供应链中关键海外垄断厂商及美国本土瓶颈环节的投资逻辑。
哈哈 https://t.co/8i18XOIKK7 我重点关注的海外股票摘要: 1. 日本帝人博谢(Nitto Boseki) - T-玻璃(T-Glass) 2. 台湾欣兴电子(Unimicron) - 基板/玻璃基板(Substrates/Glass Core) 3. 三星/SK海力士 - 高带宽内存(HBM) 4. 日本铠侠(Kioxia) - 与 $SNDK 合作的 NAND 闪存 5. 台湾南亚科技(Nanya Technologies) - 内存 我通常更偏好做多美国供应链,如 $INTC (晶圆代工)、$MU (内存) 或 $LPTH (锗),因为它们具备“美国制造”的瓶颈优势及供应链地位。但这些海外公司要么对整个人工智能(AI)基础设施建设至关重要,要么能捕获大量溢出需求。 如果你想知道如何投资海外市场? - $IBKR 像帝人博谢这样的公司有美国场外交易(OTC)等价物,如 $NBCLF。如果你使用 Robinhood,$EWY / $FLKR 可以提供对韩国大型内存公司的敞口。 像帝人博谢这样的公司接近垄断,因此这是我特意去投资的极少数海外公司。
英文原文
lol https://t.co/8i18XOIKK7
-
建议AI实验室实施KYC验证以遏制模型蒸馏风险。
今天,Anthropic 点名批评了 DeepSeek、Moonshot 和 Minimax 的蒸馏行为。 OpenAI 也面临中国实验室同样的问题。 没人做的最愚蠢的修复方案: KYC 验证。 从 $HOOD 到 $IBKR 的每家金融机构都这样做以追踪资金流向(例如 Persona 自拍/ID)。 同样的规则应适用于 token 流。 美国前沿实验室应尽快对其最新模型实施此措施。 我确信他们关心无摩擦增长,但这关乎国家安全风险,且个人/企业只需经历一次此流程。 这是最容易在一天内完成的修复方案。 如果不这样做,中国实验室将继续找到绕过方法。
英文原文
Today, Anthropic called out DeepSeek, Moonshot, and Minimax for distillation. OpenAI has also faced the same issues with Chinese labs. The most braindead fix nobody is doing: KYC verification. Every financial institution from $HOOD to $IBKR does this to track flow of funds (eg. Persona selfie / ID) Same should apply to token flows. American Frontier Labs should start doing this soon for their latest models. I'm sure they care about frictionless growth, but this is national security at risk and people/enterprises only need to go through this process once. This is by far the easiest fix that can be done in a day. Chinese labs will continue to find bypasses if this is not done.
-
市场资金从虚高标的轮动至超大规模云厂商资本支出相关的电力与半导体板块。
@rioferdy838 如果市场将许多估值虚高的标的轮动到超大规模云服务商资本支出(Capex)交易中的电力股和半导体股,这并不令人惊讶。
英文原文
@rioferdy838 Not surprising if market rotated a lot of the inflated names into the hyperscaler capex trade in power names and semis
-
解析$EWY波动率错配引发的Vega扩张现象。
这是波动率错配引发的 Vega 扩张(Vega expansion)的实际表现。 $EWY 的隐含波动率(Implied Volatility, IV)指数下降。 整个长期期权链(Long term option chain)上涨。 https://t.co/EEuYTCyxYH
英文原文
This is Vega expansion in action from volatility mispricing. $EWY SK index down. Entire long term option chain up. https://t.co/EEuYTCyxYH
-
波动率交易因间接结构导致三星/SK海力士敞口远超50%上限。
虽然存在相关敞口,但波动率交易策略追求的是对三星/SK海力士的最大化敞口。基本上我发现它通过间接的通道结构间接地大幅超过了上限。人们常犯的错误是认为上限为50%(波动率降低),但像SK Square这样的公司基本上就是SK海力士的控股公司。三星电子也通过其他三星公司几乎被多次计入。随着市值上升,权重越来越向它们倾斜,而不是对两者都限制在50%。
英文原文
There’s exposure to those but the volatility trade is wanting max exposure to Samsung / SK Hynix. Basically I found it inadvertently largely exceeds the cap through indirect pass-through structures. And people make the mistake thinking it caps out at 50% (volatility gets lower) but companies like SK Square are basically a SK Hynix holding company. And Samsung Electronics is practically listed multiple times through other Samsung companies. As the marketcaps go up, it increasingly becomes weighted toward them, not capped at 50% for them both.