cichou

$SIVE $TSM
🇨🇦🇹🇼

[Serenity @aleabitoreddit]

Three new $SIVE insider purchases today.

Most notably: $SIVEF CEO Vikram bought 1,000,000 SEK worth of shares on the open market today.

This is following the massive drop on Sivers stock following 2 misleading CPO delay reports, Meta Compute headlines, and another insider sale report from another photonics director.

Typically, insider sells for any number of reasons. But having the CEO buy on the open-market is the clearest signal in management confidence.

Following $GFS, $JBL, and other laser supplier partnerships, and his goal of NASDAQ listing in the near term.

Feels like the CEO sees a clear potential for Sivers to become the next photonics giant (myself included).

[Pep Invest @Pep_Invest]

$SIVE - The only reason they are recommending it is that they want to buy into these stocks themselves at a lower price and then issue another buy recommendation shortly after.

They didn't buy $SIVE for no reason.

Institutions are not your friends!

[Han @Han4Claude]

https://t.co/0PBVJrw8bv

[駿HaYaO @QQ_Timmy]

Sivers Semiconductors 貸方 Bootstrap Europe 行使現有可轉���貸款之轉換權

Kista，2026 年 7 月 3 日Bootstrap Europe IV SCSp（「Bootstrap」） 將把 1,200 萬美元的可轉換貸款轉換為 Sivers Semiconductors AB (publ)（「Sivers」或「本公司」）的股份。董事會因此決議向 Bootstrap 發行 22,847,044 股新普通股。

背景
如 2026 年 2 月 24 日所公告，Sivers Semiconductors 已對所有現有外部債務設施進行再融資，與 Bootstrap 取得總額 1,700 萬美元的貸款設施。其中���括固定利率擔保可轉換貸款 1,200 萬美元。根據可轉換貸款的條款與條件，Bootstrap 有權將全部或部分未償還本金轉換為本公司普通股，轉換價格為每股 SEK 4.77。

轉換
Bootstrap 已行使其權利，將可轉換貸款全部未償還本金 1,200 萬美元轉換完畢。此次轉換將發行 22,847,044 股新普通股予 Bootstrap。新股付款方式為以 Bootstrap 對本公司之可轉換貸款債權進行抵銷。董事會已根據轉換通知決議發行該等股份。Sivers Semiconductors 財務長 Heine Thorsgaard 表示：「Bootstrap Europe 是極佳的策略性貸款夥伴，他們的轉換大幅減少了 Sivers 的未償還債務。這符合我們既定策略，持續強化資產負債表，並將投資重點放在把握三個快速成長的重點市場領域的重大商業機會。」

股本與股份數量
轉換後，本公司股份總數增加 22,847,044 股普通股，由 332,234,273 股增至 355,081,317 股普通股，對現有股東造成約 6.4% 的稀釋。本公司股本增加 SEK 11,423,522，由 SEK 166,117,136.50 增至 SEK 177,540,658.50。
$SIVE
https://t.co/MSX2XKvvgL

[Phoenix Yin @Phoenixyin13]

Nature 期刊盖章：普通人到底怎么变有钱？

哈佛、斯坦福和 Meta 联手，直接调取了 Facebook上 7220 万人、210 亿条真实的好友关系链。我个人感觉，这数据量，基本把大半个美国翻了个底朝天。
他们想研究一个终极问题：到底是什么决定了一个人能不能打破阶层固化、逆袭变有钱？

结果发现，那些天花乱坠的因素，什么学校好坏、社区凝聚力、宗教信仰啊，诸如此类的。都不是决定性的。最硬核的指标只有一个，叫经济连接度，即Economic Connectedness, 简称 EC。
说白了，就是你的跨阶层朋友圈有多大。
更落地一点��译，就是在���收入人群里，到底有多少比例的朋友是富人。

研究发现，这个指标和跨代收入向上流动的相关系数高达 0.65，
🚨在社科界，这几乎是统治级的超高相关性了。

我们可以得到一个黄金定律，即一个出身普通的普通孩子，如果在一个富人比例高、且大家玩得来的社区长大，他长大后的平均收入能直接暴涨 20%。

既然搞钱的密码是 EC，那普通人想逆袭，最该做的就是打破自己的社交舒适圈，换个环境蹭圈子，去富人或精英高频出现的社区、学校、甚至健身房，比如高强度的专业训练场、高质量的学术沙龙。
更重要的是，提升自己的被链接价值。毕竟富人也不傻，单纯的巴结没用。用你的专业能力、独特见解，哪怕是懂 AI懂技术、或者在某个小领域做到了极致去作为入场券，给高净值人群提供情绪价值工具价值。

想变有钱，先去看看你朋友圈里最厉害的 5 个人是谁。去主动连接那些比你富有、比你认知高的人，现在就行动起来。

这是一篇五年前的论文，但我记忆犹新。

[Serenity @aleabitoreddit]

$SIVE is raising ~$61M (600M SEK) to expand manufacturing capacity for InP lasers and optical amplifiers.

