RSS Parrot

BETA

🦜 Zephyr / @zephyr_z9

@nitter.privacyredirect.com.zephyr.z9@rss-parrot.net

I'm an automated parrot! I relay a website's RSS feed to the Fediverse. Every time a new post appears in the feed, I toot about it. Follow me to get all new posts in your Mastodon timeline! Brought to you by the RSS Parrot.

---

Twitter feed for: @zephyr_z9. Generated by https://nitter.privacyredirect.com

Your feed and you don't want it here? Just e-mail the birb.

Site URL: nitter.privacyredirect.com/zephyr_z9

Feed URL: nitter.privacyredirect.com/zephyr_z9/rss

Posts: 120

Followers: 1

If this thesis holds, then it all comes down to which XPU designer can secure the supply chain and get as much volume as possible "At this point, does performance even matter as long as the chip works decently well? What exactly is the moat of ASICs?"

https://nitter.privacyredirect.com/zephyr_z9/status/2057490567286898808#m

Published: May 21, 2026 15:55

If this thesis holds, then it all comes down to which XPU designer can secure the supply chain and get as much volume as possible "At this point, does performance even matter as long as the chip works decently well? What exactly is the moat of ASICs?" …

Must read especially for the NAND bears

https://nitter.privacyredirect.com/zephyr_z9/status/2057488681599123676#m

Published: May 21, 2026 15:48

Must read especially for the NAND bears Ben Bajarin (@BenBajarin) The market is not ready for what is about to happen with storage thanks to agentic. I explain why, and share our beneficiary map. The uncomfortable version of this thesis is that if…

More price hikes

https://nitter.privacyredirect.com/zephyr_z9/status/2057463644955562155#m

Published: May 21, 2026 14:08

More price hikes 🐇 (@1p_semicon) 大同特殊鋼 > 半導体製造装置向けステンレス鋼については、バルブ・マスフロコントローラー・フィルター関係など全体的に足元かなり受注が増加している daido.co.jp/common/pdf/pages… — https://nitter.privacyredirect.com/1p_semicon/status/2057413193199403167#m

Great performance from legacy semis

https://nitter.privacyredirect.com/zephyr_z9/status/2057455186013282796#m

Published: May 21, 2026 13:35

Great performance from legacy semis Zephyr (@zephyr_z9) I haven't worked on an article in the last 2 months But writing this with a team is extremely fun 35 pages of pure Alpha, multiple ideas Enjoy!!! citriniresearch.com/p/semis-… —…

G6: Floppy disks near Uranus

https://nitter.privacyredirect.com/zephyr_z9/status/2057448275863884227#m

Published: May 21, 2026 13:07

G6: Floppy disks near Uranus SemiAnalysis (@SemiAnalysis_) With modern agentic workloads and long context windows, a common bottleneck in serving LLMs at scale is where to store all the KV cache. Luckily, KV cache can be extended beyond HBM into other…

Nvidia will allow customers to source their own DRAM and start selling systems without it

https://nitter.privacyredirect.com/zephyr_z9/status/2057433929356108203#m

Published: May 21, 2026 12:10

Nvidia will allow customers to source their own DRAM and start selling systems without it Jukan (@jukan05) Rumor: As memory prices have surged, memory has come to account for an excessively high share of the BOM, so Nvidia is reportedly carrying out…

RT by @zephyr_z9: Rumor: As memory prices have surged, memory has come to account for an excessively high share of the BOM, so Nvidia is reportedly carrying out some system-level optimizations. Internally, the company is said to be discussing reducing the amount of system DDR memory in some Vera Rubin configurations — translator’s note: this likely refers to LPDDR — while keeping HBM capacity unchanged.

https://nitter.privacyredirect.com/jukan05/status/2057406900367712522#m

Published: May 21, 2026 10:23

Rumor: As memory prices have surged, memory has come to account for an excessively high share of the BOM, so Nvidia is reportedly carrying out some system-level optimizations. Internally, the company is said to be discussing reducing the amount of system…

RT by @zephyr_z9: For Qwen3.7-Max, we have invested far more compute into RL training than ever before. Its top-tier AA score confirms the resulting general and agentic capabilities. This is just the start. We will firmly push forward RL scaling to build more powerful Qwen models. Stay tuned!

https://nitter.privacyredirect.com/ChujieZheng/status/2057403166589956518#m

Published: May 21, 2026 10:08

For Qwen3.7-Max, we have invested far more compute into RL training than ever before. Its top-tier AA score confirms the resulting general and agentic capabilities. This is just the start. We will firmly push forward RL scaling to build more powerful Qwen…

RT by @zephyr_z9: 8 years at Jane and started his own quant shop, now at deepseek making a cli agent? 👀

https://nitter.privacyredirect.com/dhtikna/status/2057309588278714514#m

Published: May 21, 2026 03:56

8 years at Jane and started his own quant shop, now at deepseek making a cli agent? 👀 Deli Chen (@victor207755822) Bty，the owner of this project is one of our most gifted （richest, also🤣）workmate: Tianyi(linkedin.com/in/tianyicui/), don't hesitate to…

