Nvidia insists the AI gravy train won’t derail any time soon
Shrugs off bubble talk while bragging about Rubin
Nvidia is trying to calm the cocaine nose jobs of Wall Street again, this time with its chief bean-counter insisting that fears of an AI bubble are nonsense.
Tachyum boasts of a 6GHz with 1,024 cores
Outfit promises cosmic performance from a chip that exists only on paper.
Tachyum has rolled out fresh specs for its upcoming Prodigy Universal Processor and the sheet reads like someone let a marketing intern loose with a calculator.
Nvidia shoves Rubin into production with HBM4 in hand
Vera Rubin superchip sprints from keynote glare to TSMC’s lines
Nvidia is racing its Rubin GPUs into production, turning the Vera Rubin superchip from stage prop into silicon while the ink on the slides is still drying.
Samsung gears up for 2nm and HBM4 mass production
Korean giant profits surge
Samsung is charging ahead into 2026 with plans to begin mass production of next-generation memory and chip technologies, including its long-awaited 2nm Gate-All-Around (GAA) process.
Samsung talks up HBM4E
Look at my huge bandwidth and sharp pricing
Samsung is waving fresh HBM4 and HBM4E roadmaps and inking chunky memory deals with Nvidia and AMD.
AMD bets big on 2nm for next-gen Instinct MI450 GPUs
Lines up TSMC’s finest silicon to take a swing at Nvidia’s AI throne
AMD is preparing to go all in on advanced chipmaking with its next-generation Instinct MI450-series accelerators, built on TSMC’s brand-new N2 process.
Nvidia uses radical cooling trick for Rubin Ultra AI chips
Ditches traditional liquid cooling
Nvidia is preparing a major rethink of how it cools its next-generation Rubin Ultra AI chips, reportedly moving to a new microchannel cover plate system to stop its silicon beasts from overheating.
Nvidia and AMD crank up AI chip war
Both sides rejigging their offerings
Nvidia and AMD are frantically revising their next-gen AI designs in an effort to out-muscle each other, with power budgets and memory bandwidth figures spiralling upwards.
Nvidia kicks off “cold revolution 3.0”
Suppliers told to cough up pricier heat dissipation gear
Nvidia is demanding its suppliers develop new microchannel water cooling plate (MLCP) technology, with unit prices running three to five times higher than existing solutions.
Nvidia's Rubin CPX GPU targets AI inferencing arms race
Rack-sized monster to crush token counts
Nvidia has decided that building GPUs the size of small towns is the way to stay ahead in AI, and its latest weapon is the Rubin CPX which is aimed squarely at inferencing and large context models.