sarath_svn's profile picture. Semiconductors | Physical AI

Sarath

@sarath_svn

Semiconductors | Physical AI

$ORCL $AMD “Oracle just committed to 50K $AMD MI450 GPUs for OCI starting in 2026.”

Oracle just committed to 50K $AMD MI450 GPUs for OCI starting in 2026. Add this to AMD’s new multi-year 6 GW OpenAI deal, and it’s clear AMD is on a run. $ORCL gets supply diversity; AMD gets validation from both a hyperscaler and a top AI lab. Execution is key. Is AWS…

PatrickMoorhead's tweet image. Oracle just committed to 50K $AMD MI450 GPUs for OCI starting in 2026.

Add this to AMD’s new multi-year 6 GW OpenAI deal, and it’s clear AMD is on a run. 

$ORCL gets supply diversity; AMD gets validation from both a hyperscaler and a top AI lab.

Execution is key. 

Is AWS…


$AVGO $AMD “Design the whole system, all of the stuff about the chip, the way we design these racks, the networking between them.”

The $AVGO partnership isn't just a custom chip, but entire rack-scale cluster for reasoning inference, per the Broadcom + OpenAI podcast this morning: "Andrew Mayne: So this is going to entail both compute and chip design and scaling out? Sam Altman: This is a full system...…



$INTC “Intel 18A is the first 2 nanometer node on US soil” Also, the first 2 nanometer class node to support Backside power delivery

Intel unveils details on its first product built on 18A - Panther Lake, as well as previewing data center processor Clearwater Forest, and a first look inside Fab 52. Here are 3 things to know about the #IntelTechTour 2025. intel.ly/4nT5I5y



“Bigger your fleet of robots, cheaper they become to produce & more data u can collect, which means faster ur robots can improve, creating a natural flywheel effect where industry’s early leader can begin to distance its rivals. 1st mover gets a cheaper & smarter robot over time”


Sarath 已轉發

This is how Amazon delivery bot will ring your doorbell. It is done by leveraging the same loco-manipulation framework that enables humanoid robots to move boxes, developed by Amazon's Frontier AI & Robotics team.

I've long wondered if we can make a humanoid robot do a 𝘄𝗮𝗹𝗹𝗳𝗹𝗶𝗽 - and we just made it happen by leveraging 𝗢𝗺𝗻𝗶𝗥𝗲𝘁𝗮𝗿𝗴𝗲𝘁 with BeyondMimic tracking! This came after our original OmniRetarget experiments, with only minor tweaks to RL training: relaxing a…



$INTC “Intel 18A-P is getting mobile-optimized standard cell libraries.”

Important: Intel 18A-P Mobile Support 18A has been in the news all along, and a *very important* 18A-P news just got buried in the flow. ▶️ LBT said they aren't getting "anymore" new customers for 18A. So, 18A PDK might not be under active development anymore. ▶️ Intel said…

Silicon_Fly's tweet image. Important: Intel 18A-P Mobile Support

18A has been in the news all along, and a *very important* 18A-P news just got buried in the flow.

▶️ LBT said they aren't getting "anymore" new customers for 18A. So, 18A PDK might not be under active development anymore.
▶️ Intel said…


$INTC - Margin Stacking

$INTC As 18A gets to HVM, derivative effects expected are A) intel profit margins will go up B) IFS will breakeven and go beyond C) Increased confidence in 18A process to drive external customers signing up Intel Foundry capacity D) investment on 14A, even more advanced process



$INTC “Intel 18A is the most advanced semiconductor node developed & manufactured in USA We are proud to announce that Fab 52 in Arizona is now fully operational & set to reach high-volume production of our next generation client product Panther Lake later this year. “

Delighted to visit Intel Arizona this week! Intel 18A is the most advanced semiconductor node developed and manufactured in the United States and we are proud to announce that Fab 52 in Arizona is now fully operational and set to reach high-volume production of our next…

LipBuTan1's tweet image. Delighted to visit Intel Arizona this week! Intel 18A is the most advanced semiconductor node developed and manufactured in the United States and we are proud to announce that Fab 52 in Arizona is now fully operational and set to reach high-volume production of our next…
LipBuTan1's tweet image. Delighted to visit Intel Arizona this week! Intel 18A is the most advanced semiconductor node developed and manufactured in the United States and we are proud to announce that Fab 52 in Arizona is now fully operational and set to reach high-volume production of our next…


$INTC $TSM “18A HD density as 186.38 MTr/mm2, but enables +10% *HIGHER* cell utilization due to BSPDN (PowerVia). This increases the *real density* of a 18A die.”

