zksyncGm

Build ZKSYNC

[zksyncGm @zksyncgm]

When you see 200+ t/s on a pro6000 cluster you realize that it’s the only metric that matters.

It’s magic.

[Ahmad @TheAhmadOsman]

Local AI hardware = capacity × bandwidth × software stack

- Capacity tells you what fits
- Bandwidth tells you how hard the box can breathe
- The software stack tells you how much of the spec sheet you can actually cash out.

Hardware by Memory Bandwidth
- Mac Studio M3 Ultra: up to 512GB @ 819 GB/s
- RTX PRO 6000 Blackwell: 96GB @ 1792 GB/s
- RTX 5090: 32GB @ 1792 GB/s
- RTX 4090: 24GB @ 1008 GB/s
- RX 7900 XTX: 24GB @ 960 GB/s
- Radeon PRO W7900: 48GB @ 864 GB/s
- AMD Radeon AI PRO R9700: 32GB @ 640 GB/s
- Intel Arc Pro B65: 32GB @ ~608 GB/s
- Tenstorrent Wormhole n300: 24GB @ 576 GB/s
- Tenstorrent Blackhole p150: 32GB @ 512 GB/s + 800G
- MacBook Pro M5 Max: 460-614 GB/s
- MacBook Pro M5 Pro: 307 GB/s
- DGX Spark: 128GB @ 273 GB/s (coherent + CUDA)
- Mac mini M4 Pro: 273 GB/s
- Ryzen AI Max / Strix Halo: ~256 GB/s (~96GB usable GPU)
- MacBook Air M5: 153 GB/s
- Snapdragon X2 Elite: 152-228 GB/s
- Intel Lunar Lake: 136 GB/s
- Snapdragon X Elite: 135 GB/s
- Mac mini M4: 120 GB/s
- Arc Pro B60: 24GB @ ~456 GB/s

Verdict

- GPUs are still the bandwidth kings

- Apple wins: stupid amounts of memory, don’t want to shard across GPUs
- Apple loses: when raw tokens/sec & concurrency matter more

- DGX Spark: coherent memory + NVIDIA stack

- Strix Halo / Ryzen AI Max: first real x86 unified-memory contender

- Tenstorrent: fully OSS stack, excited to see this mature

Fitting ≠ serving

Even if it fits, you still pay for
- bandwidth during decode
- KV cache growth
- dequantization
- batching + concurrency
- scheduler quality
- framework overhead

The only mental model that matters:

1. What must fit?
2. What bandwidth tier do I need?
3. What software stack can actually deliver it?

In short:
- NVIDIA → fastest raw speed
- Apple Studio M3 Ultra → biggest one-box memory
- Strix Halo → first real x86 unified
- DGX Spark → coherent NVIDIA dev appliance
- AMD / Intel Arc → rising alternatives
- Tenstorrent → fully opensource stack

Do ask: “which bottleneck am I buying?”

Not: “which hardware is best?”

[zksyncGm @zksyncgm]

They've won so much they're printing money every day.

[Serenity @aleabitoreddit]

Wow, 3 limit ups in a row with WUS TW. Pretty sad I didn't take larger positions.

WUS TW is 1 my 2 NAV arbitrage + independent growth trades.

If you want context: WUS TW is a ~$1.12B PCB player with AI DC growth.

Their stake in WUS Kunshan is $4.79B (11.4% of 42.04 billion)

So basically they're a $1.1B PCB player growing independently, sitting on ~$4.79B worth of another company that looks to IPOed on Hong Kong markets soon.

And a successful activist investor in Palliser (with Ajinomoto and Toto), is pressuring them to trim some holdings along with a few others.

Seems like a pretty insane disconnect, and I think there's a lot of potential here.

[zksyncGm @zksyncgm]

@GrailCo Solid numbers, keep going!

I have applied to join Glist and am awaiting approval.

[zksyncGm @zksyncgm]

@GrailCo up up up

[zksyncGm @zksyncgm]

Without the ability to withstand risk, investment is unsuitable. What market only goes up? If you want investments that only go up and never down, I suggest you go to sleep; you'll find everything you want in your dreams.

[10xHustler @multiBagger2026]

@aleabitoreddit Why are so many basing you over $XFAB and your other positions ? Could see so much negativity? Even after you picking solid plays so far? Did $XFAB pull of some bad shit?

