觅信聚合

我的关注

2026-09-11 07:03

原文：
Yep, I'm personally high conviction on the AI trade (especially photonics/memory) despite macro.

- $NVDA gave forecasts for $1.3T hyperscaler capex ($100B+ above BofA analyst expectations)

- $SPCX Wells Fargo forecasts are ~$263B for 2027.

- $127B spend in 2027 for Anthropic/OpenAI model training + inference from internal investor documents (WSJ).

- Even the Pentagon (US gov) is looking to fund the AI buildout, with $5B potentially into Fluidstack.

What sectors do I think capex flows hit the hardest?

Photonics + Memory.

Trendforce for example est. DRAM and NAND to be 68% of hyperscaler capex in 2027.

So when all that memory procurement spend hits the balance sheets of $MU, $SNDK, Samsung, $SKHY with 70-80%+ margins.

I don't think these companies care about short term macro scares.

And when your leading memory companies are all allocating capacity for DDR5/HBM or high end memory, with little visibility on legacy memory supply coming online.

What does that say about the durability of ESMT/Etron/Winbond and others?

And when your optical content/GPU goes up for rubin ultra + $META / $GOOGL TPU v9, etc. Alongside TAM eg. $131.4 billion (+81% revision) for 2027 optical module forecasts.

Who benefits from supplying all that lasers during a shortage other than $LITE / $SIVE / $AAOI / $COHR?

I think the overwhelming fundamentals + growth of some of the AI sector companies will outperform macro.

翻译：
是的，尽管宏观环境，我个人对AI交易（尤其是光子学/内存）高度确信。

- $NVDA 预测超大规模算力公司资本支出为 1.3 万亿美元（比 BofA 分析师预期高出 1000 亿美元以上）

- $SPCX Wells Fargo 预测 2027 年约为 2630 亿美元。

- 内部投资者文件（华尔街日报）显示，2027 年 Anthropic/OpenAI 的模型训练和推理支出为 1270 亿美元。

- 即使是美国国防部也在考虑为 AI 建设提供资金，可能向 Fluidstack 投入 50 亿美元。

我认为资本支出最集中在哪些领域？

光子学 + 内存。

例如，Trendforce 估计 2027 年 DRAM 和 NAND 将占超大规模算力公司资本支出的 68%。

因此，当所有这些内存采购支出计入 $MU、$SNDK、Samsung、$SKHY 的资产负

2026-09-11 06:03

原文：
It's a macro market right now https://t.co/12ySQqZWrO

翻译：
现在是宏观市场 https://t.co/12ySQqZWrO

https://video.twimg.com/amplify_video/2098290933058211840/vid/avc1/480x262/qp7KORc6ycggM2Y1.mp4?tag=29
Serenity (@aleabitoreddit) on X
It's a macro market right now

2026-09-10 12:49

原文：
共和党が議会で多数派を維持できたら、トランプが成人1人につき5,000ドルを配ると約束している理由が、正直よく分からない。（総額約1.35兆ドル）

それより、アメリカ中のトイレを、日本にあるような便座が温かくなるTOTO（5332.T）のトイレにアップグレードしたほうが、支持率はもっと上がるんじゃないかと思う。

計算してみた：

3億3,500万台のトイレをTOTO UltraMax II + S7Aに交換すると、約9,900億ドル。
設置費用まで含めると、総額は約1.3兆ドル。

僕の案のほうがかなり安い。

翻译：
如果共和党在国会保持多数席位，我真的不明白特朗普承诺每位成年人发放5,000美元的原因。（总额约1.35万亿美元）

相比之下，我觉得把美国各地的厕所升级为日本那种座圈加热的TOTO（5332.T）厕所，支持率会更高。

算了一下：

把3.35亿台厕所换成TOTO UltraMax II + S7A，约9,900亿美元。
加上安装费用，总额约1.3万亿美元。

我的方案要便宜得多。

2026-09-10 06:38

原文：
Wow, HBM reportedly drove up AI accelerator/card prices by 20-50% in China.

So $SKHY, Samsung, $MU, should be happy to hear this:

- Huawei has raised indicated pricing for the Ascend 950DT 20–50% above quotes from just 2 months ago

- Cambricon's upcoming 690 is reportedly +20–30%

Reuters stated that HBM acquired through grey market channels in China can cost several times regular pricing.

