K

@Joeyet3 @revoAIution Good reply, we need facts not copium

对SIVE最准确的证据排序 如果不是给“MI500方案”排名，而是给Sivers现有证据强弱排名，我会这样排： Ayar Labs关系最强 Sivers已经获得过订单、参与产品 qualification 和量产准备，Ayar也公开确认其SuperNova使用Sivers激光阵列。这是实际客户关系。 GFS平台关系第二 Sivers进入正式参考设计和SCALE平台，扩大了被下游客户采用的机会，但目前没有披露排他性、出货规模和绑定客户。 MI500关系仍为未证实 AMD确认MI500会开始引入光学，但没有公开点名GFS、Ayar、Lightmatter或Sivers，也没有明确MI500采用哪一种CPO/NPO/外置光源实现方式。 Lightmatter是可信的替代风险 它与Ayar同时属于NVLink Fusion光互连伙伴，拥有自己的Passage和Guide体系。没有AMD连接证据，但不能简单判为“最不可能”。 对你持仓真正意味着什么 这段话加强的是： Sivers拥有两条真实的产业入口：Ayar客户关系和GF参考设计； MI500确认开始导入光互连，使整个潜在市场更可信； 如果AMD最终采用“Ayar＋SuperNova”或采用绑定Sivers的GF参考设计，Sivers会明显受益。 它没有证明的是： AMD已经选择GFS； MI500一定采用CPO而非其他光学形态； GFS客户必须采用Sivers； Ayar不会引入第二供应源； Sivers已经获得MI500相关订单。 最终判断： Sivers拥有MI500的合理期权，但目前没有MI500的确认内容价值。 原文所谓“最可能GFS、第二Ayar”，本质上是在给无法量化、而且层级混乱的供应链路径强行排序。若作为投资依据，应把结论降级为：SIVE是MI500光学转型的潜在受益者之一，尚不是已确认供应商。

Dug into every possible optical linkage for $SIVE and MI500 after AMD's own executive confirmed optics will be used. MI500 launches 2027, and AMD confirmed it uses both copper and optical connectivity. Here's is what I found from most to least likely. Most likely = $GFS. TechPowerUp reported back in April that AMD is returning to $GFS specifically for CPO manufacturing on MI500 by name. $GFS's own SCALE deck lists AMD as an OCI MSA founding member alongside $AVGO, $NVDA, Meta, Microsoft, and OpenAI. The official June 2 $SIVE and $GFS press release confirms $SIVE laser arrays are integrated into reference designs on $GFS's silicon photonics platform, available in the SCALE platform for CPO and linear pluggable optics. No other laser company is named anywhere in that official announcement. Second = Ayar Labs. AMD directly invested in Ayar Labs' Series E funding round. Ayar Labs joined Nvidia's NVLink Fusion ecosystem in June and already demonstrated a full CPO scale-up rack with Wiwynn at OFC in March. $SIVE has been Ayar Labs' confirmed laser supplier since 2022. No confirmation this specific relationship is tied to MI500 by name yet though. Least likely, Lightmatter. Also joined NVLink Fusion, making them a direct competitor to Ayar Labs in that same ecosystem slot. One unverified source claims AMD backing here too. This wouldn't touch $SIVE at all, so would be the worst outcome. Lightmatter builds its own proprietary laser in-house. Other real players in this space with no confirmed AMD connection. Celestial AI, now owned by Marvell. Intel Photonics. Nubis, Xscape Photonics, Ranovus, Scintil. $LITE and $COHR, both with confirmed $2 billion investments from Nvidia specifically, not AMD. Recap. Most likely, $GFS, $SIVE is the confirmed reference laser. Second, Ayar Labs, $SIVE is the only laser supplier. Least likely, Lightmatter, no $SIVE laser involved, bad outcome for the thesis if this is the one that wins.

@ThematicTrader why

I am not playing short side right now, but if I were to play it, I'd be short $BE here $223.

@WheelieInvestor @FeroceResearch ...

@FeroceResearch I knew it likely didn’t mean anything in terms of directionality but i didn’t know exactly what it was. I need to look more into the Arbitration.

SOMEONE JUST BOUGHT OVER $100 MILLION OF $MU CALLS EXPIRING NEXT WEEK I have never in my life seen an order this big for such a short dated options contract Someone knows something. Something BIG is about to happen

@michaelsikand @ParadisLabs Loaded $AVEX?

