AI founders, CTOs & infra leaders in Chicago for #SC26:
RSVP to our full stack event today (3-5:30PM):
Join @MassedCompute (GPUs), @EverpureData (Data), @bits_io (Security & observability), @procurri (Hardware lifecycle),ย @spectrocloudinc (Infra mgmt)
๐
https://t.co/sZAfWtV4C9
OpenAI says GPT-6 Cyber found 81 high-severity Firefox flaws and 948 Linux kernel exploits, and Mozilla is already shipping fixes. Build your own AI security tools at https://t.co/Oih6uJ89ux
NVIDIA built a team of AI sub-agents to trace slowdowns across compute, networking, hardware, and scheduling on its DGX SuperPOD clusters, all on-premises. Run your own GPU workloads at https://t.co/Oih6uJ89ux
Massed Compute GPUs are powering open science at #SC26.
Visit the San Diego Supercomputer Center (SDSC) booth #4645 to see how weโre supporting the National Research Platform (NRP) cluster.
Massed Compute is expanding to the @AmericanTowerUS Edge Data Center in Raleigh, NC, putting our GPUs at the edge of the Research Triangle ๐
Read the full story at: https://t.co/Q5rsNIiT4s
Massed Compute is at #SC26 in Chicago. If you're training or scaling AI models, weโd love to meet!
JOIN US TONIGHT from 5-7 PM CT.
@digitalrealty + @DellTech are hosting a networking reception with whiskey tasting & light bites.
Space is limited. RSVP: https://t.co/vNy7rC7kno
Arete 27B just hit 754.1 output tokens per second at 32 concurrent requests on a single H100 with vLLM. The A100 is the value pick at 444.4 tokens per second per dollar. The full numbers are live and public. Launch your own GPU on Massed Compute today. https://t.co/xnhK6cnImT
SC26 kicks off in Chicago this weekend, and we're headed there!
On Wed. Nov. 18 (3-5:30 PM), get the entire #AIinfrastructure stack in one room at our "Chicago Full Stack SuperCompute" Happy Hour. Drinks, bites, and scaling AI.
Limited spots. RSVP now: https://t.co/sZAfWtV4C9
Between political cycles and market swings, conversations around AI have taken a wild ride in the last few weeks. That makes flexibility so valuable for AI builders and leaders.
Get NVIDIA GPUs without multi-year lock-in with Massed Compute: https://t.co/ewXByuZOec
CUDA 13.4 adds enforced memory limits per workload, so multiple jobs can share one GPU safely. A new open-source GPU Fractioning Operator brings it to Kubernetes. Rent GPUs at https://t.co/Oih6uJ89ux
NVIDIA showed smarter Tensor Core precision modes speed up atom-level simulation by up to 2.5x on Blackwell while keeping FP32-level accuracy. Run your chemistry and materials workloads at https://t.co/Oih6uJ89ux
Three models, three jobs, three picks. Clef labels on an A100 SXM4 at $1.38 an hour. JEV routes tickets fastest on an RTX PRO 6000 Blackwell. Kolibri serves 1,885 tokens per second on the same Blackwell.
https://t.co/VYxO8MiF1M
Food service has the highest quit rate of any US industry. Yum Brands says Taco Bell's drive-thru voice AI is making jobs easier and helping keep team members longer. Build AI that helps people at https://t.co/Oih6uJ89ux
LFM2.5-24B-A2B has three right answers. L40S for tokens per dollar at $0.97 an hour, A100 SXM4 for raw throughput, RTX PRO 6000 Blackwell for a 101.9 ms first token.
https://t.co/uDX7x87BXC
Georgia Tech's Tech AI helps companies and public agencies take AI from idea to deployment, pairing them with university researchers and infrastructure. Build your next AI project at https://t.co/Oih6uJ89ux
IBM Research's Granite Timeseries Ensemble combines four open forecasting models into one tool, with no retraining needed. Permissively licensed on Hugging Face. Run your forecasting workloads at https://t.co/Oih6uJ89ux
NVIDIA's Sentry runs AI agent security on BlueField-4 DPUs, outside the agent's reach, so a rogue agent can't switch off its own monitoring. Planned for early 2027. Run your agentic workloads at https://t.co/Oih6uJ89ux
NVIDIA showed a frontier AI model writing the code that turns robot CAD files into simulation-ready assets for Isaac Sim. The AI handles physics, joints, and collisions. Run your simulation workloads at https://t.co/Oih6uJ89ux
NVIDIA and Jane Street built Mamba-2 kernels in cuTile, so developers can write GPU kernels in Python while the compiler handles Tensor Cores. Run your state-space models at https://t.co/Oih6uJ89ux