RT by @zephyr_z9: On April 27, we called out the possibility that Socionext could announce a potential hyperscaler customer, and today Socionext’s CEO acknowledged that the company is likely to participate in Arm’s AGI CPU. As a result, the stock is now up 18%. There is just too much alpha in Citrini Chat…

https://nitter.privacyredirect.com/jukan05/status/2057309184220356953#m

Published: May 21, 2026 03:55

On April 27, we called out the possibility that Socionext could announce a potential hyperscaler customer, and today Socionext’s CEO acknowledged that the company is likely to participate in Arm’s AGI CPU. As a result, the stock is now up 18%. There is…

RT by @zephyr_z9: Bty，the owner of this project is one of our most gifted （richest, also🤣）workmate: Tianyi(https://www.linkedin.com/in/tianyicui/), don't hesitate to join us~

https://nitter.privacyredirect.com/victor207755822/status/2057304008327627067#m

Published: May 21, 2026 03:34

Bty，the owner of this project is one of our most gifted （richest, also🤣）workmate: Tianyi(linkedin.com/in/tianyicui/), don't hesitate to join us~ Deli Chen (@victor207755822) 🚀 We’re hiring! DeepSeek is forming a new Harness team to build Code Harness…

Jensen's biggest regret

https://nitter.privacyredirect.com/zephyr_z9/status/2057284012998381662#m

Published: May 21, 2026 02:15

Jensen's biggest regret Techmeme (@Techmeme) Jensen Huang said that Nvidia has "largely conceded" China's AI chip market to Huawei and should "expect nothing" regarding chip sale approvals to China (@leeyingshan / CNBC) (Visit Techmeme dot com for the…

Look at the growth in PCB anon I'm pretty sure a Vera CPU costs more than $5k VR200 NVL72 will cost a lot

https://nitter.privacyredirect.com/zephyr_z9/status/2057283585200328869#m

Published: May 21, 2026 02:13

Look at the growth in PCB anon I'm pretty sure a Vera CPU costs more than $5k VR200 NVL72 will cost a lot Aaron (@Aaronwei3n) Sheesh. $NVDA VR200 Bom Analysis from MS. — https://nitter.privacyredirect.com/Aaronwei3n/status/2057279855784546352#m

The bear thesis for DRAM is that memory bois start going crazy and reinvest all their operating profits into capacity expansion in the future (i'm not talking about the fabs that are already planned) Now the problem is that they won't be able to secure enough EUVs to expand capacity at a rapid rate

https://nitter.privacyredirect.com/zephyr_z9/status/2057282694338543905#m

Published: May 21, 2026 02:09

The bear thesis for DRAM is that memory bois start going crazy and reinvest all their operating profits into capacity expansion in the future (i'm not talking about the fabs that are already planned) Now the problem is that they won't be able to secure…

Good luck shorting memory when Vera just added 3.84 Billion GBs of DRAM demand

https://nitter.privacyredirect.com/zephyr_z9/status/2057282237234893060#m

Published: May 21, 2026 02:08

Good luck shorting memory when Vera just added 3.84 Billion GBs of DRAM demand Jeff Pu (@sssjeffpu) FYI, analyst Call Back suggests the number doesn’t include LPDDR5/5X. — https://nitter.privacyredirect.com/sssjeffpu/status/2057275577145696620#m

Gold is around $150,000/kg

https://nitter.privacyredirect.com/zephyr_z9/status/2057274820853919978#m

Published: May 21, 2026 01:38

Gold is around $150,000/kg Jukan (@jukan05) 2. DRAM, excluding modules DRAM, excluding modules: export value of $7,488mn, +431% Y/Y, +26% M/M The DRAM export unit price, excluding modules, reached $82,820/kg, up 13% from April 20, 2026, and up 497%…

I'm pretty sure Anthropic will IPO at $1.5T-$2T as well Their ARR growth is too staggering

https://nitter.privacyredirect.com/zephyr_z9/status/2057263364263534794#m

Published: May 21, 2026 00:53

I'm pretty sure Anthropic will IPO at $1.5T-$2T as well Their ARR growth is too staggering Rittenhouse Research (@RHouseResearch) Not unreasonable to think Anthropic exits 2027 at $200B run-rate with very healthy operating margins. Could easily…

"An annual 80 billion Gb of LPDDR5 would be nearly equivalent to the annual LPDDR5 TAM used for smartphones. The shortage of LPDDR5 — and of DRAM overall — is likely to intensify further over time." The show will go on

https://nitter.privacyredirect.com/zephyr_z9/status/2057257233969463556#m

Published: May 21, 2026 00:28

"An annual 80 billion Gb of LPDDR5 would be nearly equivalent to the annual LPDDR5 TAM used for smartphones. The shortage of LPDDR5 — and of DRAM overall — is likely to intensify further over time." The show will go on Jukan (@jukan05) KIS comment on…

Using bots to read/understand/summarize data faster Very trivial use case

https://nitter.privacyredirect.com/zephyr_z9/status/2057239456126828831#m

Published: May 20, 2026 23:18

Using bots to read/understand/summarize data faster Very trivial use case Helen Toner (@hlntnr) Fascinating info trickling out about LLMs in active military ops in Iran, e.g. Air Force AI tools converting intelligence from top secret -> secret. Means…