@ParadisLabs WHY I LOADED ON 7/16

This Google earnings release makes all the fear in semis from the past month look hilariously stupid. They will all print crazy earnings, from $LITE to $COHR to $SNDK to $NBIS. - Cloud revenue growth is accelerating - Gemini 950M MAUs vs. 750M in Q1 - Raise 2026 capex guidance to ~$200B from ~$185B Don't forget who held your hand through choppy waters, anon.

@babyfolio Still 80% port in $nbis?

I don't like using the term "easy money." But buying compute here feels like it. My top pick in this space is obviously $NBIS, but I think the broader ecosystem, AI clouds and colocation providers will continue to benefit. Some will do better than others, but rising tide will lift all the boats.

@michaelsikand Lol

$AVEX - This $3.6B Anduril competitor could be the next 5x drone stock. Calling it at $34/share as the most asymmetrical defense stock since $KRKNF. AEVEX went public on the NYSE last Thursday but FinX totally missed it. This 650 employee baby Anduril has shipped 10K drones to Ukraine, is guiding for $800M 2026 revenue, and has 100K sq. ft of manufacturing space that can push out 1000/units a month. Did I mention it beat out Anduril on an Army program? At $32/share it’s MC is over $1B less than $ONDS despite doing 5x the forward revenue, having a suite of highly integrated, battle tested products/software, a deep history as a government contractor, and winning two sole source Pentagon contracts worth $1.2B combined. It also trades at half of $KTOS on forward revenue, despite growing way faster (85% guide) and being a pure play drone stock. Has a $500M+ backlog and an $8B identified backlog amplified by $50B in proposed UAS spending by the DoD this year. What differentiates AEVEX from other wannabe drone stocks is they’ve been doing ISR missions since 2008 (founded by OG UAV Air Force guys) and have been pumping out thousands of attritable mass units for Ukraine for years. In 2022 the Pentagon needed thousands of cheap drones in Ukrainian hands in weeks, and $AVEX was the only one who could turn on the line. They delivered 4,400 units across 7 platforms for roughly $554M in contract value, with the fastest runs moving from order to warfighter in 9 days. The follow on was EUCOM AOR Deep Strike which awarded another $645.7M for 4,800 aircraft across 7 platforms, sole sourced again. What’s crazy is the diversity of $AEVX’s platforms. Their website feels like Anduril in that they have so many exciting products/designs from loitering munitions (Dagger) to long range drones (Raker) to even unmanned surface vessels (Mako). They also have a software platform called CompassX they expect every product to run on by year end (and be compatible with other manufacturers) AND a crazy moonshot product called ForgeX which is a mobile manufacturing facility for drones. ForgeX is literally a container placed near the battlefield and can pump out attritable weapons. I think there’s a strong margin of safety on the valuation side but the risks are worth mentioning. Basically all their revenue is coming from programs for Ukraine which could flip on a dime if the war ends or if the U.S. pulls support. AEVEX is spending way less on R&D than Anduril and Shield, which is not encouraging in a fast-evolving landscape. IPO proceeds are mostly being used to retire debt than fund growth and re-investment into the business. Lock up expires in 6 months in mid-October and MDP (a PE firm) holds 79% of the float. My rebuttal is that even without Ukraine the DoD is spending $50B of UAS with the new $1.5T budget and new theaters are emerging in Iran and the Pacific. R&D as a percentage of revenue is actually higher than public peers AVAV and KTOS. On the liquidity side, $375M of new credit facilities will help bolster things. I can’t find any reason to believe MDP is a bad actor or “worse” than other PE firms, but governance is always going to be a risk in this kind of set up. Overall, I don't think a drone prime as high quality and proven as $AVEX should be trading anywhere near the MC of a diluting, pump holdco like $ONDS, especially with the new defense budget and Iran War tailwind. Retail deserves a higher quality drone prime with credible contracts, manufacturing, and capacity. I think $AVEX is that stock and presents a phenomenal long for anyone trying to play the future of defense. $AVEX is at 4.5x forward sales at $32/share. $AVAV at 4.5x. $KTOS is 7.8x Anduril at 14x. $ONDS at 23x. Shield AI at 24x. Bear case: $31 ($AVAV floor) Base case: $56 ($KTOS, +74%) Bull case: $170+ (Shield AI / ONDS multiple, +430%) To read my free deep dive report on AEVEX I’ve linked it below.

