Lux, an AMD-built AI supercomputer, opens to scientists at Oak Ridge
Half of its annual capacity is reserved for the Genesis Mission, and a second machine in the same DOE and AMD arrangement is due in 2028.
Graphics processing unit: a chip built to do many simple calculations at once. Designed for graphics, it turned out to suit AI, and it is now the main hardware for training and running models.
Nvidia dominates the market, which is why its supply and prices shape the whole AI industry.
Half of its annual capacity is reserved for the Genesis Mission, and a second machine in the same DOE and AMD arrangement is due in 2028.
Shipping is expected in 2028, and key specifications such as memory and FP64 allocation remain undisclosed.
If the claim holds, the cost of moving AI workloads off Nvidia hardware could fall sharply.
Operators in Malaysia and five other countries may soon need export licences for AI chips, with existing facilities given 90 days to comply.
Isolation depends on the mode chosen: isolated mode still trusts the parent instance, while confidential mode needs attestation checks to hold.
Owners of older Maxwell, Pascal and Volta cards keep only security updates until October 2028, while Turing and newer cards stay on regular driver updates.
The gap between Nvidia's private assurances and its trade body's public objections could shape how the Chip Security Act is written.
The flaw cannot be patched on cards already sold, so the effect on the used market will depend on how widely the technique spreads.
If the claims hold beyond AMD's own tests, memory-constrained servers could need far less installed DRAM, though the rollout timing is still unclear.
Federal agencies have three days to patch both products, which were added to CISA's Known Exploited Vulnerabilities catalog.
Users who have not updated Chrome 157 remain exposed to an attack Google has not yet described in detail.
Organisations running SonicOS should patch quickly, since the flaw may already be in use against systems that rely on SAML-based multi-factor authentication.
The expanded collaboration pairs national AI infrastructure with inference software, agent security tools and faster semiconductor simulation.
The rewrite targets a leaner client, but Discord has not yet set a public release date.
If NVIDIA's own and partner-reported figures hold up, agent workloads could run on fewer CPUs per unit of power, though none of the numbers have been independently benchmarked.
At 2 to 3 Hz against a 30 Hz target, npu-driver is a research prototype, but early figures suggest it could free the GPU for other work.
Because rules judge only the processor, memory worth about $4,000 a unit in China is moving through Hong Kong and Shenzhen, with enforcement expected to tighten.
Its Dubhe design aims to fit more AI racks into the same floor space by cutting conversion stages and copper cabling.
The features are still in preview and Microsoft's results will vary between games, with no release date set.
Anyone who loaded the affected models with trust_remote_code enabled should treat those machines as compromised and rotate any reachable credentials.