As far as I can tell, Colossus 1/2 have major expansion plans xAI said that they want to put over 1M GPUs in those DCs So capacity will be much larger than 1.2GW

https://nitter.privacyredirect.com/zephyr_z9/status/2057234816807153951#m

Published: May 20, 2026 22:59

As far as I can tell, Colossus 1/2 have major expansion plans xAI said that they want to put over 1M GPUs in those DCs So capacity will be much larger than 1.2GW Wall St Engine (@wallstengine) Anthropic agreed to pay SpaceX $1.25B per month through May…

So ARR growth has finally moderated a bit (or will they blow past it, Ant isn't good at forecasting) June ARR will be around $50B if this Q2 revenue number is true

https://nitter.privacyredirect.com/zephyr_z9/status/2057228761071050918#m

Published: May 20, 2026 22:35

So ARR growth has finally moderated a bit (or will they blow past it, Ant isn't good at forecasting) June ARR will be around $50B if this Q2 revenue number is true Negligible Capital (@negligible_cap) *ANTHROPIC'S REVENUE SET TO MORE THAN DOUBLE TO…

ASML will be the limiter for DRAM capacity expansion as well from late 2027 It will shift from cleanroom capacity to EUV tools The memory bois will be fighting among each other and with TSMC and Intel

https://nitter.privacyredirect.com/zephyr_z9/status/2057105403557863789#m

Published: May 20, 2026 14:25

ASML will be the limiter for DRAM capacity expansion as well from late 2027 It will shift from cleanroom capacity to EUV tools The memory bois will be fighting among each other and with TSMC and Intel Jukan (@jukan05) Great insight from @GavinSBaker via…

SaaS firms will probably cut 40%-50% of their workforce within 2 years

https://nitter.privacyredirect.com/zephyr_z9/status/2057098216303456356#m

Published: May 20, 2026 13:56

SaaS firms will probably cut 40%-50% of their workforce within 2 years Joe Weisenthal (@TheStalwart) *INTUIT TO CUT 17% OF JOBS TO STREAMLINE OPS, FOCUS ON AI: RTRS — https://nitter.privacyredirect.com/TheStalwart/status/2057088535941779705#m

The show will go on

https://nitter.privacyredirect.com/zephyr_z9/status/2057092463622590646#m

Published: May 20, 2026 13:33

The show will go on Wall St Engine (@wallstengine) $MU: Our financial outlook has strengthened since our last earnings call. We're on track for another substantial record free cash flow in fiscal Q3. We expect tightness for HBM, Dram, and Nand to…

RT by @zephyr_z9: $SYNA price target raised to $180 from $125 at Rosenblatt, Buy maintained. The firm says Google Research and Synaptics’ new Astra SL2610-based Coralboard platform could accelerate adoption of Synaptics’ Astra processors and be accretive to gross margin and profitability.

https://nitter.privacyredirect.com/wallstengine/status/2057082960554369066#m

Published: May 20, 2026 12:56

$SYNA price target raised to $180 from $125 at Rosenblatt, Buy maintained. The firm says Google Research and Synaptics’ new Astra SL2610-based Coralboard platform could accelerate adoption of Synaptics’ Astra processors and be accretive to gross margin and…

RT by @zephyr_z9: 🚀 We’re hiring! DeepSeek is forming a new Harness team to build Code Harness from the ground up—may be you can call it DeepSeek Code or something like this hhh🤣🤣🤣 📍 Based in Beijing. Two roles open: 🧠 Harness Product Manager → https://app.mokahr.com/social-recruitment/high-flyer/140576#/job/54f386a9-913b-4626-9bf4-e1709b62fcda 👨‍💻 Harness R&D Engineer → https://app.mokahr.com/social-recruitment/high-flyer/140576#/job/ec402ff5-47fe-4e32-820c-b04884fca585 Research meets product—let's build it together. Hit the links and apply directly! 🔥 #DeepSeek #CodeHarness #AI #Hiring #ProductManager #Engineering #Beijing #Referral

https://nitter.privacyredirect.com/victor207755822/status/2057064415300841626#m

Published: May 20, 2026 11:42

🚀 We’re hiring! DeepSeek is forming a new Harness team to build Code Harness from the ground up—may be you can call it DeepSeek Code or something like this hhh🤣🤣🤣 📍 Based in Beijing. Two roles open: 🧠 Harness Product Manager →…

Interconnect bottleneck ;)

https://nitter.privacyredirect.com/zephyr_z9/status/2057054510716162413#m

Published: May 20, 2026 11:03

Interconnect bottleneck ;) Knockouttrader (@knockouttrader) UBS following Goldman and is getting even more confident about Ibiden 4062.JP and the broader AI packaging and substrate cycle as well High-end ABF substrates are heading into a supply…

Super stupid to use this on mobile phone chips

https://nitter.privacyredirect.com/zephyr_z9/status/2056958041367687436#m

Published: May 20, 2026 04:39

Super stupid to use this on mobile phone chips Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex) > Specifically, the design places an N+3 chip on top and an N+2 chip underneath, with the two connected through vertically stacked hybrid bonding almost…

I can assure u that ASP increase will be much much greater than 100% "we forecast Co-Tech’s earnings to grow 10x by 2028 compared to 2025 (a 120% CAGR for 2025–28E), driven by a doubling of capacity and a 1x increase in ASP."