As someone who is long memory, I'm happy about memory causing this much inflation.

翻译：
哇，HBM据称在中国把 AI 加速器/卡的价格推高了 20%~50%。

所以 $SKHY、Samsung、$MU 应该会为此感到高兴：

- 华为将 Ascend 950DT 的指示价格提高了 20%~50%，比仅两个月前的报价高出这么多
- 据称，Cambricon 即将推出的 690 价格上涨了 20%~30%

路透社称，通过中国灰色渠道获取的 HBM 价格可能是正常价的数倍。

作为一个长期持有内存的人，我对内存导致如此高的通胀感到满意。

2026-09-10 06:28

原文：
$NVDA could cause another bottleneck like EML/CW but this time with high end probe card / test socket capacity.

- US related beneficiaries might be $FORM and $COHU from my own read through. (don't own any)

- WinWay, MPI and CHPT where the three cited by the article for probe card and socket markets.

The article also described a capacity spillover into:

- MSScorps (* hello) / MAtek / iST, which were cited in the source.

Sort of reminds me of what they did with EML/CW recently? The news just came out today so sometimes markets needs time to digest implications.

Eg. Winway said sock lead times went from 8w to 13-14w.

So it's worth looking into.

翻译：
$NVDA 可能会像 EML/CW 那样再次出现瓶颈，但这次是高端探针卡/测试插座产能。

- 我个人阅读后认为美国相关受益者可能是 $FORM 和 $COHU。（我并未持有）

- 文章中提到的探针卡和插座市场的三家公司是 WinWay、MPI 和 CHPT。

文章还描述了产能溢出至：

- MSScorps（* hello）/ MAtek / iST，这些在来源中被提及。

有点让我想起他们最近对 EML/CW 的操作？该消息今天刚发布，市场有时需要时间消化其影响。

例如，WinWay 表示插座交付周期从 8 周延长至 13-14 周。

因此值得关注。

2026-09-09 22:44

原文：
Anthropic's Economic Scenario publication was a pretty interesting read:

1. Anthropic's "extreme" scenario has GDP growth reaching 15.4% in 2030

So around 7.3x current rates (~2.1% Y/Y), which is consistent with what Elon Musk is saying about outgrowing the national debt with AI.

2. Almost all of the divergence happens after 2027.

Their model shows relatively little separation in 2026-2027, then the curves bend sharply into 2028-2030.

So my read is that it points to 2026-2027 as buildout years and 2028-2030 as economic acceleration.

Kinda gets mapped as well when you look at $AMZN / $META / $GOOGL capex, and short term FCF headwinds (but massive projected profitability post 2028)

3. In an extreme scenario, unemployment of knowledge workers goes to 17.9%. With 11.9% economy-wide unemployment

"At the individual level, it means coders and call service center agents may have to switch to jobs like electrician and nurse, which are less exposed to AI."

GG computer science majors.

4. Non-cognitive wages: +33.6% vs. baseline in an extreme scenario.

So your electricians/construction-type labor becomes extremely valuable? (but this isn't factoring in humanoids yet...)

Whereas -11.5% for knowledge workers in 2030 vs. baseline.

5. Aggregate capital income is ~81% higher than it would have been without AI

Productivity explodes -> aggregate labor income barely changes -> essentially all of the incremental GDP accrues to capital.

So that makes ownership of capital rather than labor... so equities like compute, power, semis, networking, robotics, etc. become pretty important...

TLDR:

- Economic growth go brr
- Jobs that require thinking = unemployment + lower wages
- Physical labor = more important, like electricians.
- AI equities go brrr.

