AMD Nears Samsung HBM4 Supply Deal For New AI Chip Shipping With 432GB Of Memory Per GPU
AMD is closing in on a definitive supply agreement with Samsung Electronics for high-bandwidth memory used in its next-generation artificial intelligence chips, with the new MI455X accelerator packing 432 gigabytes of the hardware per GPU.
The company announced that its Helios rack-scale AI systems have entered full production, with initial shipments slated for the end of the third quarter. AMD also detailed the MI500 accelerator series, which will feature optical interconnects and aims to outpace Nvidia’s Rubin Ultra system in high-bandwidth memory and scale-up capabilities.
From the sources (25 posts)
@tradfi*AMD SAYS AI ACCELERATOR MARKET TO REACH $1.4 TRILLION BY 2030 $AMD
@financialjuiceAMD: The AI accelerator market is expected to reach $1.4 trillion by 2030. $AMD
@stockmktnewz$AMD CEO Lisa Su just said she expects the AI Accelerator market to hit $1.4 Trillion by 2030
@financialjuiceAMD sees the data center CPU market hit $220 billion by 2030. $AMD
@wallstengine$AMD SEES DATA CENTER CPU MARKET HIT $220B BY 2030
@financialjuiceAMD sees the overall compute market reaching $2 trillion by 2030. $AMD
@stockmktnewz$AMD SEES THE OVERALL COMPUTE MARKET REACHING $2 TRILLION BY 2030
@yahoofinanceAMD CEO Lisa Su says the AI accelerator market will hit $1.4 trillion by 2030. "By the end of the decade, the AI accelerator market is going to approach the size of the entire semiconductor market today," Su said, adding that GPUs will ma
@financialjuiceAMD announces Helios AI server computer rack. $AMD
@stockmktnewzAMD CEO Lisa Su just said she expects the AI Accelerator market to hit $1.4 Trillion by 2030 "By the end of the decade, the AI accelerator market is going to approach the size of the entire semiconductor market today"
@wallstengine$AMD CEO Lisa Su expects the AI accelerator market to reach $1.4 trillion by 2030, within an overall compute market projected to reach $2 trillion. Monthly AI token consumption has already increased 158x over the past two years. https://t.
@financialjuiceAMD CEO calls the MI450 AI Accelerator the fastest in the industry. $AMD
@financialjuiceAMD CEO: Helios in full production; shipments will start soon. $AMD
@wallstengine$AMD unveiled Helios, its new rack-scale AI system powered by the MI450 accelerator. CEO Lisa Su called the MI450 the fastest AI accelerator in the industry. Helios is now in full production, with shipments set to begin at the end of Q3 an
@stocksavvyshay$AMD CEO Lisa Su expects the AI accelerator market to reach $1.4T by 2030 nearly matching the size of today’s entire semiconductor industry. That growth is already showing up in usage with monthly AI token consumption up 158x over the past
@polymarketmoneyBREAKING: AMD CEO Lisa Su expects the AI accelerator market to reach $1,400,000,000,000.00 by 2030.
@yahoofinanceAMD is launching its Helios AI rack, calling it the "most powerful rackscale AI infrastructure." CEO Lisa Su says its in production already, and shipments will begin soon.
@yahoofinance$AMD CEO Lisa Su says that in 2026, for the first time, the world is using more AI compute to run models than to train them.
@stocksavvyshay$AMD unveils Helios, its MI450-powered AI system which CEO Lisa Su calls “the most powerful rack-scale AI infrastructure.” Helios is already in full production with shipments beginning at the end of Q3 and ramping through Q4.
@stockmktnewz$AMD CEO Lisa Su just said “In 2026 for the first time the world is using more AI compute to run models than to train them”
@financialjuiceOpenAI expects to deploy AMD Helios at a massive scale. $AMD
@financialjuiceOpenAI Infrastructure Chief At AMD Event: OpenAI got its hands on the AMD Helios GPU racks three months ago. $AMD
@financialjuiceOpenAI Infrastructure Chief: OpenAI plans to work with AMD on MI500 series AI chips and beyond. $AMD
@wallstengineOPENAI GOT ACCESS TO $AMD’S HELIOS GPU RACKS THREE MONTHS AGO, VP OF COMPUTE STRATEGY SAYS
@stocksavvyshay$AMD CEO Lisa Su says 2026 marks the first year the world uses more AI compute “to run models” than to train them. She expects inference to consume ~60% of global AI capacity as agentic AI scales.