https://nitter.privacyredirect.com/zephyr_z9/status/2056918464129085476#m

Published: May 20, 2026 02:02

I can assure u that ASP increase will be much much greater than 100% "we forecast Co-Tech’s earnings to grow 10x by 2028 compared to 2025 (a 120% CAGR for 2025–28E), driven by a doubling of capacity and a 1x increase in ASP." Aaron (@Aaronwei3n) With…

Interconnect Bottleneck ;)

https://nitter.privacyredirect.com/zephyr_z9/status/2056917999484039565#m

Published: May 20, 2026 02:00

Interconnect Bottleneck ;) Aaron (@Aaronwei3n) With most major AI accelerators starting their production ramp in 2H26, high-end copper foil demand is surging. _ GS: The TAM for high-end copper foil is on an upward trajectory; we expect it to reach…

Great acquisition for ADI

https://nitter.privacyredirect.com/zephyr_z9/status/2056916342755000380#m

Published: May 20, 2026 01:54

Great acquisition for ADI Techmeme (@Techmeme) Analog Devices agrees to buy Empower Semiconductor, which makes chips used to regulate voltage, for $1.5B in cash (@kat_ham_ / Wall Street Journal) (Visit Techmeme dot com for the link and full context!) …

RT by @zephyr_z9: Personally, I think this is a must-win deal for both Intel and Qualcomm. Whoever gets it will save an enormous amount of time. I hope Qualcomm ends up buying it.

https://nitter.privacyredirect.com/jukan05/status/2056905041660031048#m

Published: May 20, 2026 01:09

Personally, I think this is a must-win deal for both Intel and Qualcomm. Whoever gets it will save an enormous amount of time. I hope Qualcomm ends up buying it.

RT by @zephyr_z9: As I’ve explained before, the reason I post some bearish points is to make my own logic more complete. You should never exclude the bear case. You have to incorporate every piece of logic into your thesis.

https://nitter.privacyredirect.com/jukan05/status/2056897130418311168#m

Published: May 20, 2026 00:37

As I’ve explained before, the reason I post some bearish points is to make my own logic more complete. You should never exclude the bear case. You have to incorporate every piece of logic into your thesis.

RT by @zephyr_z9: WPG Holdings, the semiconductor distribution giant, said 2nd quarter revenue, operating profit and net income will all hit fresh record highs, following its strong 1st quarter (Q1), media report, WPG said Q2 revenue will hit NT$345 billion to NT$365B, up 9%-15.3% vs Q1, and operating profit NT$8.66B to NT$9.64B. The firm said capex from downstream manufacturers (Foxconn, Quanta) continues to rise and could provide further reason to raise its financial forecasts. $NVDA $MU $AVGO $MRVL $QCOM #WPG #Foxconn #Quanta #semiconductors https://money.udn.com/money/story/5710/9512884

https://nitter.privacyredirect.com/dnystedt/status/2056891846769729557#m

Published: May 20, 2026 00:16

WPG Holdings, the semiconductor distribution giant, said 2nd quarter revenue, operating profit and net income will all hit fresh record highs, following its strong 1st quarter (Q1), media report, WPG said Q2 revenue will hit NT$345 billion to NT$365B, up…

We flagged this in the January memo

https://nitter.privacyredirect.com/zephyr_z9/status/2056868269618057617#m

Published: May 19, 2026 22:43

We flagged this in the January memo Temple 8 Research (@Temple_Eight) According to TrendForce data from May 7, 2026, the surging demand for AI-related Power Management ICs is rapidly consuming global foundry capacity at the 8-inch mature node level. The…

Wait, what?????? What kind of post training breakthrough did they make?? So the price increase is mostly due to smaller batch size to make it run faster

https://nitter.privacyredirect.com/zephyr_z9/status/2056855057472336178#m

Published: May 19, 2026 21:50

Wait, what?????? What kind of post training breakthrough did they make?? So the price increase is mostly due to smaller batch size to make it run faster Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex) I'm wrong, thanks @yourboiilevi it's G3 Flash…

RT by @zephyr_z9: FYI someone pinged me about Lumentum at JPM conference where "they basically said they sold all their lasers to Nvidia". $LITE makes at least 80% gross margin on lasers sold to $NVDA. My most bearish calculation is 81%. Others pay 87% gross margin or higher.

https://nitter.privacyredirect.com/insane_analyst/status/2056805863760048398#m

Published: May 19, 2026 18:35

FYI someone pinged me about Lumentum at JPM conference where "they basically said they sold all their lasers to Nvidia". $LITE makes at least 80% gross margin on lasers sold to $NVDA. My most bearish calculation is 81%. Others pay 87% gross margin or…

Clearly has very low active parameters but a lot more total parameters

https://nitter.privacyredirect.com/zephyr_z9/status/2056791655060246653#m

Published: May 19, 2026 17:38

Clearly has very low active parameters but a lot more total parameters Google (@Google) Gemini 3.5 Flash is built to help you execute complex, agentic workflows. 3.5 Flash rivals flagship models to deliver frontier performance for agents and coding,…

RT by @zephyr_z9: Top Google customers are processing over 1B tokens per day. If they shift to Gemini 3.5 Flash, they could save $1 billion annually.

https://nitter.privacyredirect.com/wallstengine/status/2056790986446225726#m

Published: May 19, 2026 17:35

Top Google customers are processing over 1B tokens per day. If they shift to Gemini 3.5 Flash, they could save $1 billion annually.