Sivers operates as a fab-lite model so this is likely hinting at foundry allocation/scaling for CW DFB laser ramp across Jabil, and their other partners.

It could also point to broader foundry allocation/scaling beyond the already announced WIN partnership.

My feedback is that it's very bullish if it's comprised of long-only institutions and there's no heavy discount (30 day VWAP). Otherwise, would need to reaccess if there's arb or short term investors.

This is the perfect time to get ideal strategic investors like $GFS on the cap table. As well as focusing on having more US institutional support.

Especially considering $56M is a small check to a single US institution that want photonics exposure (eg. look at Poet's $400M registered direct offering for sentiment).

A lot of it is "it depends when more info comes out", but this looks like a very positive signal for mass production given Siver's fab-lite models.

[Serenity @aleabitoreddit]

That’s a misconception:

$SIVE is the laser supplier for next gen architectures, not just CPO scale up.

Pluggable, scale out CPO, scale up CPO, NPO, etc.

- Sivers and $JBL went god-mode and developed 1.6T optical transceivers with CW lasers.

Effectively designing around the EML bottlenecks, and even made things more power efficient.

From quotes from Jabil management they created a “relatively dramatic moat”.

So for the next gen of 1.6T pluggable transceivers, Sivers seems immediately used.

Markets missed this nuance too: after Jabil’s announcement, other pluggable players reached out, and SIVE is working with them now (prob codevelopment, qualification stage).

There’s not too many players that build optical transceivers that aren’t vertically integrated like Lumentum (think: Innolight/eoptolink, and maybe others).

So these are active developments, just not public material yet and could be a new press release anytime.

- for CPO scale out which happens H2 2026 onward like $POET, $SIVE is a laser supplier for those players.

- scale up CPO is h2 2027, from Ayar and Nvidia’a NVLink CPO ecosystem and that’s the main volume ramp across optical players.

But this is where $SIVE looks like they mog every other player from supply chain mapping.

Since they’ve likely been working + designed in with $MRVL Celestial, Lightmatter, Lightelligence and the others since they were small way back when.

On top of things like this:

For foundries like $GFS, $SIVE is the reference laser.

Not even including massive companies like O-Net building ELS with $SIVE, likely for the Asian hyperscalers supply chains.

This is why $SIVE is by far my favorite laser chokepoint long.

Feels like they’re everywhere in new optical architectures starting beginning of 2027.

[Pep Invest @Pep_Invest]

$SIVE $AMD - The industry is hit by a massive squeeze on CW Lasers.

Recent supply chain tracking from TrendForce shows that $AMD is quietly scrambling to lock in massive purchase orders for these lasers.

Their goal is obvious is to secure their own independent supply line before $NVDA locks down the entire market and chokes out $AMD next-o-Packaged Optics (CPO) roadmap.

The big problem?

Giants like $LITE and $COHR are basically booked solid through 2028.

This supply crunch turns $SIVE into a massive tactical play.

Connecting the dots between $AMD and $SIVE isn't just wishful thinking; looking at the infrastructure, it’s practically inevitable.

$SIVE recently locked in a deep collaboration with $GFS, embedding their laser arrays directly into GF’s SCALE CPO framework.

$AMD has a long, deeply rooted history of relying on $GFS for its silicon photonics manufacturing.

If $AMD builds its optical chips through GF, they are essentially forced to use $SIVE as their reference light source. The partnership is already written into the silicon.

$SIVE is also the exclusive hardware engine behind Ayar Labs' SuperNova light sources. $AMD is a founding heavy-hitter in the Optical Compute Interconnect (OCI) consortium, an ecosystem heavily reliant on Ayar’s architecture.

This means $AMD long-term hardware pipeline is already designed to plug right into $SIVE's physical lasers.