@michaelsikand Lol

Holy sh*t guys. $KRKNF is a generational "fuck the institutions" trade for the little guys. Big funds can't own this stock. Citadel, Blackrock, T Rowe, DE Shaw etc. It's trading on a tiny junior exchange in Canada (it is the #1 company on it and outgrew it a while ago). Because of this, institutions own just 20% of the stock. We have 80%. This is one of those rare situations where we got rich and the elites didn't. As long as you ditch Robinhood and other useless brokers for one that enables OTC or TSX, you can buy the hottest stock the big sharks can't. For as long as this stock is on the TSX.V (I give it 9-12 months tops), there is tremendous upside for retail. We can keep pushing this higher to match U.S. drone comps valued way higher and force the institutions to pump our bags when they inevitably have to get in on an uplisting to the TSX and then again to the NASDAQ. This inevitable capital shift alone will contribute multi-bags. Next, this is our revenge against the private market elites depriving us from investing in the companies we literally care more about than them. Marc Andreessen and Josh Kushner out here writing a $4B check to Anduril like its pocket change. Just another easy 20x for them. There's no way we get to buy Anduril at IPO for less than $100B, maybe even less then $250B. Kraken is retail's way to share in the returns of what is basically guaranteed to be the next $1T defense company. Everyone keeps asking me if it's "too late" or if the stock has ran too much. It'll be early as long Ken Griffin and Larry Fink aren't in this stock.

@ren_stocks Lame af excuse

Yes to the 10 comments that corrected me. Yes AEHR instead of AHER. Mfs. I’m a Pikachu I can’t read for shit.

If this so happens to be the bottom, since I'm getting some sparks over here... I would be scooping up the following stocks as fast and as hard as possible: $SNDK $NBIS $SKHY or SK Hynix 000660 $AAOI $PENG $AHER Wishing you all success.

@TradexWhisperer @JackZhang70 Wtf u trash cunt test the prompt then post a reply. U r zero value as is. Womans day as it were

@JackZhang70 Is this real? If so signs of distillation just like Deepseek answers.

Kimi K3 costs as much to run as GPT 5.6 Sol Max. More than GPT 5.6 Terra Max. It's not just OpenAI and Anthropic burning through HBM anymore. Every new frontier model from every competitor adds another layer of memory demand at the silicon level. $MU $SKHY $DRAM

@michaeljburry Also fix ur wonky lid fuck

For a weekend deep think, consider this.

@michaeljburry Mr

@michaeljburry Get a real job

@AnikaSomaia @SemiAnalysis_ Changing em dash to normal dash doesn’t mean u real

a week ago @SemiAnalysis_ wrote that Chinese labs are "simply too compute poor to truly reach the frontier." today one of those "too compute poor" labs, a 300-person startup actually, shipped a model that compares to opus 4.8 the entire western consensus – export controls, the $650B hyperscaler capex race, the "compute moat" investment thesis – is built on one assumption: flops gate capability. if that were true, chip controls would keep chinese labs permanently behind the frontier. but after reading through moonshot's stack i no longer think it is. training is efficiency-compressible: MoE routing, INT4-native quantization, better data curation, infra built around scarcity (their Mooncake stack exists because they don't have gpus!). a small lab with taste can compress the compute needed to make a frontier model, even if it can't afford to serve one the frontier is no longer something money can buy

@Thecontrarianr1 @revoAIution Don’t get ur panties in a twist

Goldman Sachs published research on the global optical module market. Here is what it means for $SIVE. Total optical module TAM growing from $34.2 billion in 2025 to $72.6 billion by 2027. High speed 1.6T and 3.2T revenue alone growing from $3.9 billion in 2025 to $44.9 billion by 2028. How this relates to SIVE Silicon photonics revenue share is projected to surge from 28% in 2025 to 62% by 2028. SiPh penetration by volume climbs from 6% to 18% over that same window. Silicon cannot generate its own light. Every single SiPh module still needs an external laser fed in from outside. As SiPh eats more of this market, the demand for that external light source grows right alongside it, not the other way around. This report is about the broader module market and InnoLight specifically, not a confirmed $SIVE data point directly. But the mechanism is the same one running through $SIVE's exposure to Ayar Labs, $GFS SCALE, and the rest of the ecosystem. The bigger the SiPh wave gets, the more laser demand it creates underneath it. The more potential for $SIVE

@jukan05 Source?

My conclusion so far is that Kimi K3 may be the first model to narrow the gap with leading U.S. closed-source models to less than three months.

@PimDeWitte @NonyaBi10393887 Also u really think a bot reads half cropped image and makes a comment? Not that far. Prob this just engagement bait

@NonyaBi10393887 Thank you for your advise on cropping, Nonya Bidness 10393887 (definitely not a bot, by the way)

kimi is truly a revolutionary model

@PimDeWitte That's an interesting error message you tried to cut out of the screenshot. Doesn't look like the tool you're using is actually using Kimi - which is why you received this response. Nice try though... repost since i dodge the other