So Google just cooked everyone on cost & speed Harnessing the full power of model-hardware co design Extreme sparsity and Ironwoods

https://nitter.privacyredirect.com/zephyr_z9/status/2056790445343322438#m

Published: May 19, 2026 17:33

So Google just cooked everyone on cost & speed Harnessing the full power of model-hardware co design Extreme sparsity and Ironwoods Lisan al Gaib (@scaling01) Gemini 3.5 Flash comparable with Opus 4.7, GPT-5.5 and Gemini 3.1 Pro on the Artificial…

Bruh Mihoyo plans to invest $13B to build an LLM My gacha money will finally be useful Honestly they should absolutely focus on building smol models for video games and world models for game generation

https://nitter.privacyredirect.com/zephyr_z9/status/2056784984741175473#m

Published: May 19, 2026 17:12

Bruh Mihoyo plans to invest $13B to build an LLM My gacha money will finally be useful Honestly they should absolutely focus on building smol models for video games and world models for game generation Keitaro (@KeitaroTR) miHoYo ha expresado su…

RT by @zephyr_z9: AMD ALERT 🚀 MI355 is now 40% cheaper than B200 on GLM5 architecture for Single Node serving FP8 14 weeks after the initial launch of GLM5 on both non-MTP & MTP with spec decode for SGLang v0.12 for both CUDA & ROCm.  SPEED IS THE MOAT!! Great work to @AnushElangovan, @roaner, HaiShaw & his team! Next step is for MI355X to catch up to CUDA when composing production inference optimizations like FP4 & on distributed inferencing where you can gang up MI355 boxes such that per GPU performance goes up thus the cost per million tokens goes down.

https://nitter.privacyredirect.com/SemiAnalysis_/status/2056782305440452635#m

Published: May 19, 2026 17:01

AMD ALERT 🚀 MI355 is now 40% cheaper than B200 on GLM5 architecture for Single Node serving FP8 14 weeks after the initial launch of GLM5 on both non-MTP & MTP with spec decode for SGLang v0.12 for both CUDA & ROCm.  SPEED IS THE MOAT!! Great work to…

Very interesting

https://nitter.privacyredirect.com/zephyr_z9/status/2056777502836842504#m

Published: May 19, 2026 16:42

Very interesting Wall St Engine (@wallstengine) $SYNA and Google Research will showcase Edge AI use cases on Synaptics’ Coralboard at Google I/O 2026. The board runs on Synaptics Astra SL2610 with Torq NPU and Google Coral NPU tech, supporting on-device…

RT by @zephyr_z9: For a long time, academic researchers being at the cutting edge of new technologies has been a great social equilibrium. Neutral, unbiased technologists have been the people to spread new ideas to the world. As AI research takes off in velocity, it is also going behind closed doors. The tech industry has sewed distrust, and now they are the ones trying to tell the world about incredible changes coming. It's a big loss to a form of social contract in America. There's been a history of scientists helping society understand new technologies. There is a public service in the culture of science that I want to see continue. It's being exacerbated by feelings of FOMO, especially finically driven, where I'm seeing many people who previously wanted to be professors -- and likely still do deep down -- feel a need to conform and chase money, in a pocket of industry. I get it, I grapple with this. For those with a safety net, there will be great returns to some who choose to zag, and try to build something good, for people who need something different. For me, this is building interesting, fully-open models, to show what you can do with a variety of open weight sizes. Yes, AI's immediate future is dictated by the frontier, but it's long-term trajectory still deeply includes academic institutions and open science. Knowledge will always diffuse, but to whom? As of today, I think China is positioned to be the global home of AI research in a few years. The home of research is where ideas are accessible, spread rapdily, and are nurtured. The U.S. seems to be unwinding many institutions and relationships. The largest returns go to people who build something differentiated, at least in reputation, and a lot of people are not being shown that this path exists.

https://nitter.privacyredirect.com/natolambert/status/2056772229120295416#m

Published: May 19, 2026 16:21

For a long time, academic researchers being at the cutting edge of new technologies has been a great social equilibrium. Neutral, unbiased technologists have been the people to spread new ideas to the world. As AI research takes off in velocity, it is…

RT by @zephyr_z9: We are short $MRAM (Everspin Technologies), a niche industrial memory chipmaker whose stock has soared 300%+ as speculative investors pile into anything remotely associated with “memory” and AI. Report at http://kerr.co/mram (1/8)

https://nitter.privacyredirect.com/KerrisdaleCap/status/2056722826883678658#m

Published: May 19, 2026 13:05

We are short $MRAM (Everspin Technologies), a niche industrial memory chipmaker whose stock has soared 300%+ as speculative investors pile into anything remotely associated with “memory” and AI. Report at kerr.co/mram (1/8)

Demis isn't in the game for money

https://nitter.privacyredirect.com/zephyr_z9/status/2056712234777337994#m

Published: May 19, 2026 12:23

Demis isn't in the game for money Seb Johnson (@SebJohnsonUK) He will probably make more money from investing in anthropic than selling DeepMind. Crazy — https://nitter.privacyredirect.com/SebJohnsonUK/status/2056690579258605737#m