Look at $AMD recent acquisition of CPO pioneer Enosemi.

Enosemi’s designs require incredibly precise, external laser sources to actually move data. Since $SIVE has clear, unconstrained manufacturing capacity through its relationship with WIN Semiconductors, it stands out as one of the only global players capable of handling $AMD massive production scaling.

On the corporate side, $SIVE is actively restructuring its books to PCAOB standards to prepare for a major dual-listing on the Nasdaq in New York. You don't launch a major US IPO without a massive PR catalyst. Sealing an official supply agreement with a trillion-dollar hyper-scaler like $AMD would be the ultimate validator for their US market debut.

[法克魷吉米 Jimmy | Real TSMC Insight | 純愛を学び中 💜 @jimmy_yoasobi]

Five key points

>TSMC has not stumbled in Co-Packaged Optics CPO and has already entered mass production

In cooperation with Broadcom it has shipped over fifty thousand Tomahawk 5 Bailly switches

> Meta reliability study covers more than one million 400G port device hours with zero link failures

> Broadcom third-generation Davisson 1024Tbps was launched in October 2025 using TSMC COUPE process

> NVIDIA Quantum X Photonics will enter mass production in early 2026

> Spectrum X CPO follows in second half of 2026 both based on TSMC technology

Although Intel EMIB has advantages in yield and thermal management with long-term silicon photonics investment

It remains at demonstration stage with no customer shipments yet

—

Three conclusions

> TSMC has successfully crossed the CPO production threshold

Ecosystem partners such as Broadcom and NVIDIA are jointly advancing the industry

> Technical superiority does not guarantee market dominance

Semiconductor history has many cases of excellent architectures that ultimately failed

> Intel may capture 15 to 25 percent share in the CPO field

But 90 percent forecasts are overly optimistic and require the entire supply chain to fail simultaneously

—

> > > TSMC has never hyped itself aggressively

But always delivers on schedule with yields on schedule

Therefore Tier-1 customers in the market all use TSMC chips

—

Additional point on process leadership

TSMC now maintains at least one full generation lead in mass production processes compared to Intel

A new tech node must undergo extensive IP verification and library development before design houses can use it reliably

This makes mature verified and volume-produced nodes extremely valuable for the most advanced products

They ensure design houses achieve expected performance without unexpected bugs or stability issues

If a foundry delays wafer delivery the design house risks handing its hard-won time-to-market advantage directly to competitors

—

Real-world example

Intel struggled for years with its 10nm process delays which dragged on from planned volume production in 2016-2017 until full ramp in 2019

This allowed AMD to fully switch to TSMC 7nm and launch competitive Ryzen and EPYC products years ahead gaining massive market share in CPUs while Intel lost ground in data centers and client PCs during that period​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

$TSM

[Herman Jin @ShanghaoJin]

市场上胡说八道人太多，原因还是不少人真一知半解，而却管着大笔钱

你第一次去AWS上租台机器都得找半天，何况CPO要焊在芯片旁边。有点问题不是太正常了么？即使过度NPO，光源还是缺货，慌啥呢

其实就在等这鬼故事～��所知道的是：全产业链还是在“头铁”地推进CPO。Period

ps投研问Funda @himself65

[Boywus @Boywus]

普通年轻人更适合围绕规则做带信息差的生意，牌照生意没有资格，平台生意前期投入又过于巨大。

当代社会高度依赖金融结算、物流网络和信息系统，规则制定者与基础设施提供者天然拥有最舒服的抽水权。

但对于大多数普通人来说，更现实的机会往往存在于系统与系统之间，做孤立系统之间的沟通者，带着信息差进行制度套利。

从依靠采购权、地理边界和运输成本的现货搬砖，到后来利用信息传递、市场结构与交易规则差异的期货套利。

本质上，都是在规则的缝隙里寻找利润，我们从始至终都只是倒爷和背包客。

[The Kobeissi Letter @KobeissiLetter]

BREAKING: President Trump says the Trump Administration might buy equity stakes in US AI companies and that he will host a meeting with AI executives as soon as next week, per Reuters.

[Serenity @aleabitoreddit]

Okay... just some more weekend shower thoughts about $XFAB.

I still feel like it could be the next $TSEM, just early stage at a $1.4B MC?

