![\"AMD](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/07\/news-p.v1.20260724.ba9002b62f8f4f28983ebc39fc213b34_P1.png\")
\uc0ac\uc9c4 \ud655\ub300 AMD CEO Lisa Su (left) and Antropics co-founder Tom Brown are presenting together at AMD Advancing 2026 in San Francisco on the 23rd (local time). [Picture = AMD] <\/p>\nAnthropic, Claude HW Transplantation Automation
AMD Introduces Claude Across Development Organizations
OpenAI also launches GPT-grade trials at Helios <\/p>\n
\uc0ac\uc9c4 \ud655\ub300 AMD CEO Lisa Su (left) and Antropics co-founder Tom Brown are presenting together at AMD Advancing 2026 in San Francisco on the 23rd (local time). [Picture = AMD] <\/p>\n
“One engineer connected Claude to the machine and asked him to ‘set this machine up’ and left it for the whole weekend. When I came back on Monday, the performance graph of our latest model kept going up.”<\/p>\n
Tom Brown, co-founder of Anthrax, said on the stage of AMD’s ‘Advanced AI 2026’ keynote speech at the Mosconi Center in San Francisco on the 23rd (local time) that his artificial intelligence (AI) model Claude has virtually completed the initial driving work of AMD’s new AI accelerator alone. Getting a new hardware platform up and running for the first time is usually considered a challenge that large engineering organizations cling to for months.<\/p>\n
“It’s always a huge task to raise a new hardware platform, so I expected it to be a big project again this time,” Brown said. “But the actual experience started and ended with one engineer.” AMD CEO Lisa Su, who was on the stage together, also added, “Our team reported that the anthology engineer was touching MI355, so when I asked him if he needed support, he answered that he didn’t need help.”<\/p>\n
The meeting took place immediately after large-scale cooperation announced by AMD and Anthropic on the 22nd. Anthropic will deploy Instinct MI450 series graphics processing units (GPU) on AMD’s rack-unit system ‘Helios’ with a maximum scale of 2 gigawatts (GW). The first 1GW operation will begin in the first half of 2027. AMD has also agreed to invest up to $5 billion in equity in Anthropic. According to Toms Hardware, the deal will make AMD the sixth semiconductor source to support Claude after Nvidia, Amazon Traynium and Google Tensor Processing Unit (TPU).<\/p>\n
What is noteworthy is the content of the multi-year engineering cooperation between the two companies. Anthropic will put Claude into AMD’s GPU workload optimization and software development, while AMD will introduce Claude throughout its engineering and product development organizations. “Claude’s main specialty is software engineering, but now he’s helping with normal software development and other tasks like design and layout,” Brown said. “We can work together at a point where we make the next generation of chips better.”<\/p>\n
Sachin Katy, general manager of OpenAI Infrastructure, who went on stage, then made the same diagnosis. “One of our key bets is recursive,” he said, “and AI recursively improves the system it will return to, and ultimately AI conducts the research itself.” Although professional engineers are now spending enormous time on kernel optimization and hardware-specific model implantation, it is already proven in the actual service environment that AI automates much of it.<\/p>\n
“OpenAI received Helios Rack three months ago and is already running GPT-class workloads, and large-scale deployments will begin later this year and accelerate throughout 2027,” Carty said. OpenAI signed a contract with AMD to introduce up to 6GW infrastructure last year.<\/p>\n
AMD also released Helios’ performance figures on the same day. On its own metrics, MI455 experienced up to 34 times higher throughput and 18 times more token throughput per dollar than previous generations, averaging 10-15% higher performance and up to 30% more token throughput per dollar than its rivals at the same rack power.<\/p>\n
“What the industry as a whole is now realizing is that this is not a rack-level, but that the entire data center is a system,” Katy said. “Instead of being held back by human speed, we are making AI do it directly.” [Silicon Valley correspondent Wonho-seop]<\/p>\n","protected":false},"excerpt":{"rendered":"Anthropic, Claude HW Transplantation AutomationAMD Introduces Claude Across Development OrganizationsOpenAI also launches GPT-grade trials at Helios \uc0ac\uc9c4 \ud655\ub300…\n","protected":false},"author":2,"featured_media":116777,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[53,3154,182,372,717,43312,136],"class_list":["post-116776","post","type-post","status-publish","format-standard","has-post-thumbnail","category-anthropic","tag-anthropic","tag-anthropic-claude","tag-claude","tag-engineering","tag-hardware","tag-helios","tag-software"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/116776","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=116776"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/116776\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/116777"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=116776"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=116776"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=116776"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}