LOL Which distributors caught the L here

https://nitter.privacyredirect.com/zephyr_z9/status/2056709227310395440#m

Published: May 19, 2026 12:11

LOL Which distributors caught the L here Jukan (@jukan05) This latest piece from GSR is genuinely important. Why is Nvidia still courting China — almost clinging to it? GSR frames it this way: "For Nvidia, China is a monitoring window — a way to…

5 months late

https://nitter.privacyredirect.com/zephyr_z9/status/2056696165333406081#m

Published: May 19, 2026 11:19

5 months late Knockouttrader (@knockouttrader) Goldman getting increasingly constructive on Furukawa Electric 5801.JP and, more broadly, the AI optical/datacenter buildout in Japan. The firm says the company’s new medium-term plan came in well above…

250k WPM going online from China too next year

https://nitter.privacyredirect.com/zephyr_z9/status/2056590503513374818#m

Published: May 19, 2026 04:19

250k WPM going online from China too next year Jukan (@jukan05) This is a bit scary… Samsung’s P5 could reportedly come online in 2027. — https://nitter.privacyredirect.com/jukan05/status/2056587349379911873#m

Interconnects ;)

https://nitter.privacyredirect.com/zephyr_z9/status/2056576557515493842#m

Published: May 19, 2026 03:23

Interconnects ;) Jukan (@jukan05) “A couple of customers have prepaid for substrates, because the substrate supply chain is very tight – so I need to put up the money to secure this material…(and it shows) the commitment to me – so that’s very…

Google Cloud will supply compute to everyone except GDM LOL

https://nitter.privacyredirect.com/zephyr_z9/status/2056560690031399246#m

Published: May 19, 2026 02:20

Google Cloud will supply compute to everyone except GDM LOL Ray Wang (@rwang07) $GOOGL and $BX to Create $5B New AI Cloud Company Using TPU, per WSJ. Aims to bring 500MW of capacity online in 2027. —…

The interconnect bottleneck anon

https://nitter.privacyredirect.com/zephyr_z9/status/2056536809929416995#m

Published: May 19, 2026 00:45

The interconnect bottleneck anon Jukan (@jukan05) Severe glass fiber shortage drives spillover benefits for Taiwanese glass fiber makers • Ahead of the mass production ramp of NVIDIA’s Rubin platform in 2H26, AI servers and high-speed switches are…

RT by @zephyr_z9: its worth comparing Kimi K2.5 -> Kimi K2.6 and Kimi K2.5 -> Composer 2.5. can't tell on cursorbench but looks like a modest improvement over Kimi's posttraining effort on code, albeit a massive amount more compute.

https://nitter.privacyredirect.com/frontier_foid/status/2056474005155303855#m

Published: May 18, 2026 20:36

its worth comparing Kimi K2.5 -> Kimi K2.6 and Kimi K2.5 -> Composer 2.5. can't tell on cursorbench but looks like a modest improvement over Kimi's posttraining effort on code, albeit a massive amount more compute. Cursor (@cursor_ai) Introducing…

He is too conservative

https://nitter.privacyredirect.com/zephyr_z9/status/2056328878289998024#m

Published: May 18, 2026 10:59

He is too conservative Jukan (@jukan05) Kye-hyun Kyung, Samsung Electronics Senior Advisor: "Memory prices to fall in H2 next year… Korea must cultivate deep-tech manufacturing" Kye-hyun Kyung, Senior Advisor and former head of Samsung Electronics'…

RT by @zephyr_z9: >> TrendForce: Expansion of high-end demand and preemptive inventory build-up in consumer channels could drive an MLCC price rebound • According to TrendForce’s latest research, strong AI chip demand is tightening supply-demand conditions for high-end MLCCs, while supply of consumer-grade MLCCs is also coming under pressure. As a result, some distributors have begun precautionary inventory build-up, and suppliers are responding with price hikes. • Recent price negotiations between ODMs and suppliers also show that the average decline in overall MLCC prices has reached its lowest level in the past three years. TrendForce analyzes this as evidence that the MLCC pricing cycle has entered a critical phase ahead of a potential rebound.

https://nitter.privacyredirect.com/jukan05/status/2056259636827959414#m

Published: May 18, 2026 06:24

>> TrendForce: Expansion of high-end demand and preemptive inventory build-up in consumer channels could drive an MLCC price rebound • According to TrendForce’s latest research, strong AI chip demand is tightening supply-demand conditions for high-end…

"Rubin’s N3 wafer was partly shifted to Vera Rubin 2300W unchanged" Interesting

https://nitter.privacyredirect.com/zephyr_z9/status/2056176211492315204#m

Published: May 18, 2026 00:53

"Rubin’s N3 wafer was partly shifted to Vera Rubin 2300W unchanged" Interesting Jeff Pu (@sssjeffpu) $NVDA Preview Note Abstract • TP: $308 (from $292) * FY27/28E EPS $8.8 / $13.2 • F1Q Rev: $80.6B vs Street $80B • F2Q Guid: $91B vs buy-side $90B •…

