CoreWeave

The NVIDIA HGX B300 is now generally available on CoreWeave, and the performance is a game-changer for agentic AI.️ Reasoning loops need memory and speed. We’re delivering both: 🔹 4.93x faster latency (DeepSeek-R1) 🔹 3.42x faster tokens (Kimi K2.5) 🔹 270GB HBM3e memory 🔹 800Gbps networking See the results below ⬇️

#GTC26 was one for the books! ⚡️ From F1 cars to developer nights, the energy at CoreWeave House was unmatched. We came for the AI innovation and stayed for the community (and the bucket hats!) Thanks to everyone who made this week legendary. See you next year! @NVIDIAGTC

700+ experiments in 8 hours ⚡️ When @karpathy’s AutoResearch agent needed to break the one GPU bottleneck, @skypilot_org leveraged a @CoreWeave cluster to get it done. The result? Parallelism made the experiments 9x faster. This is how research scales in 2026. Full technical breakdown below ⬇️

We co-hosted an exclusive look at the future of #PhysicalAI and breakthrough partnerships with @AstonMartinF1 at CoreWeave House during #GTC26. From the track to the cloud, elite performance engineering is the ultimate differentiator. Check it out below! #CoreWeave #AMF1 #NVIDIAGTC

Some highlights from the floor of #GTC26, with some great booth demos from the #CoreWeaveCrew 👇

From custom kicks to the ultimate co-sign from Jensen Huang himself, the CoreWeave booth was the place to be. For everyone that came by and got some swag from the booth, CoreWeave Blue looks good on you 💙

Our AI Wall with @decart_ai is turning #NVIDIAGTC attendees into astronauts, action heroes, and much more! Low latency, high frame rates, and pure AI magic – powered by CoreWeave Cloud and Lucy 2.0 The future of interactive AI is here.️ #AI #CoreWeave #Decart #GenAI

What happens when The Essential Cloud for AI meets the peak of Formula One engineering? At #NVIDIAGTC, CoreWeave House revealed all the answers. In partnership with the @AstonMartinF1 Team, CoreWeave brought together engineers, builders, and racing fans for a closer look at how the AI.R platform is reshaping the way the cars are designed, tested, and pushed to their limits. From Richard Ahlfeld and Justin Hodges doing a deep dive on the AI.R Tunnel’s role in accelerating AMR26 development to conversations with Shergul Arshad on what’s ahead for the 2026 season, one thing was clear: the future of racing is being built in the cloud. But it wasn’t all fireside chats, English breakfasts, and Aston Matcha bars. Guests stepped into the driver’s seat of the F1 simulator, connected with the teams behind the technology, and got their hands on exclusive AMF1 x CoreWeave gear—limited drops inspired by the speed, precision, and performance we’re building together. This is the intersection of game-changing AI and world-class engineering at full throttle. #GTC26 #CoreWeave #AMF1

See you there! #NVIDIAGTC

The "Inference Inflection" isn't a prediction – it's happening right now. ️ Great to have Jensen Huang stop by the @CoreWeave booth to check out the NVIDIA HGX B300 systems we just made generally available on CoreWeave Cloud. We’ve been working with @NVIDIA since the very beginning, and it’s incredible to see that shared history culminate at Booth 913 this year 🤝 Check out some of the highlights from the floor below!

Day 1 at #NVIDIAGTC was one for the books. From the opening keynote to the closing toast, the vibes were at an all-time high. Check out the highlights below and see what we've got in store for the rest of the week here: coreweave.com/nvidia-gtc-2026

Agentic AI is growing exponentially, but general-purpose clouds are holding it back like an F1 car on cobblestones. NVIDIA HGX B300 is now GA on CoreWeave to eliminate those bottlenecks. Why Pioneers trust CoreWeave for HGX B300: ✅ Frictionless Scale: Non-blocking NVIDIA Quantum-X800 InfiniBand networking. ✅ Sustained Peak: Every server is liquid-cooled for optimal GPU temps. ✅ Predictable Speed: Move from run to results faster with Mission Control. ✅ Expert Backed: AI-native infrastructure optimized with @NVIDIA. Unlock the performance your agentic breakthroughs deserve. hubs.la/Q047f8t20

All smiles for Day 2 of #NVIDIAGTC 😄

The W&B mobile app is finally here!! Now you can check on those weekend training runs without having to open your laptop 🫡 Great to see @wandb making the ML workflow even more seamless with the new iOS app!

Following successful validation on @NVIDIA Grace Blackwell, we’re deploying @PhaidraAI agents across our liquid-cooled fleet. Alongside Phaidra, NVIDIA, and @APLDdigital, we’re setting a new standard for reliability, SLAs, and efficiency in the age of AI. The data: 📉 80% reduction in thermal spikes ⚡️ <10s response time ⚖️ IT/OT convergence at GW scale We’re turning the data center into a single, self-learning unit of compute. Full breakdown here: phaidra.ai/blog/breakthro…

Big news for developers! @Cline will integrate @CoreWeave’s W&B Inference to power the next generation of autonomous coding agents. This integration provides: ✅ Low-latency scaling ✅ Access to state-of-the-art models ✅ Built-in observability with W&B Weave Catch the demo at GTC this week at Booth 913! Full announcement here: hubs.la/Q0475lKj0 @wandb

Staff Engineer Zach Merendino is on the floor at @Switch as we deploy the ultimate infrastructure for autonomous agents and reasoning models. The Performance Breakdown: 🔥 Faster Token Generation 3.42 times faster for reasoning models like Kimi 2.5. 🧠 270GB HBM3e: 50% more memory per GPU vs. NVIDIA HGX B200. 🔌 2x Bandwidth: Powered by non-blocking NVIDIA Quantum-X800 InfiniBand. ❄️ Liquid-Cooled: Zero thermal throttling for sustained peak utilization. Ready to move from run to results? hubs.la/Q046ZSZ10

Big news from #GTC2026! We announced some big upgrades to CoreWeave Cloud, further enabling us to power the next phase of production-scale agentic and physical AI. What’s New: ✅ NVIDIA HGX B300 GA: Expanded accelerator memory and liquid-cooled reliability ✅ W&B Integration: Environment-Free RL for "on-the-job" agent training at 40% lower cost. ✅ Robotics AI: New side-by-side multimodal comparison for embodied AI development. ✅ Mobile-First AI: Introducing the first iOS app to monitor training runs on the go. ✅ NVIDIA Vera Rubin Platform: Deploying NVIDIA Vera Rubin NVL72 & NVIDIA Vera CPU in H2 2026. More here: hubs.la/Q046-GsR0 @Cursor_AI