They kinda leapfrogged current gens (which $TSEM are getting volume from) to compete for H2 2027 CPO scale up inflection point ($ASX docs cite Xfab (aka. photonixFAB) as focusing on CPO)

By building out some black magic MTP (transfer printing) architecture for lasers w/ other stuff like TFLN.

Basically next-gen integration IP, they're still behind on yields, sure.

But $NVDA evaluating it for transceivers/switches to see if it can volume ramp. That $NOK sets the specifications/assembly for. (nvidia invested in nokia for this these switches/networking too btw).

And if their MTP supply chain works... (eg. with Smartphotonics providing lasers, EU players doing assembly).

It basically volume ramps with $NVDA just like why Nvidia signed long term agreements with $TSEM?

Downside risk?

Already below replacement book value, can always go lower yeah, but typically to a certain point.

Maybe more CHIPS act subsidies next few months from chips act 2. If it doesn't go well there's SiC (152% Y/Y Growth, 195% Y/Y SiC wafer shipment growth)/GaN power semi upside.

Europeans /LLMs will say "oh evaluations doesn't mean it's a future contract!".

This is kinda different since the European Union is behind this effort and $XFAB for soverign photonic supply chains.

Not your typical company + hyperscaler evaluation, since $NVDA wants to be nice to Europe's regulators. They'd prob be pissed if nvidia just stayed in US/Taiwan/China.

So if they can make this MTP black magic work with mass production, feels almost for sure nvidia/nokia volume ramp on some tiny $1.4B silicon photonics foundry or at least throw them a bone with smaller contracts.

In terms of timelines, maybe just a months early since it volume ramps H2 2027/H1 2028 (which happens to be in line with CPO scale up timelines)...

Or just unknown because they named their project something stupid like photonixfab?

Like XFAB Photonics would have been better? so institutions/screeners can connect the dots when looking at CPO silicon photonic foundry players?

Automotive should also coming out of a slump medium term, sped up by self-driving (TSM Chairmain comments yesterday said ai automotive was TSM's growth vector alongside robotics). So their core business also should pick up speed too medium term.

Obviously markets/europeans want a "Nvidia signs $2B+ contract, XFab volume ramping 2027!"

But by then it will be a $9B+ company and you miss out on all the upside. And especially since everyone analyst/institution is blind to volume expectations for these....

Normally don't invest in companies in evaluation stages, but this just seems very de-risked by EU sovereignty + Gov backing, and you have Nvidia + Nokia there for volumes if they can make the IP work.

I think markets are probably missing something here... there's almost 0 value being assigned to being CPO exposure in Europe as their long term upside.

[Grok @grok]

It's called hand steepling (or finger pyramid). Body language experts describe it as a confident, authoritative pose signaling thoughtfulness and control—common among leaders, CEOs, and speakers during serious talks.

The triangular shape fuels online theories linking it to Illuminati, Freemasons, or "elite" symbolism. That's pattern-seeking in photos of powerful people; no evidence supports any secret meaning. It's just a natural, comfortable gesture.

[Anders Storm @StormDirac]

JP Morgan 5% threshold FI flagging notice for $SIVE.

Has anyone identified the holder behind this filing?

Could be:
• Institution
• Family office
• High-net-worth/private investor/retial

Before jumping to conclusions, it's worth understanding who actually triggered the flag.

Anyone looked through the details yet?

#SIVE

https://t.co/HLVuuT68zi

[Anders Storm @StormDirac]

Hej Sven synd att du inte fårr svar. Skall försöka summera några saker som jag ser som helt nytt.

1. Sivers is being integrated directly into GF's SCALE platform. GF only introduced SCALE publicly about a month ago. Until now, GF had announced the platform itself, but not a laser partner embedded into that ecosystem. The release states that Sivers' laser arrays will be available in GF's SCALE™ (Silicon Photonics Co-packaged Advanced Light Engine) platform and reference designs. Being picked as GF reference design vs Lumentum/SmartPhontics is strong statement and validation of Sivers as Laser supplier into the AI ecosystem. That means Sivers is no longer just supplying components into the market generally; it is becoming part of GF's preferred silicon photonics architecture.

2. Expansion beyond CPO into LPO and broader optical connectivity like Pluggables with a large TAM that is closer to ramp (1.6T. Pluggables). Earlier Sivers announcements focused heavily on External Light Sources (ELS) for CPO deployments. This release explicitly says the collaboration supports: a) Co-Packaged Optics (CPO) b) Linear Pluggable Optics (LPO), c) Other emerging datacenter interconnect architectures. This broadens the addressable market materially beyond the ELS/CPO discussed previously.