[zksyncGm @zksyncgm]

Agreed. Without the ability to withstand risk, investment is unsuitable. What market only goes up? If you want investments that only go up and never down, I suggest you go to sleep; you'll find everything you want in your dreams.

[Serenity @aleabitoreddit]

I think u must be new here.

Many of my ideas get intense backlash at the start, especially the more original they are.

$AXTI - endless hate to the point I got banned from the $RDDT WSB forum.

They thought it was some scam Chinese company, but Reuters, Epiwafer company earnings, and institutions validated their InP substrate position many months later.

$RPI - everyone called it some meme stock.

Literally Bloomberg, Financial Times, and others went out and called it a meme stock with no fundamentals.

Analysts went and said the idea was stupid and said it was going to crash “as a fact”.

Earnings came out? Blew away any projection with 58% fwd revenue growth.

But they seem to have forgot all the backlash they threw out at me, while they’re citing it as a high growth ai hardware company no.

$SIVE? Everyone called it a “meme stock”. I get bunch of hate from Swedish media all the way up.

Bunch of people didn’t understand the technical nuances, and they keep throwing personal attacks for some strange reason.

But as you know it’s probably my most successful idea and it’s validated so far by institutional buying from Fidelity Research, JP Morgan as well as formally announced partnerships from $JBL to $GFS.

Same hate with:

- $AAOI at $30, when everyone called management a “scam” or “shady”

- $LITE at $300 when everyone called photonics a “bubble”

- $RKLB at $20 when everyone thought it was a low revenue launch company that was a bubble. I actually got temp banned from WSB from posting about Rocketlab since moderators didn’t like the stock.

- $HOOD at $20 when everyone recalled them freezing GME buys/sells

- $IQE at $12 when people thought it was just some random crap $100m company over in the UK with “no actual photonics partnerships”

- $SOI at $44 when European bank analysts thought it was “overvalued”and my thesis wasn’t anything new

- $NBIS at $75 when everyone in the $IREN camp said I was spreading Russian propaganda and that they had no moat

- $INTC at $115 when everyone thought they couldn’t compete with $TSM

- $MRVL at $85 when everyone thought they were losing ASIC share to Broadcom.

- $AEHR at $35 when everyone misread their earnings and thought they had no revenue

- $EWY at $115 when everyone was crying KOSPI was a bubble and LNG/helium/oil would disrupt the memory trade

Can go on and on��

I do read a lot of the comments, which is why I remember a lot of the hate (probably either jealously, impression farming, or lacking the technical depth is my guess).

But I think at this point, people can just see each thesis validated over and over again.

And the success in markets drown out the old noise.

Good thing is markets are the final arbiter of what’s right or wrong, not the angry comments or posts on X.

[zksyncGm @zksyncgm]

Love that there are different pack tiers for different collectors

[Grail.xyz @GrailCo]

STARTER PACKS: $15 for 1 digital card backed by real world grails. Open to all

EXPANSION PACKS: $90 for 3 digital cards backed by real world grails. GLIST only

LAUNCH PACKS: Grail digital cards backed by real world grails to commemorate new gtoken launches. GLIST only

[zksyncGm @zksyncgm]

@GrailCo If you are real, GLIST will find you. At least I hope so

[zksyncGm @zksyncgm]

Whether AI CapEx stays strong through 2026 and 2027

How supply chain localization and geopolitical issues affect optical components

The competitive landscape, especially if new players or traditional optical module giants accelerate their entry

[zksyncGm @zksyncgm]

I personally agree that we're still in the very early
stage.

