AMD Courts 13 Korean AI Firms and Plans a Research Center in Korea by Year-End
Lisa Su pitched pairing AMD processors with Korean neural processing units for inference systems, and named MangoBoost as AMD's first local cloud partner.
TechnologyAICybersecurityThe Systems Shaping Business
Section
Page 2 of 3: earlier stories.
Lisa Su pitched pairing AMD processors with Korean neural processing units for inference systems, and named MangoBoost as AMD's first local cloud partner.
Samsung has floated HBM4 asking prices about three times what current HBM3E chips fetch, according to The Korea Herald.
A 20-year contract pays for 890 megawatts of new output from existing plants, and a separate 15-year deal adds 2,700 megawatts on the PJM grid.
The Wyoming deal runs through 2048 and pairs 590 megawatts of grid service with a privately managed microgrid of about 2.1 gigawatts of third-party resources.
Validating resolvers that still trust only KSK-2017 could fail to resolve names under every top-level domain once KSK-2024 takes over.
The North Dakota-focused AI data center builder expects first power in 2028 and is already pitching the undisclosed Finnish site to hyperscalers.
Visiting Taipei and Hsinchu, AMD's CEO said demand for its processors is still outrunning supply and that partners are now planning three to five years out.
The 380-megawatt unit shipped from Changwon, Korea, and is one of a dozen turbines the Korean manufacturer has contracted to supply to U.S. customers.
Qualcomm blames a leaked license-termination threat for a $170 million hit to a Meta chip deal; Arm says Meta changed course on its own.
The agreement with Virginia-based AIR Control Concepts centers on air-cooled chillers that lean on outdoor air, and could extend to liquid-cooling hardware.
Third-quarter sales of NT$3.03 trillion beat analyst forecasts, and the iPhone and AI server maker expects AI-related growth to continue into the year-end peak.
The debt-free rescue will cost BT about £400 million this fiscal year, and the culture secretary has ordered a fast public-interest review.
Partners including Synopsys, Cadence, Infosys and TOPPAN will help companies without in-house chip teams design custom parts for the Japanese foundry.
A Culpium report says TSMC could build and run a fab serving only Tesla, SpaceX and xAI, a move that would raise questions about Intel's 14A role.
AWS chief Matt Garman paired the Built Together program with a rebuttal of common complaints about water, power prices and diesel generators.
The special purpose vehicle would raise debt and sell a 10% stake to outside investors, moving Grace Blackwell hardware off Amazon's balance sheet while it keeps using the chips.
We Energies says the Point Beach deal would cut fuel costs for other customers, but regulators must still sign off and a rate increase is on the table.
OpenAI's hardware chief says the company chose AMD's x86 server CPU to de-risk its first in-house inference rack, despite close ties to Nvidia and Arm.
Counterpoint data put Micron at 24% of second-quarter DRAM revenue to SK hynix’s 25%, while Samsung held the top spot and won back ground in HBM.
The Planet-built prototype will run inference in short bursts on about a kilowatt, while a Joule paper sets out the launch prices orbital data centers would require.
Each Texas fab could cost at least $20 billion, but the plan hinges on Congress renewing a 35% chipmaking tax credit that lapses at year-end.
The reported five-year, $7 billion lease would be Tencent's largest overseas compute deal and relies on U.S. rules that block chip sales to China but not rentals abroad.
The 690-megawatt contract gives Constellation the revenue certainty to seek another license extension for Maryland’s only nuclear plant and to add capacity by 2032.
Japan's largest power generator would feed the site from its own LNG supply chain, with Apollo lined up as financing partner and operations targeted for around 2028.