翻译：
Anthropic 的《经济情景》报告相当有趣：

1. Anthropic 的“极端”情景预测 2030 年 GDP 增长率达到 15.4%

约为当前水平的 7.3 倍（约 2.1% 同比），这与 Elon

2026-09-09 22:44

Musk 所说的 AI 将使经济增长超越国债的观点相吻合。

2. 几乎所有分歧都出现在 2027 年之后。

模型显示 2026‑2027 年的差距相对较小，随后曲线在 2028‑2030 年急剧分离。

因此我的解读是：2026‑2027 年是建设期，2028‑2030 年是经济加速期。

这与 $AMZN / $META / $GOOGL 的资本支出以及短期自由现金流压力（但 2028 年后预计利润大幅提升）相呼应。

3. 在极端情景下，知识工作者的失业率升至 17.9%，全经济失业率为 11.9%

“从个人层面来看，程序员和呼叫中心客服可能需要转行做电工、护士等不太受 AI 影响的工作。”

GG 计算机科学专业的同学们。

4. 非认知类工资：在极端情景下比基线高出 33.6%

于是电工/建筑类劳动力变得极其有价值？（但尚未计入类人机器人）

而知识工作者在 2030 年相较基线下降 11.5%。

5. 总资本收入比没有 AI 时高出约 81%

生产率爆炸式增长 → 总劳动收入几乎不变 → 几乎所有新增 GDP 都归资本所有。

这意味着拥有资本而非劳动更为重要……因此计算、能源、半导体、网络、机器人等相关股票变得相当关键。

TLDR：

- 经济增长飞速
- 需要思考的工作面临失业和工资下降
- 体力劳动更受重视，如电工
- AI 相关股票大幅上涨

2026-09-09 17:48

原文：
If you hid the ticker, you would think this was an AI bottleneck name. https://t.co/6rR1BWYRb9

翻译：
如果你隐藏了 ticker，你会觉得这像是一个 AI 瓶颈的名字。 https://t.co/6rR1BWYRb9

2026-09-09 13:00

原文：
I’m not surprised $LAPTOP is down -96.16% after launch.

I frown on these types of events that are not positive sum for retail. https://t.co/YbrzeOQpuT

翻译：
我并不惊讶 $LAPTOP 在上市后跌了 -96.16%。
我对这类对散户没有正和效应的事件感到不满。 https://t.co/YbrzeOQpuT

2026-09-09 12:16

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...

翻译：
Citibank TMT：“激光和光纤正被定位为‘下一个HBM’。”

自2025年以来，我就一直在讨论 $LITE / $AAOI 等激光，以及向连续波（CW）架构的转变，早于当前行业共识。

我完全赞同 Lightmatter/Citi 的观点。

仅供参考：

激光方面：$LITE / $SIVE / $AAOI / $COHR / $MTSI / Furukawa / Sumitomo 是较为知名的公司。

光纤方面：$GLW / Furukawa / Fujikura / Sumitomo 更为人熟知（显然在供应链中还有更多），*披露，个人对该主题的敞口。

补充说明：“供应紧张将为这些资产提供稀缺溢价”。

我们已经看到激光价格上涨，例如今天在深圳 CIOE 2026 上对 $SIVE 的渠道调查以及 $LITE 的业绩评论。

因此我认为2027–2028年的光子超级周期将非常类似于2025–2026年的存储周期。

尤其是根据高盛修订后的估计，光学内容/GPU-ASIC 将显著增长……

2026-09-09 07:31

原文：
Demand for Astra is unprecedented to the point OpenAI might pause new Pro subscriptions.

If you want a translation:

AI buildout + compute go brrr.

翻译：
对 Astra 的需求空前，以至于 OpenAI 可能会暂停新的 Pro 订阅。

如果你想要翻译：

AI 建设 + 计算算力嗖嗖

2026-09-09 05:53

原文：
Special thank you to awodias for the crowdsourced channel check all the way from CIOE 2026 Shenzhen!

- Confirmed that $SIVE is actively engaged with Chinese pluggable makers

- $SIVE has raised lasers prices given tight supply

- 70mW lasers seem the most tight given association with 800g pluggables

It’s pretty cool that our little community reaches all the way around the world.

(Also thank you if other people did channel checks and I missed it, this was the first one I saw)

翻译：
特别感谢 awodias 在 CIOE 2026 深圳进行的众包渠道检查！

- 确认 $SIVE 正在积极与中国可插拔厂商合作
- 由于供应紧张，$SIVE 已上调激光器价格
- 70mW 激光器因与 800g 可插拔产品关联而最为紧缺

我们的社区能够遍及全球，真的很酷。

（如果还有其他人做了渠道检查但我漏掉了，也谢谢你们，这是我看到的第一个）

2026-09-08 18:58

原文：
How https://t.co/pkCKZ7zIrK

翻译：
如何 https://t.co/pkCKZ7zIrK

2026-09-08 16:46

原文：
Goldman Sachs in their recent note upgraded their global optical module market forecast to:

- $67.7 billion (+33% revision) for 2026
- $131.4 billion (+81% revision) for 2027
- $148.5 billion (+115% revision) for 2028

For $NVDA Rubin Ultra, GS increased its modeled optical modules per GPU:

1.6T: 0 -> 2
3.2T: 3 -> 5

GS also raised their 1.6T and above forecast for 2027 by 61%... so that's also very material, with 80% silicon photonics penetration rate estimated.

