InfraGroq 3 LPX enters production, built together with NVIDIA's platformA rack-level inference accelerator, co-designed with Vera Rubin NVL72. It combines GPUs with dedicated fast-response modules and handles token generation in agentic systems. Already deploying at Nebius, CoreWeave and SpaceXAI.InfraAWS Glue 6.0 cuts the price by 30 percent and adds full Iceberg v3 supportAWS shipped version 6.0 of Glue, its data-processing service, at 30 percent lower cost than previous versions. Underneath is a new engine, and on top, full support for the Iceberg table format v3.InfraGitHub explained the 17 August outage: capacity did not keep up with its own growthOn 20 August GitHub published a post-mortem of the 17 August failure, which held the platform down for seven hours and 47 minutes and took out github.com, authentication, Actions, APIs and Copilot. Per the company the cause was not a change in code or configuration but capacity that did not arrive in time.InfraCerebras launched CS-4 and promised up to thirty times faster than GPU systemsThe fourth generation of their system stands on three processors the size of a whole silicon wafer. The company claims up to 30 times faster inference. The number is theirs; there is no independent check yet.InfraNVIDIA started locking up land, power and buildings the way it used to lock up chipsJensen Huang announced a partnership with SB Energy for a site in Ohio that will host NVIDIA compute. The tenant is OpenAI. The company calls land, power and shell the next strategic resource.InfraNVIDIA wants AI factories to move to 800 volts direct currentThe bottleneck is not how much power there is but how many times you convert it on the way to the chip. NVIDIA, Google and Microsoft are writing a common standard through OCP; more than 80 manufacturers are already building to it.InfraNVIDIA is selling compute as infrastructure, not as machinesThe goal is over $500 billion of outside capital flowing into AI factories the way it flows into roads and power plants. NVIDIA backstops up to 25 per cent of residual value per project.InfraNVIDIA brought a 30-billion model down to the home RTX cardNemotron 3.5 Lightning is an open mixture-of-experts model built to run locally: on an RTX computer, not in someone else's cloud. Alongside it ships NeMo Switchyard, an open router between models.
42 stories