$AMD $MU $SKHY $DRAM AMD's MI450 is reportedly set to feature 432GB of HBM4 memory per GPU, around 150GB more than NVIDIA's Vera Rubin GPU, which is expected to carry 288GB." Memory Advantage "While the system may trail NVIDIA's VR200 NVL72 in raw compute power, it is expected to hold an advantage in HBM4 memory capacity, a key factor for memory-intensive AI workloads such as large language model inference." Full Rack Specs "The system will feature AMD's 6th-generation EPYC Venice CPUs with up to 256 cores, paired with 72 Instinct MI455X AI accelerators. It will pack 31TB of HBM4 memory and deliver 1,400TB/s of memory bandwidth." "AMD… estimates Helios will deliver around 2,900 FP4 dense PFLOPS of compute performance." Adoption "AMD said that eight of the top 10 AI companies are already running workloads on its Instinct GPUs, including OpenAI, Cohere, and xAI… AMD expects tens of billions of dollars in AI data center revenue starting in 2027, driven largely by Helios deployments." "AMD's Helios AI Rack Challenges NVIDIA with HBM4 Memory Edge; Microsoft Joins as Latest Customer" Pricing "Helios could cost $5 million–$5.5 million per rack, compared with $3.5 million–$4 million for NVIDIA's next-generation Vera Rubin platform."

$AMD AI GPU revenue is modeled to jump ~3x in 2027 as MI450 and Helios ramp with units nearly doubling and ASPs up ~40% eventually reaching ~$160B by 2030. What makes that even more interesting is AMD rack capacity reaching ~3 GW in 2027 and ~9 GW by 2030 which lines up with GPU revenue per GW rising to $18B. And once you add server CPUs you can see that the total data center revenue gets closer to ~$220B by 2030.

$AMD CEO Lisa Su is heading to Korea to secure HBM4 supply from $SKHY and Samsung as MI450 ramps with 432GB of memory per GPU. That makes HBM supply a big piece of AMD’s 2027 ramp while Samsung’s push for a major AI customer gives Su more leverage on allocation.

NEW: AMD $AMD will invest up to $5 billion in Anthropic and supply up to 2 gigawatts of its Instinct MI450 AI chips, with shipments set to begin in the first half of 2027, per WSJ.

AMD $AMD could secure more than $20B in spending from Anthropic $ANT under its initial AI computing arrangement, according to Anthropic's updated IPO filing. The agreement involves AMD's Helios rack-scale systems and MI450-series GPUs, with initial capacity planned for 2027 and the possibility of additional orders. I think this is worth watching because AMD is increasingly getting opportunities to compete for large AI infrastructure deployments that have traditionally favored NVIDIA $NVDA. However, the outstanding giant in this is, and will always be, Jensen Huang / NVIDIA.

AMD, TSMC and OSAT Channel Check: 1. AMD's AI GPU production (MI455/MI450) is expected to remain 100% at TSMC CoWoS in 2027. Production estimates land at a mid-point of 600k units for both SKUs next year per recent MS forecasts. 2. While ASE and Amkor still lead OSAT capacity, their capacity ramps are rumoured to be progressing slower than expected. This means that TSMC may be required to support CoWoS-L wafers for AMD's Venice CPU prodction which would further constrain internal capacity for MI455/MI450 production. Per MS, this would equate to TSMC supporting ~80k CoWoS-L for Venice production. $AMD $TSM $AMKR $ASX

561 commits. 242 contributors. 64 new. Kimi K3 stable, plus early support for next-gen NVIDIA Rubin and AMD MI450 hardware. Here's what's new in @vllm_project v0.27 from @mgoin_ in <4 minutes:

$AMD reports Q2 after the close today. The focus has moved past PC swings to what's structural: server CPU share gains, and execution on the Helios and MI450 rack-scale AI platforms in H2. With cloud penetration and agentic-AI demand, server CPU sales look likely to beat Helios is the one to watch — 432GB HBM4 per GPU, 23.3TB/s bandwidth, ~34x the prior generation's peak compute. It ships late Q3 and ramps in Q4, with Meta, OpenAI, Anthropic and Microsoft already lined up The real question after the bell: how much of those deployments converts into recognized revenue guidance — and whether new large-scale customers show up. UBS sees 2026 revenue reaching $51.1B → NFA. For informational purposes only