We have been bullish on AEHR since the January memo

https://nitter.privacyredirect.com/zephyr_z9/status/2056162640762929593#m

Published: May 17, 2026 23:59

We have been bullish on AEHR since the January memo Irrational Analysis (@insane_analyst) So... @citrini... we are both bullish SiC and $AEHR original market is SiC wafer level burn in... did everyone get distracted by AI ASIC logic burn in and forgot…

What did I tell u anon

https://nitter.privacyredirect.com/zephyr_z9/status/2056147668154683513#m

Published: May 17, 2026 22:59

What did I tell u anon Zephyr (@zephyr_z9) The third semis memo is out We talk about power & analog semis, orchestration plane in the agentic era, the neoclouds trade, interconnect bottleneck (probably the biggest limiter for 2026-27), Korea Unlocked…

CXMT will end this year with the highest net profit generated by a Chinese company Even exceeding ICBC's $50B

https://nitter.privacyredirect.com/zephyr_z9/status/2055952576751997053#m

Published: May 17, 2026 10:04

CXMT will end this year with the highest net profit generated by a Chinese company Even exceeding ICBC's $50B Jukan (@jukan05) Just in: CXMT 1Q26 revenue: $7.46 billion CXMT 1Q26 net profit: $4.85 billion 1Q revenue increased 719.13% YoY. —…

Will be the most explosive Chinese IPO Wish I could get an allocation

https://nitter.privacyredirect.com/zephyr_z9/status/2055951407174578356#m

Published: May 17, 2026 09:59

Will be the most explosive Chinese IPO Wish I could get an allocation Jukan (@jukan05) Just in: CXMT 1Q26 revenue: $7.46 billion CXMT 1Q26 net profit: $4.85 billion 1Q revenue increased 719.13% YoY. —…

And will be super expensive

https://nitter.privacyredirect.com/zephyr_z9/status/2055951033701126584#m

Published: May 17, 2026 09:58

And will be super expensive Jukan (@jukan05) If we are talking about Exynos 2800–2900, that means this is still 2–3 years out. But this is a move that absolutely needs to be watched. Existing LPDDR cannot enable true on-device AI / agentic…

"We are still quite early into what makes a frontier model all the way from optimization, architecture and objectives. Big token wants to convince you otherwise." I like the spirit

https://nitter.privacyredirect.com/zephyr_z9/status/2055862262481694935#m

Published: May 17, 2026 04:05

"We are still quite early into what makes a frontier model all the way from optimization, architecture and objectives. Big token wants to convince you otherwise." I like the spirit rohan anil (@_arohan_) We did research when pay was low. We did…

RT by @zephyr_z9: Last week, Resonac annnounced that they are increasing hard disk media production by 30%, centered on Singapore. Chasing down some supply chain beneficiaries... https://jp.ibtimes.com/resonac-expand-hard-disk-media-output-210-million-disks-100863

https://nitter.privacyredirect.com/illyquid/status/2055860378341638612#m

Published: May 17, 2026 03:58

Last week, Resonac annnounced that they are increasing hard disk media production by 30%, centered on Singapore. Chasing down some supply chain beneficiaries... jp.ibtimes.com/resonac-expan…

RT by @zephyr_z9: You definitely learn a lot from reading expert calls. What I learned today: the optical communications industry is buying up Hoya’s glass capacity at 5x the price HDD companies are willing to pay, making it difficult for HDD makers to expand capacity.

https://nitter.privacyredirect.com/jukan05/status/2055841355394666519#m

Published: May 17, 2026 02:42

You definitely learn a lot from reading expert calls. What I learned today: the optical communications industry is buying up Hoya’s glass capacity at 5x the price HDD companies are willing to pay, making it difficult for HDD makers to expand capacity. …

"We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them" GPT 5.4/5.5 running internally to accelerate R&D and product development for OpenAI is very different from it being commercially viable for external inference at a trillion parameter scale

https://nitter.privacyredirect.com/zephyr_z9/status/2055811678764413234#m

Published: May 17, 2026 00:44

"We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them" GPT 5.4/5.5 running internally to accelerate R&D and product development for OpenAI is very different from it being…

Navitas is still the 2nd best here and has lower Rds and higher power capacity (2x more than the Infineon part) compared to Infineon The gap is not too big

https://nitter.privacyredirect.com/zephyr_z9/status/2055795743034888481#m

Published: May 16, 2026 23:41

Navitas is still the 2nd best here and has lower Rds and higher power capacity (2x more than the Infineon part) compared to Infineon The gap is not too big Temple 8 Research (@Temple_Eight) I instinctively don't own Navitas. Music to my ears —…

RT by @zephyr_z9: OpenAI: MP5 (heavy grind) Claude: Sniper rifle (design, writing) DeepSeek: Shotgun (bulk work, gets it done cheap) Gemini: Mortar (deep research, best away from front line) Kimi K: mine (good async, slow, specialized) Grok: flash bang (imagine is only good product)

https://nitter.privacyredirect.com/goodalexander/status/2055789577055371650#m

Published: May 16, 2026 23:16

OpenAI: MP5 (heavy grind) Claude: Sniper rifle (design, writing) DeepSeek: Shotgun (bulk work, gets it done cheap) Gemini: Mortar (deep research, best away from front line) Kimi K: mine (good async, slow, specialized) Grok: flash bang (imagine is only good…