So more optical content per chip = a lot more CW lasers, more SOI substrates, happy optical companies.

翻译：
高盛在最新报告中上调了全球光模块市场预测至：

- 2026 年 $67.7 billion（+33% 修正）
- 2027 年 $131.4 billion（+81% 修正）
- 2028 年 $148.5 billion（+115% 修正）

针对 $NVDA Rubin Ultra，GS 将每颗 GPU 的光模块模型从：

1.6T: 0 -> 2
3.2T: 3 -> 5

GS 还将 2027 年 1.6T 及以上的预测上调了 61%……这也非常重要，预计硅光子渗透率为 80%。

因此，每颗芯片的光学内容增加 = 更多 CW 激光器、更多 SOI 衬底，光学公司将受益。

2026-09-08 16:22

原文：
Think optical rally today from $SIVE to $AAOI was kinda multifaceted.

Goldman Sachs upgraded their optical forecast:

$67.7 billion (+33% revision) -> $131.4 billion (+81% revision) -> $148.5 billion (+115% revision) from 2026 -> 2028.

- According to the report, the 1.6T silicon optical module uses four 70mW CW Lasers, with a total value of $15-20 ($3.75–5/laser) EML uses 8 200G EML, which is about $160.
- Reported 80% SiPH penetration (good for $SOI) and four CW lasers per 1.6T module
- Rubin Ultra moves from three 3.2T modules/GPU to five, plus two 1.6T modules (more optical content per chip)
- $GOOGL TPU v9 adds another substantial source of demand.

Basically: Much higher TAM revision, more optical content per chip (really really good for laser companies)

$QCOM deal with $AMZN adds another layer for 1.6T related demand, which was explicitly included. Qualcomm had some optical acquisitions recently so that's probably related.

Same with $GLW for fiber (but different parts of the supply chain)

翻译：
我认为今天从 $SIVE 到 $AAOI 的光学行情有点多面性。

高盛上调了他们的光学预测：

- 根据报告，1.6T 硅光模块使用四个 70mW 连续波激光器，总价值 $15-20（每个激光器 $3.75–5），EML 使用 8 个 200G EML，约 $160。
- 报告称 SiPH 渗透率为 80%（对 $SOI 有利），每个 1.6T 模块配备四个连续波激光器。
- Rubin Ultra 从每 GPU 三个 3.2T 模块提升到五个，并增加两个 1.6T 模块（每颗芯片的光学内容更多）。
- $GOOGL TPU v9 带来另一大需求来源。

基本上：TAM 大幅上调，芯片光学内容增加（对激光公司非常非常有利）。

$QCOM 与 $AMZN 的合作为 1.6T 相关需求再添一层，这一点已明确计入。Qualcomm 最近进行了一些光学收购，可能与此有关。

对 $GLW 的光纤业务也是如此（但属于供应链的不同环节）。

2026-09-08 15:38

原文：
Okay I know this is US time and this is a Taiwan stock...

But I really am starting to think $ESMT (3006) can cook like $SNDK. And I’m excited.

So hopefully in a few months, people can something say similar...

FUN FACT CHAT:

Did you know.. 1 old DDR2 memory chip is.. ~$0.96? (from LCSC data)

Obviously different per spec since some can be $2.5...

But if you 3x ASP Hike this.

That's a whole... ~$2 added to final BOM cost.

So about the same cost as a $COST hot dog.

It goes into things like a D-Link security camera, maybe like $33 final cost...

and adds like $2. (which people won't mind)

But... 3X ASP hike to ESMT? Is very, very material.

That's just for one product category, and there's many others from SLC NAND (eg. +120–170% H2) to NOR Flash products for H2 ASP hike projections.