Intel 18A vs TSMC N2 Density: Why PowerVia (BSPDN) totally rocks!!! Many media outlets constantly ignore "Cell Utilization". Probably due to lack of understanding. 18A HD density as 186.38 MTr/mm2, but enables +10% *HIGHER* cell utilization due to BSPDN (PowerVia). This…

Silicon_Fly's tweet image. Intel 18A vs TSMC N2 Density:

Why PowerVia (BSPDN) totally rocks!!!

Many media outlets constantly ignore "Cell Utilization". Probably due to lack of understanding.

18A HD density as 186.38 MTr/mm2, but enables +10% *HIGHER* cell utilization due to BSPDN (PowerVia). This…


Sarath 已轉發

🏭 From Intel Arizona's new Fab 52 comes the future: Panther Lake PC processors with 50%+ faster performance, Clearwater Forest server processors with 288 E-cores, all on Intel 18A - America's most advanced semiconductor node. Gaming, AI, data centers, robotics - read more…

intelnews's tweet image. 🏭 From Intel Arizona's new Fab 52 comes the future:  
Panther Lake PC processors with 50%+ faster performance, Clearwater Forest server processors with 288 E-cores, all on Intel 18A - America's most advanced semiconductor node.  
Gaming, AI, data centers, robotics - read more…
intelnews's tweet image. 🏭 From Intel Arizona's new Fab 52 comes the future:  
Panther Lake PC processors with 50%+ faster performance, Clearwater Forest server processors with 288 E-cores, all on Intel 18A - America's most advanced semiconductor node.  
Gaming, AI, data centers, robotics - read more…

Sarath 已轉發

Panther Lake is the company’s first product built on Intel 18A and will begin ramping high-volume production this year, with the first SKU slated to ship before the end of the year and broad market availability starting January 2026. intc.com/news-events/pr…


$INTC Intel 18A based Panther Lake Intel18A = GAA + BSPDN Fab 52 in Arizona is set to reach high-volume 18A production by the end of this year.

JUST IN: Intel $INTC announces first chip built with 18A (1.8nm) tech, Panther Lake. Says Arizona Fab 52 is fully operational and set to reach high-volume 18A production later this year.

LuxAlgo's tweet image. JUST IN: Intel $INTC announces first chip built with 18A (1.8nm) tech, Panther Lake.

Says Arizona Fab 52 is fully operational and set to reach high-volume 18A production later this year.


Chinese Humanoid Robots : Hardware Stack ( $INTC, $NVDA)

AI data center chips get all the attention. But the next battlefield is edge compute, which is crucial for autonomous systems. Many Chinese robots rely on Intel CPUs and Nvidia GPUs, including Unitree. But Chinese chipmakers have also been developing their own alternatives.…

kyleichan's tweet image. AI data center chips get all the attention. But the next battlefield is edge compute, which is crucial for autonomous systems.

Many Chinese robots rely on Intel CPUs and Nvidia GPUs, including Unitree. But Chinese chipmakers have also been developing their own alternatives.…


$AMD $NVDA $AVGO

Get your money out of AMD while you still can. Many of you have piled into the stock after it already gapped up 30%. You are in a world of pain once everyone sobers up. Empty promises, don’t make money. Here are the cold HARD facts. 1) AMD has to hit $600 for OpenAI to get…



$NVDA

All roads lead to $NVDA

danielnewmanUV's tweet image. All roads lead to $NVDA


$TSM $NVDA $AMD

TSMC N3P $/xtor is similar to N2 from what I understand, although N2 has a slight perf/W advantage. AMD and Nvidia will chose the one that makes more sense to them. AMD will use N2 since they don't need gigantic dies like Nvidia. N3P can fab large dies & can support more volume.



$NVDA

The United States has approved NVIDIA’s sale of billions of dollars worth of chips to the United Arab Emirates.



$NVDA $AVGO New Nvidia orders for 800G Modules : 60% to China’s Innolight & Eoptolink. 40% to US companies like Coherent, Lumentum, Broadcom.

> About 60% of the new Nvidia orders for 800G modules are going to China’s Innolight and Eoptolink, and about 40% to US companies like Coherent, Lumentum, Broadcom Can't even buy the Chinese players



$TSM $INTC

There is no spreadsheet, barring total AI meltdown/letdown, that says we have enough foundry capacity through 2030. If you have it, please share. If Intel can execute on external (non-Intel chips) for 18A and 18AP, it will add some anchor wafer tenants. Expect TSMC to make some…



$AMD $INTC

After yesterdays news, you can add Sam Altman to the list of people who want Intel to work @benthompson:

buccocapital's tweet image. After yesterdays news, you can add Sam Altman to the list of people who want Intel to work

@benthompson:


United States 趨勢

Loading...

Something went wrong.


Something went wrong.