Jun

[Jun @Junjunjun120511]

$RAM

Moomooさん早くしてくれー😩

#RAM

[Roundhill Investments @roundhill]

Attention traders ready to level up. Meet $RAM, the Roundhill T-REX 2X Long DRAM Daily Target ETF. Trade $DRAM, now with 2X the exposure.

Learn more here: https://t.co/4SUVZHADM1

$RAM $DRAM @REXShares

[Jun @Junjunjun120511]

@kurokawaakito48 確かにそーですよね😅
私はスペースエックス2ブルの損切り分の税金余力があるので、発売開始されたら少量買うつもりです🧐

[Jun @Junjunjun120511]

ポストも動画もかっ��いー😍

[BKLZ〜雰囲気投資家の矜持〜 @Biz_zatukora]

6/23🇺🇸-297078ドル

NVDA -4.1

MU -13.1
SNDK -13.6
STX -5.0
DRAM -14.2
NBIS -2.9
AEHR -9.6
LITE -7.3
IONQ -0.81
SPCX +0.98

擦り傷。ダメージゼロ。

[Jun @Junjunjun120511]

@micron_girl うわー！
めっちゃ気になりますよね‼️

[Jun @Junjunjun120511]

寄り天だけはやめてください🙏

[Jun @Junjunjun120511]

@paurooteri 変態DRAM🤣🤣🤣

[Jun @Junjunjun120511]

@micron_girl こちらこそです👍
過去の経緯からもMU決算でサプライズ分割発表しそうな気がしますが、どー思いますか？？？

[Jun @Junjunjun120511]

@TradexWhisperer いつも有益な情報提供ありがとうございます‼️
ワールドカップファイナルは🇯🇵×🇺🇸？🤣

[Jun @Junjunjun120511]

$DRAM
に追い風🤣

[Trade Whisperer @TradexWhisperer]

$DRAM OpenAI and Broadcom Unveil LLM-Optimized Intelligence Processor

Key Available Data on HBM for OpenAI's Titan Chip

Samsung Deal: Up to 800 million gigabits of 12-layer HBM4 memory supplied exclusively for the first-generation Titan accelerator. This represents roughly 7% of Samsung’s projected full-year HBM output and supports initial deployments.

Memory Configuration: 12-layer HBM4 stacks provide higher capacity and bandwidth than previous HBM3E generations. Reports and patents indicate aggressive designs, potentially supporting a high number of stacks per chip (up to 20 in some referenced architectures) using advanced packaging to maximize memory for inference workloads.

Scale Context for 10 GW: 10 gigawatts represents massive infrastructure. Each modern AI accelerator pairs with hundreds of GB of HBM for efficient large-model inference. Across the full rollout (H2 2026 through 2029), total HBM demand likely reaches tens to hundreds of thousands of stacks, with the Samsung allocation covering an initial phase. Additional supply from Samsung, SK Hynix, and others will support later generations.

Partnership: OpenAI and Broadcom have a multi-year strategic collaboration (ongoing for about 18 months as of the announcement) to co-develop custom AI accelerators and networking systems optimized for large language models (LLMs), particularly for inference workloads.

Scope: Plans to deploy 10 gigawatts of OpenAI-designed AI accelerators. OpenAI designs the chips and systems; Broadcom handles development, manufacturing aspects, and provides Ethernet networking solutions for scale-up and scale-out in AI clusters.

Timeline: Chips expected to ship starting in late 2026 (with initial deployments targeted for H2 2026). A second generation is planned on more advanced process nodes.

Focus: The processor is tailored for efficient LLM inference, addressing surging demand for AI compute. It builds on Broadcom's expertise in custom ASICs, networking (Ethernet), and high-performance silicon. This helps reduce reliance on third-party GPUs (e.g., Nvidia) for OpenAI's workloads...

[Jun @Junjunjun120511]

@Biz_zatukora ノリポチッはヤバそうですね🤣

[Jun @Junjunjun120511]

@ice_ice_coffeee 最後のポエット‥
笑いました🤭

[Jun @Junjunjun120511]

@ichiokutameo1 超絶神決算からの爆上げだと思います！
おまけにサプライズもありそうな‥😏

[Jun @Junjunjun120511]

@Bvr3ZG0LGo19720 なるほど🧐
そういう見方もありますね🫡

[Jun @Junjunjun120511]

@Bvr3ZG0LGo19720 サプライズ分割発表されたら保有枚数ヤバいことになりそうですね！

[Jun @Junjunjun120511]

@mamimumemo_kabu 間違いなく出るでしょう🧐

[Jun @Junjunjun120511]

@Nnyanchusan 逆を言うとその暴落が買い時ですね！

[Jun @Junjunjun120511]

$MU

明日の決算がめっちゃ楽しみになるマイクロンCEOのコメント

#mu
#マイクロン

https://t.co/cCL7AiZ8tb