That said, whether the big revenue ramp in 2027 can actually happen as expected will depend on a few key factors:

[Serenity @aleabitoreddit]

非常感谢中文社区对我最看好股票的投资思路进行的深度剖析！

总的来说，��� $AAOI 和 $SIVE 这样的激光公司是我个人的最爱，因为它们在拓展营收方面有着极大的想象空间。它们可以不仅仅局限于销售激光器，还能制造完整的光模块 以及各种光引擎或 ELS (外部光源) 组件。

又或者像 $COHR 那样，甚至可以向上游延伸进军衬底制造领域，从而实现垂直整合，并不断提升自身的毛利率和总潜在市场规模 TAM

这其中很大一部分的关键在于要率先精准踩中投资主线，毕竟去年很多人甚至都不觉得“光子学（photonics）”���成为一个真正的热门概念。

而且我认为最会让大多数人感到惊讶的是，我们目前仍处于极其早期的阶段。绝大多数的营收放量（爬坡）要到明年，也就是 2027 年的上半年乃至下半年才会真正开始！

不过，能看到中文社区大家持续不断的支持，真的让我感觉非常棒。我会尽我所能，不辜负大家给我起的 “白毛股神” 这个称号。

[Serenity @aleabitoreddit]

The 1st large scale X study on stock market net worth has finished!

From the 6000+ responses of those active on X finance, here's the results:

5.8% - $10M+
15% - $1m - $10M
35.9% - $100k - $1M
43.3% - under $100k

This is probably more representative of active followers.

Where 20.8 out of every 100 people have over $1,000,000+.

And 5.8 out of 100 have over $10,000,000.

And 43.3% of the population are probably happier in life than the rest.

Was just curious, fascinating figures.

Definitely not representative of the total population, but just those active on Finance X.

[Serenity @aleabitoreddit]

$ALRIB general meeting notes came out today.

-2nd ROSIE System expected to be delivered shortly to a leading quantum computing player” in the US.

- They’re also intensifying BD for their products used for BTO/STO on photonics and recorded strong interest.

Just for background, they’re the MBE duopoly with Veeco, which got re-rated heavily recently (Disclosure own Riber), with exposure to Quantum.

Positive development overall.

[Steven Cheng @stevencheng]

Building a robot hand that is strong, precise, reliable, and mass-producible is also one of the hardest challenges facing Optimus.✋

[zksyncGm @zksyncgm]

We remain bullish on the robotics sector

[zksyncGm @zksyncgm]

You can now apply for reimbursement from Binance.

[土澳大狮兄BroLeon | Crypto | AI | Stocks @BroLeon]

为了找到罪魁祸首，我用上了花1500刀接入的Arkham api。

让AI帮我去分析，果然效率比人工高了太多。

AI告诉我昨晚这波 aster-2:native 暴跌的砸盘主力就是 @wmt_ventures 冬季，因为他有很多钱包，人工比较容易看漏。

Wintermute → 全部 CEX 合计 ≈ $9.89M(即约 $1M 去了 OKX/Bybit 等其他所,$8.8M 全砸币安)。

而币安 24h ASTER 净流入 +$10.9M —— Wintermute 一家就占了币安净流入的80%,它就是这波往交易所灌货的主力。

果然不请自来的都没什么好心眼🤣
@Leonard_Aster @Kathy_Aster @Aster_DEX

[zksyncGm @zksyncgm]

I've been holding long positions in Nvidia for a long time with 6x leverage; I need a quick breakout and a price surge soon.

[NVIDIA AI Infrastructure @NVIDIAAIInfra]

📣: MiniMax M3 has landed, joining models like DeepSeek V4 and Kimi-K2.6 at the frontier of open agentic models — and NVIDIA Blackwell is already delivering leading performance on it.

NVIDIA Blackwell Ultra delivers up to 5x higher AI factory throughput than NVIDIA Hopper on MiniMax M3, while achieving over 300 TPS/user for highly responsive applications.

⚡ And this is just the start — performance is expected to improve further as NVIDIA deploys the extreme co-design stack with optimized kernels, NVFP4, disaggregated inference through NVIDIA Dynamo, and beyond.

[zksyncGm @zksyncgm]

@GrailCo gTOKEN

[zksyncGm @zksyncgm]

@GrailCo 😮

[zksyncGm @zksyncgm]

The US and China will be the two most powerful countries in the AI ​​era, with South Korea likely to rank third.

[zksyncGm @zksyncgm]

Great test article, I learned a lot.

[Przemek Chojecki | PC @prz_chojecki]

Kimi 2.7 ranked 2nd after Fable 5 and before GPT-5 xhigh

We have re-run our ErdosBench smoke test on 14 problems with Kimi 2.7, Qwen 3.7 Max, Grok 4.3 and compared it with the top performers from previous runs.

Kimi 2.7 is amazingly good. More below.