AMD Exemplary EU reveals Openai CEO Sam Altman

Advanced micro devices New information about the next generation AI chips on Thursday, new information about the instinct Mi400 series will be presented next year.

MI400 chips will be able to collect thousands of plugs to a full server rack called AMD, which allows the AMD to be connected in a way that allows you to connect as a “RAF scale” system.

“For the first time, we accept each part of the shelf as a single system,” said the CEO of AMD, San Jose in San Jose, San Jose.

Open CEO Sam Altman, came to the stage with water and said the company will use the AMD plugs.

“When I started talking about the features for the first time, I was not like that, just sounds completely crazy,” Altiman said. “It will be something incredible.”

The AMD’s rack-scale setup will appear as a system that develops a user as a system like a system, such as a system, a system that develops large language models and large language models. These customers want to cover all data centers and ask the AI ​​computers “Hiperscale” groups that may be able to apply a massive amount of power.

“Think of Helios as a function like a massive computing engine,” he said, “he said,” he said Nvidia Vera Rubin racks expected to be broadcast next year.

The AMD’s rack-scale technology allows the latest chips to compete with the Blackwell plugs, built with 72 graphical processing units. NVIDIA is a large and only opponent of GPUs in the GPUs in the AMD’s Great Information Center to develop and place AI applications.

Chip said that Openai – prominent NVIDIA client – AMD has commented on AMD on MI400 Roadmap. With the Mi400 chips and this year’s Mi355X chips, a company company, a company company, company company, company company, company company power consumption, and place NVIDIA with the “aggressive” prices of the AMD.

So far, the NVIDIA is dominated by the Central GPU’s GPU’s market, because it is the first company to use the required program for AI developers, graphics for 3D games. In the past decade, before the AI ​​boom, he payed attention to AMD reciprocal Intel In the server CPU.

The water said that the AMD’s Mi355X could produce NVIDIA’s Blackwell chips using NVIDIA’s “Ownership” program.

“He says we always know, but this shows that the open program frameworks make great progress,” he said.

AMD shares are true in 2025 so far, and the Wall Street still signals that he has not seen as a great threat to the dominance of Nvidia.

AMD’s General Manger, AMD’s General Manger, AMD’s General Manger, said that the AMD’s AI plugs will be completed to work and obtain the AU plugs on Wednesday.

“Throughout the board, then our performance has a competitive advantage, so we have a meaningful purchase value that we shared the two-digit interest deposits,” Dieckmann said.

Over the next few years, large cloud companies and countries are preparing to spend hundreds of billion dollars to build new data center clusters around GPU to accelerate the development of advanced AI models. This includes $ 300 billion This year, scheduled capital expenditures from Megacap technology companies.

AMD, for the AI ​​chips to exceed $ 500 billion in the total market in 2028, although the market can claim how much this market can claim and currently has more than 90% of the market Analyst calculations.

Both companies stressed how important it is for companies where both companies are on an annual basis on an annual basis, and the more important AI chip technology Microsoft, Orace and Amazon.

AMD purchased or invested in 25 AI companies in the past year, including Purchase of ZT systems earlier this yearA server manufacturer who developed technology was needed to set up the rack-sized systems.

“These AI systems are super complicated and fully assembled solutions are really critical,” he said.

AMD said that the largest 10 largest AI customers of the instinct chips, including Openai, including Openai TeslaXai and CO.

Oracle plans to offer customers to clusters with 131,000 mi355x plugs.

Officials Meta On Thursday, the AMD’s CPUS and GPU will draw results for the LLA model and plan to buy the next generation servers of AMD.

Microsoft’s representative said that Copilot EUs from the AMD chips to serve the features of the EU.

AMD refused to say how much the chips have been completed – it does not sell themselves and end users generally take them through a hardware company Dell or Super micro computer – However, the company plans to fight the Mi400 chips.

Santa Clara combines GPUs with Pensando’s 2022 Pensando’s purchase and network plugs to build Helios racks. This should also benefit the rest of the AI ​​chips to the rest of the AMD. NVIDIA owner uses open source network technology to closely integrate Ualink’s RAF systems against NVLLLLINK.

AMD claims that Mi355x can give 40% more token – AI output – A size of AI is a dollar for a dollar for using a size than the opponent of the plugs.

The data center GPUP can cost tens of thousands of dollars for a chip and cloud companies usually receive them in large quantities.

AM AI chip business is still smaller than Nvidia. This was financially 2024, the AI ​​sales will be $ 5 billion sale, but JP Morgan analysts expect a 60% increase in the category this year.

Follow: AMD CEO Lisa Water: Chip Export Control is a hood, but we still see the opportunity to grow