Weekly AI Update AMD–ANTHROPIC AI DEAL $AMD and Anthropic have reportedly agreed to an AI infrastructure deal worth tens of billions of dollars. Anthropic plans to acquire up to 2 gigawatts of AMD's next-generation Instinct MI450 chips starting in 1H27, while AMD could invest up to $5 billion in Anthropic based on deployment milestones, strengthening AMD's challenge to $NVDA. NVIDIA EXPANDS JAPAN AI ECOSYSTEM $NVDA unveiled its Cosmos 3 Edge AI model for robots and vision AI agents while expanding its physical AI ecosystem in Japan through partnerships with Fujitsu, Hitachi, Kawasaki Heavy Industries, and healthcare firms. The initiative strengthens Nvidia’s presence in industrial automation, robotics, and AI-driven drug discovery across Japan. ANTHROPIC BOOSTS AI SAFETY FUNDING Anthropic pledged an additional $20 million to Public First Action, bringing its total support to $40 million for AI safety education and policy initiatives. The company continues advocating for stronger AI safeguards, including mandatory safety testing, tighter chip export controls, and measures to reduce national security risks from advanced AI.

𝗔𝗠𝗗 𝗤𝟯 𝟮𝟬𝟮𝟲 𝗘𝗮𝗿𝗻𝗶𝗻𝗴𝘀: 𝗥𝗲𝘃𝗲𝗻𝘂𝗲 𝗝𝘂𝗺𝗽𝘀 𝟰𝟱% 𝗮𝘀 𝗗𝗮𝘁𝗮 𝗖𝗲𝗻𝘁𝗲𝗿 𝗡𝗲𝗮𝗿𝗹𝘆 𝗗𝗼𝘂𝗯𝗹𝗲𝘀 𝗼𝗻 𝗔𝗜 𝗗𝗲𝗺𝗮𝗻𝗱 @AMD reported Q3 2026 revenue of $13.4 billion, up 45% YoY and above the high end of guidance, driven by a record quarter for Data Center. However, a 40% YoY drop in Gaming and rising memory costs weighing on PC demand point to growing pressure on the client side. 𝗞𝗲𝘆 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀: • AMD is now a data center company: Data Center revenue grew 97% YoY to a record $8.5 billion, now 64% of total revenue, with operating margin rising to 34% from 25%. • Helios turns AMD into a system supplier: With MI450 launched and the first Helios racks live, AMD now offers the first credible rack-scale alternative to NVIDIA. AI revenue is expected to overtake server CPU revenue in Q4. • AMD set to out-earn Intel: Q4 guidance of ~$17 billion (up ~66% YoY) tops the $16.2 billion consensus and Intel’s ~$15.5 billion midpoint. AMD expects 80% revenue growth in 2027, even as PC units may fall by double digits. 𝗦𝘂𝗯𝘀𝗰𝗿𝗶𝗯𝗶𝗻𝗴 𝗰𝗹𝗶𝗲𝗻𝘁𝘀 𝗰𝗮𝗻 𝗮𝗰𝗰𝗲𝘀𝘀 𝗼𝘂𝗿 𝗚𝗹𝗼𝗯𝗮𝗹 𝗗𝗮𝘁𝗮 𝗖𝗲𝗻𝘁𝗲𝗿 𝘅𝟴𝟲 𝗖𝗣𝗨 𝗔𝗻𝗮𝗹𝘆𝘀𝗶𝘀 𝗮𝗻𝗱 𝗜𝗻𝘀𝗶𝗴𝗵𝘁𝘀 𝗿𝗲𝗽𝗼𝗿𝘁 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗹𝗮𝘁𝗲𝘀𝘁 𝗼𝗻 𝘁𝗵𝗲 𝗔𝗠𝗗–𝗜𝗻𝘁𝗲𝗹 𝘀𝗲𝗿𝘃𝗲𝗿 𝗿𝗮𝗰𝗲 𝗮𝗻𝗱 𝘁𝗵𝗲 𝘁𝗿𝗲𝗻𝗱𝘀 𝘀𝗵𝗮𝗽𝗶𝗻𝗴 𝗱𝗮𝘁𝗮 𝗰𝗲𝗻𝘁𝗲𝗿 𝗰𝗼𝗺𝗽𝘂𝘁𝗲: #AMD# #Semiconductors# #DataCenter# #AI# #AIInfrastructure# #Helios# #EPYC# $AMD