And my favorite thing is, they've already shown they're willing to partake in the price hike games (seems like there's a long way to go)…

Currently they make ~$111M net income a month from July earnings ($2.8B MC), which annualized is like ~2.1X P/E. *I have positions.

So... I think $ESMT has room to cook like masterchef Sandisk.

TLDR:
- Repeated rounds of price hiking = seems absurd for the company's net income... but has potential due to starting low component cost.

- Reminds me of early $SNDK but for other segments (legacy/niche memory).

翻译：
好，我知道现在是美国时间，这只是一只台湾股票…

但我真的开始觉得 $ESMT (3006) 可能会像 $SNDK 那样大幅涨价，我很期待。

所以希望几个月后，大家能说出类似的话…

趣味聊天：

你知道吗…一颗旧的 DDR2 内存芯片大约是 ~0.96 美元？（来自 LCSC 数据）

不同规格当然价格不同，有的甚至要 $2.5…

但如果把它的 ASP 提高 3 倍。

这就相当于在最终 BOM 成本上多了约 $2。

大致相当于一根 $COST 的热狗的成本。

它会被用于比如 D-Link 的安防摄像头，最终成本可能约 $33…

再加上约 $2（用户不会介意）。

但是…对 $ESMT 来说，ASP 提高 3 倍？影响非常大。

这仅是一个产品类别，其他还有 SLC NAND（例如 H2 期间提升 120–170%）到 NOR Flash 产品的

2026-09-08 14:31

原文：
Wait… this was real? I thought this was The Onion type meme.

Honestly would be more bullish on $META if they admitted they trained on thousands of adult films… instead of claiming “personal use”.

What could Zuckerberg be cooking? https://t.co/JXtt3P6BbS

翻译：
等一下…这是真的吗？我还以为这是《洋葱新闻》式的梗。

说实话，如果他们承认是用成千上万部成人电影进行训练，而不是声称“个人使用”，我会对 $META 更加看好。

扎克伯格到底在酝酿什么？ https://t.co/JXtt3P6BbS

2026-09-08 14:14

原文：
$NBIS is back, and coming in hot with a new partnership with $PLTR.

Palantir also named Nebius its preferred sovereign AI infrastructure partner.

The implications are potentially significant considering Palantir’s existing relations with the US Gov + Enterprises. https://t.co/izCDuR4r6Q

翻译：
$NBIS 回来了，并且势头强劲，携手 $PLTR 开启新合作。

Palantir 还将 Nebius 命名为其首选的主权 AI 基础设施合作伙伴。

鉴于 Palantir 与美国政府及企业的既有关系，这一影响可能相当重大。 https://t.co/izCDuR4r6Q

2026-09-08 13:50

原文：
$AMZN signs a new deal with $QCOM, giving Amazon the right to buy a ~$4B stake of Qualcomm through warrants.

At this rate.. Amazon will be more of a semi ETF than $NVDA?

Given their existing equity/warrants with AlChip, $MRVL, $ALAB, $AAOI, and others.

This of course is linked with up to $60B in milestone revenue from Amazon/Qualcomm custom silicon partnership.

Good for both companies, moreso Qualcomm.

翻译：
$AMZN 与 $QCOM 签署新协议，赋予亚马逊通过认股权证购买约 40 亿美元的高通股份的权利。

按此速度……亚马逊将更像是半 ETF，而不是 $NVDA？

考虑到他们已持有 AlChip、$MRVL、$ALAB、$AAOI 等公司的股权/认股权证。

这自然与亚马逊/高通定制硅合作带来的最高 600 亿美元里程碑收入挂钩。

对两家公司都有利，尤其是高通。

2026-09-08 13:03

原文：
Wow, $HOOD has taken minority ownership in Crypto‌.com.

It does feel like the future of the finance industry is slowly consolidating around Stripe/Robinhood.

Especially after the successful expansion of Robinhood Chain, Prediction Markets, and Credit Cards/Banking. https://t.co/FXZTmVaIo5

翻译：
哇，$HOOD 已取得 Crypto‌.com 的少数股权。

确实感觉金融行业的未来正逐渐围绕 Stripe/Robinhood 整合。

尤其是在 Robinhood Chain、预测市场以及信用卡/银行业务成功扩张之后。 https://t.co/FXZTmVaIo5

Loading...