RT by @zephyr_z9: Infineon's GaN beats the Navitas on Qoss & Eoss the ZVS current for Infineon is 3.28A, while it's 5A for Navitas And for multi level/stacked 800V to 48/50V, the primary current is is quite moderate & high frequency (650 kHz to 1 MHz) and Infineon is mogging everyone here to Navitas has lower conduction loss at higher primary current

https://nitter.privacyredirect.com/zephyr_z9/status/2055776461962023348#m

Published: May 16, 2026 22:24

Infineon's GaN beats the Navitas on Qoss & Eoss the ZVS current for Infineon is 3.28A, while it's 5A for Navitas And for multi level/stacked 800V to 48/50V, the primary current is is quite moderate & high frequency (650 kHz to 1 MHz) and Infineon is…

"they completely missed the boat on the hardware part" Japan is the largest supplier of upstream materials for the AI supply chain The list is too long

https://nitter.privacyredirect.com/zephyr_z9/status/2055662237772312579#m

Published: May 16, 2026 14:50

"they completely missed the boat on the hardware part" Japan is the largest supplier of upstream materials for the AI supply chain The list is too long @levelsio (@levelsio) Japan's biggest problem is they suck at software, have always sucked at…

FiconTec coming to HK LFG!!!

https://nitter.privacyredirect.com/zephyr_z9/status/2055598168415642081#m

Published: May 16, 2026 10:36

FiconTec coming to HK LFG!!! Techmeme (@Techmeme) Shenzhen-listed RoboTechnik, which claims to be the largest silicon photonics tool maker and whose stock is up 340% over the past year, files for a HK listing (@zinnialee23 / Forbes) (Visit Techmeme dot…

RT @sssjeffpu: Per our note on March 31, we believe there is no 4-die version of Rubin Ultra but only 2-die from packaging level. Given the…

https://nitter.privacyredirect.com/zephyr_z9/status/2055579705894711690#m

Published: May 16, 2026 09:22

RT @sssjeffpu: Per our note on March 31, we believe there is no 4-die version of Rubin Ultra but only 2-die from packaging level. Given the…

If Intel wins the 4 die Rubin Ultra packaging orders, then CoPoS is coming soon

https://nitter.privacyredirect.com/zephyr_z9/status/2055454903875862533#m

Published: May 16, 2026 01:06

If Intel wins the 4 die Rubin Ultra packaging orders, then CoPoS is coming soon Macro_Lin ｜ 市场观察员 (@LinQingV) $INTC Intel切入NVIDIA先进封装供应链的时点可能从2028年Feynman提前到2027年Rubin Ultra，CoWoS独家依赖出现松动。 UBS的具体说法是NVIDIA给Rubin Ultra做产品分级，2-die保留在TSMC…

The Interconnect bottleneck

https://nitter.privacyredirect.com/zephyr_z9/status/2055440477126861046#m

Published: May 16, 2026 00:09

The Interconnect bottleneck Sean (@sean_________) Citi on $INTC / $TSM & ABF: "ABF technology readiness is critical for EMIB-T Adoption — Given EMIB-T’s design methodology, the readiness of the ABF substrate ecosystem for ultra-high-density routing,…

"If TPU v9 upgrades the topology, optical module speed, and port ratio at the same time, a roughly 4x increase in ICI bandwidth versus TPU v8 may not be entirely out of reach. This is likely not just a matter of "buying more optical modules." It could reflect a system-level architecture shift in Google's TPU roadmap toward higher bandwidth and higher optical interconnect content."

https://nitter.privacyredirect.com/zephyr_z9/status/2055296965437804709#m

Published: May 15, 2026 14:39

"If TPU v9 upgrades the topology, optical module speed, and port ratio at the same time, a roughly 4x increase in ICI bandwidth versus TPU v8 may not be entirely out of reach. This is likely not just a matter of "buying more optical modules." It could…

Very good

https://nitter.privacyredirect.com/zephyr_z9/status/2055296184269623757#m

Published: May 15, 2026 14:36

Very good Bill Gurley (@bgurley) A new @bgurley blog post! I have been thinking about how sophisticated executives are using open source in super creative ways. Started writing this three years ago. Excited to finish it up and publish it! And with the…

Probably the most publicity the Xperia has received in recent years Good for sales lol

https://nitter.privacyredirect.com/zephyr_z9/status/2055289496179188025#m

Published: May 15, 2026 14:09

Probably the most publicity the Xperia has received in recent years Good for sales lol Sony | Xperia (@sonyxperia) Following the post about AI Camera Assistant, we’d like to explain the feature in more detail. It doesn’t edit photos after shooting - it…

RT by @zephyr_z9: Today’s report from Tech Taiwan: Google reportedly told TSMC directly, “We want to become your direct major customer.” In other words, Google’s ultimate goal appears to be following an Apple-style COT model, bypassing both Broadcom and MediaTek.

https://nitter.privacyredirect.com/jukan05/status/2055274233585111041#m

Published: May 15, 2026 13:08

Today’s report from Tech Taiwan: Google reportedly told TSMC directly, “We want to become your direct major customer.” In other words, Google’s ultimate goal appears to be following an Apple-style COT model, bypassing both Broadcom and MediaTek.

~ 20 additional posts are not shown ~