SpaceX has just given Foxconn a major order for NVIDIA GB300 AI server racks, significantly upping its training- and inference-related compute footprint. Foxconn has now added SpaceX to its growing roster of hyperscaler deals According to a report by a Taiwan-based publication, SpaceX has just awarded Foxconn a part in a new $52 billion order for around 13,000 racks of NVIDIA GB300 AI servers, with shipments expected to begin in Q4 2026 and continuing until Q1 2027, with each rack costing around $4 million. Also, the order spans 936,000 NVIDIA GPUs. Do note that it is not yet clear if […]

Read full article at https://wccftech.com/spacex-awards-foxconn-a-huge-52-billion-order-for-13000-racks-of-nvidia-gb300-ai-servers-where-each-rack-costs-4-million/

Jim Keller isn't bothered by Cerebras's recent IPO and says that he welcomes it, but Tenstorrent will still beat them on everything. Tenstorrent CEO, Jim Keller, Signals Deal With Intel or Qualcomm While Promising To Beat Cerebras "on everything" Tenstorrent recently introduced its latest BlackHole Galaxy server, a system with which it can disrupt the entire AI segment, with performance levels that crush the competition. We covered the announcement last month when the company demoed its Blackhole server undercutting a NVIDIA GB300 with up to five times better TCO. Keller Accepts The Challenge To Beat NVIDIA, Cerebras & Others At […]

Read full article at https://wccftech.com/jim-keller-cerebras-ipo-was-helpful-tenstorrent-to-beat-them-on-everything/

CPUs or GPUs, but require lots of memory for running Agentic AI, and this demand is spiraling to unseen levels as DRAM constraints persist. CPUs Running Agentic AI Will Be Equipped With Up to 400 GB of Memory, Further Crushing The DRAM Supply Chain Memory makers are earning big profits but are also unable to meet the demand. We have seen reports on how major manufacturers are rapidly expanding their production facilities, but these are yet to become operational, and Samsung itself has stated that 2027 will be worse for the DRAM industry than 2026, so it's looking like a […]

Read full article at https://wccftech.com/agentic-ai-pushes-cpus-to-pack-400-gb-of-memory-4x-more-than-today/

NVIDIA has become one of the first to submit the 'extensive' MLPerf Inference v6.0 benchmarks, delivering the highest performance relative to "all competitors" combined. NVIDIA's Blackwell Ultra, Combined With Extreme Co-Design Laws, Manages to Dominate With MLPerf v6.0 Benchmarks When it comes to benchmark submissions and showcasing the 'prowess' of its computing platforms, NVIDIA has been at the forefront, particularly with MLPerf, where the firm is one of the few entities to complete a rigorous round of benchmarks. This time, according to the company's latest blog post, NVIDIA has discussed its latest submission to MLPerf v6.0, noting that, with Blackwell […]

Read full article at https://wccftech.com/nvidia-is-among-the-first-to-submit-mlperf-inference-v6-0-benchmarks/

NVIDIA's GB300 NVL72 AI racks have been tested across DeepSeek's latest open source models, and through fine-tuning and optimized inference, the results are indeed promising. NVIDIA's Blackwell Ultra Scores Up to a 1.5x Lead Over GB200 NVL72 In Latency-Sensitive Workloads With GB300, NVIDIA's primary focus has been on delivering optimal long-context performance in order to capitalize on the agentic AI wave. In a recent post, we discussed how Blackwell Ultra delivers a 50x increase in throughput per megawatt compared to Hopper GPUs through its extreme co-design approach. Now, the Large Model Systems Organization (LMSYS) has tested GB300 NVL72 for long-context […]

Read full article at https://wccftech.com/heres-how-nvidia-blackwell-ultra-is-dominating-long-context-deepseek-workloads/