Lambda

Lambda is first to adopt #NVIDIA co-packaged optics to make large-scale AI clusters more efficient, resilient, and ready for agentic AI. Large GPU clusters are becoming power-bound, reliability-bound, and network-bound. Lambda is taking an early look at co-packaged optics to reduce network overhead and improve tokens per watt. More compute in the same footprint. Fewer interruptions. Better infrastructure for agentic AI. Read more: https://lnkd.in/eSyAVN5y

Heading to #CVPR2026 in Denver? Save Saturday night. Lambda, NVIDIA, and Kodiak are hosting AI After Dark. The official program ends. The real conversations start. June 6. Researchers, engineers, and builders. Spots are limited; approval required. Register: https://luma.com/in9oubm7 #CVPR2026 #AIAfterDark

📣 NVIDIA Blackwell sets a new STAC-AI LANG6 record for LLM inference in quantitative research and algorithmic trading, delivering the highest compute-per-watt and lowest token cost. We tested Llama 3.1 8B and 70B with NVIDIA TensorRT-LLM across multiple NVIDIA platforms. Systems tested: ✅ NVIDIA HGX B200 on Lambda ✅ NVIDIA RTX PRO 6000 Blackwell Server Edition system from Supermicro ✅ NVIDIA Grace Hopper-based server from Hewlett Packard Enterprise See the results 👉 https://nvda.ws/4fFM5ww

Lambda is heading to CVPR 2026 (Denver). Here's where to find us: - Workshops on world models and human-AI learning - Two accepted papers: PixARMesh and fractured object recovery - Live demos at Booth 523: 3D vision, embodied AI, and research agents - Booth 523 | June 3-7 | Colorado Convention Center, Denver Plan your visit: https://lnkd.in/gE4mp4Wj

⚙️ New podcast episode: Robert Brooks IV of Lambda In this episode of The Deep View Conversations, Jason Hiner talked with Brooks, Lambda's chief commercial officer, about the massive AI infrastructure buildout now underway. Lambda’s mission is to build supercomputers for superintelligence. But Brooks argues that the story is bigger than GPUs, data centers, and rising demand. It is about why compute is becoming one of the most strategically important resources in the AI economy, and why Lambda believes compute is not a commodity. The conversation goes deep on Lambda’s vision for democratizing AI, why the company invests in research, and how its experience building physical infrastructure shapes what it can offer AI labs, hyperscalers, enterprises, and researchers. Topics covered include: + Why Lambda thinks “one GPU per person” is achievable + The hidden complexity behind modern AI data centers + Why compute demand keeps surprising the industry + His $40,000 robot experiment and what it taught him about the future of work + How AI is changing the way leaders spend their time If you want to better understand the physical and economic foundations powering the AI boom, this conversation is worth your time. 📺 Watch on YouTube: https://lnkd.in/g2ftJceg 🎧 Listen in your favorite podcast player: https://lnkd.in/gfbAyPsJ Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology. And don't forget to sign up for The Deep View daily newsletter. We don’t just cover AI, we decode it. In a world flooded with hype, we deliver sharp, no-nonsense insights to keep you ahead of the curve and help you put AI to work every day: subscribe.thedeepview.com

DeepSeek V4 was the most anticipated open-source model release of the year. It also landed in the most competitive landscape yet. A few months ago, a 1.6T-parameter open model under MIT license would’ve dominated the news cycle. Now it shares the headline with infrastructure announcements, closed-model updates, and younger labs reaching new heights. The real story in V4 is the engineering, not the capability. A hybrid attention mechanism combining Compressed Sparse Attention and Heavily Compressed Attention cuts single-token inference FLOPs and KV-cache memory at long context. Cost-to-serve drops over 10x versus v3.2. The 1M context window works on price. The capability uplift hasn't followed yet. This is also the floor. V4 is still in pre-release, just as V3 was. The full model is what the community is waiting for. Lambda is deploying both DeepSeek v4 Pro (1.6T, single NVIDIA HGX B200 node) and V4 Flash (284B). Read the full breakdown from Zachary Mueller on our blog: https://lnkd.in/gKCFiwqi

Lambda is partnering with Hudson River Trading to power quantitative research and development using NVIDIA HGX B200 systems — the platform that delivers the highest compute per watt, lowest token cost, and longest useful life. Read the announcement: https://nvda.ws/3Rpl9qM

Every GPU cloud puts a hypervisor between your workload and the silicon. At rack scale, that overhead stops being a rounding error and becomes a tax on every training run. Lambda Bare Metal Instances remove it. Direct hardware access. API-driven lifecycle. No abstraction layer. NVIDIA BlueField DPUs run in Zero Trust Mode, inaccessible from the host OS. The host is yours. Production deployments start with NVIDIA GB300 NVL72. Read more: https://lnkd.in/dWXzYvHs

Hudson River Trading is partnering with Lambda to accelerate quantitative research and development. Lambda will help power HRT’s research roadmap with a full-stack architecture, including NVIDIA HGX B200 systems, advanced networking, storage, orchestration, and uptime. Read more: https://lnkd.in/gKaGf5Yr

Quants and FSI teams don't need generic AI benchmarks. They need to know if the model holds up when the full desk is querying during earnings season. Lambda is the first to publish audited STAC Research's STAC-AI LANG6 on NVIDIA HGX B200. Read more: https://lnkd.in/gUf8A9N2