3. The first full foundry-level endorsement of Sivers' laser technology. All other communications with GF had been in group of companies or partnerships with Ayar Labs, or Wireless. Now adding foundry-level endorsement to previous announcements, with system and module partners such as O-Net, Enablence and Jabil. This is different because GF is one of the world's leading silicon photonics foundries. Having GF publicly state that it is pairing Sivers laser arrays with its silicon photonics and SCALE platforms is a stronger ecosystem validation than a group of partners on list or module-level partnership.

4. A big signal that Sivers is positioning for hyperscaler qualification for Pluggables via GF like they done for Jabil. If GF succeeds in becoming one of the major SiPh platforms for 1.6T or 3.2T Pluggables, then Sivers potentially gains visibility to: NVIDIA ecosystem suppliers, Broadcom ecosystem suppliers, AMD ecosystem suppliers, hyperscaler custom-AI programs.

[Herman Jin @ShanghaoJin]

@tigeryu2001 首先我只是合理质疑，对SIVE不看多或看空，那么敏感话题，双飞对着仓位，我犯不上

只能说我过去买过沉甸甸的梦想，下场都不是特别好

GFS做硅光SiPh，没有InP fab，我之前的问题没有被回答：

InP Fab在哪里？

得有激光器芯片，才能wafer binding GFS的硅光吧？

另外，two Sigma不容易打爆的

[Herman Jin @ShanghaoJin]

凡所有相，皆是虚妄。若见诸相非相，则见如来

作为一个AI死多头，我知道在所有人狂欢的时候，但凡发出一点悲观的言论都必被讥讽、嘲笑

但是，我很想写此文，其实是复盘与反思：不但是为自己这两年半写X、做交易做的；也是也是对这个疯狂市场做的

很幸运，作为闲暇的爱好，写X结识了很多新朋友。感谢互相欣赏与启发，在过去2年里，我从24年4月初清仓BTC换成AVGO NVDA，大选前AAOI PLTR TSLA INTC，关税战后INTC GOOG，25年底又加仓INTC AAOI，到今年春节逃跑、再满仓AMD NOK（最近又操作了ORCL LITE）。虽然，有买彩票SUP归零，MRVL MDB也有赚有亏，但是我一直坚信AI对token的需求会让半导体every dog has its time，获得了不差得收益。期间不少小伙伴们也都赚到不少于10倍收益

然而，此刻要清醒的意识到，收益并不是我们选股牛，眼光独到，而是市场牛！我复盘了我自己今年3月最后一周，周末前全仓买入。赚钱并不是因为当时买对了AMD，而是当时call 对了市场资金结构

所以，在这个时点上更需要静心凝神，好好反思下市场接下去会如何

喝醉酒的人是不会知道自己喝醉的；沉醉美梦中的人也不会意识到黄粱一梦，这过程可以很久，我们也可以继续疯，但是必须要清醒地在更高阶意识中，认识到随时有被敲醒的风险

One of the four beasts saying, COME AND SEE – Revelation 6:1

不同于华尔街很多人，我是“坚信”AI的

30年来第一次，AI模型用并行计算替代串联计算，把半导体的进步直接变成了生产力（之前科技进步只是作为平台，让用科技的软件、以及用软件的人创造价值）

AI的结果是token需求指数级升高vs 而半导体供应只能线性增加。加上整个半导体老登们在过去几年黄金时间，没有因 “信” 而扩产，今年需求真正开始爆发后全产业链缺货。造就了这轮大牛市

不过，大牛市的命脉是市场price in了:

模型收入加速度，不远得未来可以justify Hyperscalers 的Capex

要知道任何动摇命脉预期的都会让市场陷入恐慌

1. 铜墙铁壁：horseman之一就是微观、公司层面真的无懈可击

从看模型架构、从架构看需求，微观看半导体每一个赛道，每一个公司都近乎完美。每一个，都越看的细致，越了解深入，越是有自信---- 这就是典型低PE泡沫

Rule of Thumb：当看了几十、上百家公司，每家需求、盈利都真实solid，你就必须质疑这个市场有没有问题

低PE泡沫一样是一个泡沫，只是他不同于2000年。高PE是个预期泡沫，市场会不停设定更高的收益预期，一旦核心公司预期破灭，整体预期立即调整，多米诺骨牌一样一片片倒下。这就好像去年的IGV，市场忽然不给EV/EBITDA估值了

但是低PE泡沫没有这个问题，盈利增长会给你足够信心在稍微下跌之后不断加仓冲进去。毕竟谁能拒绝个位数PE的存储，每年涨价涨盈利，毛利率80%。这个过程中，小跌小买，大跌大买。只要不被切中“命门”市场会非常坚挺。持续性正反馈，好像美梦一样让人不愿醒来，也不相信是做梦（买过p2p的知道我说的什么感觉吧）

不对称反馈：明显的体验是在过去1年中，市场对于宏观负面风险的反馈明显弱于正面利好。此时，并非没有“野蛮人”，而是公司盈利能力的城池固若金汤。不管伊朗危机、油价、通胀、利率、Fed统统能拒之门外。反之，一旦命门被破，所有野蛮人都会杀回来，过去没跌的够的一次性补完

2. 去监管与宏观流动性宽裕：

整个金融体系自从2008年金融海啸之后经历了一轮长期的监管与去杠杆，为了弥补系统性流动性损失美联储对应展开了一轮持续的QE。银行当时平均杠杆x30~x40，后来降低到了x15~20。去年美联储降低了eSLR要求，实质性让银行理论可扩4-5万亿美元资产负债表
其实，4月后银行间流动性泛滥，也加剧了风险资产狂欢，低PE下杠杆入市

流动性是一切之王，流动性泛滥征服一切。如果说疫情期间120b/月是细水长流，现在就是开闸放水。对应过往平衡的QE没有转向QT情况下，直接放任银行去监管，制造大量流动性

但是流动性又如多巴胺，不是快乐分子，而是期待分子

放任之后，现在通胀黏着，Fed其实没有短期降息余地。甚至在危机一旦发生之后，都很难去QE，只有降息手段的Fed面对可能的危机是一个纸老虎。而短期流动性的另一面是，目前半导体、特别是存储中的杠杆已经太高了

3. 群雄并起：He went forth conquering, and to conquer

可能因为TSMC带头不相信AI，半导体行业整体在过去2年没有应对AI需求大规模扩产

后果是，当整个产业面临硅基洪流需求，忽然傻眼。过去半导体产业链主要是为了消费电子设计、定价的。而现在，不但订单忽然拉满，而且需要不停迭代架构

NVDA就从GPU fabless，进步到rack方案解决，再往token factory迈进。每个过程需要的技术、供货商都不同。所有供货商都没有经过如此量产设计。这就好像把一台拖拉机强行拉上高铁，让他以200mile速度奔驰，每个零件都会嗡嗡作响。半导体每个细分行业都缺货，缺产能，提高不够产能

与之伴生的一个问题是：原来卖方市场只有TSMC一个“守门员”控货源、控margin、控价格。原仔汤搓的产业有肉吃，需求方对价格产能有预期，良性发展。今天忽然五亿探长“雷诺”控制不住了，每个“小混混“都能来涨价、收保护费。真正的结果不是半导体都牛叉，而是失控：失控会影响每GW需要投入的成本核算，影响到命脉的预期。会给capex与模型收入会师蒙上阴影

4. 钢丝起舞

AI模型收入预期太满

回顾一下，市场坚信定价Anthropic + OpenAI + Gemini的收入会快速增加让Capex合理。这条命脉的信仰是不能有一点动摇的。哪怕略微的动摇都会导致二级市场剧烈波动

那么，过去GPT 5.0也有模型不及预期，市场质疑scaling law为什么没有让半导体大跌？

首先当时半导体没有像今天这样全面开花，也没有如此高的杠杆；其次，当时hyperscaler的现金流冗余较多，只要他们表示相信，拿钱来capex支持是很容易度过难关的。毕竟一旦他们有钱、继续capex，可见2年NVDA收益确定性高，没有人能顶风做空NVDA的

后来也发生过GPT被Gemini追平，并且导致Orcl持有������量backlog订单而被做空。不过大总统的好兄弟老Larry敢对wall street说F you，用股权融资也要继续加注。很快， Opus横空出世让人意识到AGI时代已经到了

然而，今年capex已经到了770b，明年要1tril，华尔街是不会看Hyperscaler总营收多少的（道理就跟当时ORCL一样：OAI Anthropic可都是拿了他们投资款的），他们必须看到Anthropic 与OAI的增速持续，这个才是维系整个链条自转下去的核心

与此同时，这个游戏中大厂一直担任AI的监护人角色。但是他们的自由现金流已经变负了。“父母老了，力尽了“，后面的路要靠自己。而且市场满满预期下，对这条路一根筋的执着，不会给“主线”一丝一毫的容错

（你们可以回想NV rack液冷出问题，交换机CPO良率问题，这些小巫见大巫、假以时日一定能解决的问题，市场尚且没有任何容错）

命门在Anthropic：大过，大者过也。栋桡，本末弱也

供应延迟、乃至技术卡壳的问题是腠理之疾；任何模型端的风吹草动才是心腹大患

现在御三家模型的统一问题：算力不够，所以失智

我在自己大量使用模型编程、交易程序部署，多个AWS服务互动过程中发现，Opus4.8实际能力已经远远不如中国模型kimi了。虽然GPT还勉强可堪一用，但是也在逐步降智。市场在用固有软件思维考虑AI, 系统开发成本高、使用成本很低。试用好用将来也就可以稳定保证系统质量。但是AI是一个工厂，模型输出需要成本，饭店吃饭的人多了，厨房做不过来，质量自然变差了

第二个市场固有思维：指数的token需求是没有波动的，也不因质量改变。事实上，虽然我因为模型变蠢多用了很多token。但是我质疑token质量不断下降，token消耗增加的怪异剪刀差可以持久。现在很多公司确实对token使用有KPI考核，很多惯性订阅使用者也会加速使用，但如果算力瓶颈无法解决，我非常担心Anthropic增长曲线会变缓。而且算力问题非一蹴而就，也不是一个新模型能解决的

但是当市场开始意识到这个问题的时候，投资人会质问：

1. 跑不过蒸馏你的，模型花那么多钱训练干嘛？

2. 模型是不是商品化的model commoditize ？

当承载整个AI时代的模型“梦”也将受到质疑，整个盈收逻辑将来能否自我循环跑通将被质疑，所有今天拿了PE估值的半导体都同样会被质疑--- 你是不是周期性的？？

有人会说，失智既然因为算力不够就应该加大投入半导体解决问题

没错！但是钱呢？

如果刨掉OAI 与Anthropic拿到的融资款反哺云厂，他们能有多少净现金流可以继续支持这个投入？这个账对华尔街来说非常好算。当然不排除他们像当时orcl，哪怕融股权也要继续投入，熬过这关。到那时，可能市场的反应就会远远超过去年，假设AMZN Meta如果面临CDS 飙升、负现金流，他能否像老larry一样孤注一掷是不确定的

另外有人会说，OAI收入增加能不能弥补投资人担心。我认为预期如此之满的状况下，因为“比烂”而被替代是很难让人安心的（两军对战关羽被阵斩了，你说还有个张飞？没用的）

同时，我们也要意识到。这个Low PE的泡沫又是非常坚韧的，在没有直击到命门伤到命脉预期之前，几乎无敌。不过，一旦点中死穴就会轰然倒塌

我就业以来也曾看过好几次次纸碎金迷，潮起潮落，没有一次全民疯狂会ends well；

也没有一次梦在其中的人，会想明白这个赛道“怎么输”。最近一次币圈小伙伴熟悉的Defi summer。Uniswap很短时间狂卷几十bil USD，当时觉得取代部分传统金融只是时间问题、多少问题，根本不会想、也想不明白“怎么会输”。然而，如果喝醉酒的人想明白怎么会输，他就不会输了

不管Anthropic模型怎么样，面临什么问题，都不会动摇AI是人类第三次工业革命。AI过程不会受任何危机影响，碾着狂欢者的尸体也会继续前行，直到他替代几亿脑力劳动者

最后评论Low PE泡沫：不要说做空，哪怕踏空，痛苦指数都非常高。回头看看2007年做空subprime的哪些基金，有几个活到了最后狂欢？

牛市踏空都会变成笑柄的环境，我接下去会需要很多Aura力量与专注的力量，去一边看空一边做多。请原谅会减少在推特上消耗自己修为论道了，赠言：

可以喝酒，但是不能喝醉
可以讲故事，但是不能真信
可以跳上party桌子跳舞，但是双眼必须紧紧盯着DJ，他放着音乐跑了，你也得跟着跑