Tensormesh

Great to see the continued momentum coming out of AMD Advancing AI 2026. Thanks to the TensorWave team for highlighting the event! We're excited to continue pushing AI inference forward together. Shout out to our CEO & Co-Founder, Junchen Jiang!

That’s a wrap on #FMS2026! 🎉 A big thank you to everyone who stopped by to connect with the Tensormesh team! We're grateful for the conversations, insights, and connections that made this conference such a great experience. A highlight for us was the panel discussion featuring Yihua Cheng, digging into one of the most pressing challenges in modern AI infrastructure: the KV cache. The panel explored how capacity, bandwidth, tiering, offload, compression, sharing, and reliability all shape latency, throughput, and infrastructure cost. Yihua shared Tensormesh's perspective on the software and systems techniques needed to manage KV cache across GPUs, host memory, and storage, plus where deeper co-design across the stack is still needed. It was also great to see that nearly all the storage vendors here use LMCache Lab to showcase their KV cache performance. Events like FMS are a great reminder of how much innovation is happening across the industry — and we're excited to be part of it. Until next time, FMS! Nicolas Barcet, Junchen Jiang, Karsten Wade, Laura Kerl, Sandro Mazziotta, PhD

𝗧𝗵𝗶𝘀 𝘄𝗲𝗲𝗸 𝗮𝘁 𝗙𝗠𝗦 𝟮𝟬𝟮𝟲 Our CTO, Yihua Cheng, will join the panel discussion, “𝗧𝗵𝗲 𝗙𝘂𝘁𝘂𝗿𝗲 𝗼𝗳 𝗞𝗩$(Cache),” alongside leaders from Dell, WEKA, and Samsung. As context windows grow and inference workloads scale, the KV cache is becoming one of the largest and most performance-critical working sets in modern AI systems. The panel will explore how KV cache capacity, bandwidth, tiering, offload, compression, sharing, and reliability increasingly shape latency, throughput, and infrastructure cost. Yihua will share Tensormesh’s perspective on the software and systems techniques needed to manage KV cache across GPUs, host memory, and storage, and where deeper co-design across the stack is still needed. Open-Source Software, Ballroom F August 6, 2026 9:45 AM–10:50 AM Santa Clara Convention Center #FMS2026 #AIInfrastructure #KVCache #MemorySystems #Tensormesh

We're excited to share that Tensormesh will be exhibiting at the Future of Memory and Storage (FMS) 2026 conference from August 4–6! We're looking forward to connecting with industry leaders and showcasing how we're helping shape the future of AI inference infrastructure. If you're attending #FMS2026, stop by Booth 852 to meet the team (Sandro Mazziotta, PhD, Nicolas Barcet, Karsten Wade, Junchen Jiang, Laura Kerl) and see what we're building, and chat about the future of AI infrastructure. #FMS2026 #AI #ArtificialIntelligence #AIInfrastructure #LLM #Inference #KVCache #Memory #Storage #EnterpriseAI

Hot off the press! We're excited to see Tensormesh featured in the latest Business Insider article exploring how companies are helping enterprises reduce AI costs and maximize ROI. In the piece, our CEO & Co-Founder, Junchen Jiang, highlights an often-overlooked factor in AI infrastructure: "KV cache never appears in people's token bills, but it certainly affects their costs." At Tensormesh, we're helping organizations do more with the infrastructure they already have by optimizing KV cache and model weight management which allows more queries to be served on fewer GPUs while improving efficiency at scale. Thank you to Business Insider for including Tensormesh in the conversation around the future of cost-efficient AI infrastructure. Read the full article here: https://lnkd.in/g5b2gFdq

Thank you to everyone who stopped by the Tensormesh booth at the AI Summit at Stanford University! It was great sharing how Tensormesh is helping organizations maximize GPU utilization, reduce inference costs, and enable more efficient LLM deployments through intelligent KV cache and model weight management. A highlight of the summit was seeing the incredible turnout for Junchen Jiang's keynote, The Next Big Data Layer for AI, and his panel discussion, Infrastructure of Intelligence: Building the AI Stack. Thank you to everyone who joined the conversations and asked such thoughtful questions. We're looking forward to continuing these discussions and helping shape the future of AI infrastructure. #AISummit #Stanford #AIInfrastructure #LLM #Inference #EnterpriseAI #GenAI #GPU 

Excited to share that Tensormesh co-founder and CEO Junchen Jiang will be speaking on a panel at The AI Summit at Stanford University today! "The Infrastructure of Intelligence: Building the AI Stack" 📅 Saturday, August 1st 🕐 1:30–2:00 PM PT 📍 NVIDIA Auditorium, Jen-Hsun Huang Engineering Center Junchen will be joined by an incredible lineup: 👉 Jen Hoskins (NVIDIA) 👉 Yasmina B. (NVIDIA) 👉 Shreyas Pimpalgaonkar (Bespoke Labs) If you're around Stanford this weekend, come join the conversation! 🔗 Learn more and register: https://lnkd.in/gJa4DjG8 #AI #AIInfrastructure #Tensormesh #LMCache #AISummit #Stanford

📷 𝗧𝗵𝗶𝘀 𝘄𝗲𝗲𝗸 𝗮𝘁 𝗧𝗵𝗲 𝗔𝗜 𝗦𝘂𝗺𝗺𝗶𝘁 @ 𝗦𝘁𝗮𝗻𝗳𝗼𝗿𝗱 Our Co-founder, 𝗝𝘂𝗻𝗰𝗵𝗲𝗻 𝗝𝗶𝗮𝗻𝗴, gave a keynote on “The Next Big Data Layer for AI”, exploring why memory and data movement are becoming a critical part of the infrastructure stack for modern AI systems. As AI workloads continue to scale, the bottleneck is increasingly shifting beyond compute. Junchen will share how the next generation of AI infrastructure needs to rethink how data is stored, moved, and reused across inference workloads. If you're building or operating AI infrastructure at scale, come join us at Stanford at the Tensormesh booth! #AISummit #Stanford #AIInfrastructure #LLMInference #LMCache #Tensormesh

LLM inference has long been viewed as a compute problem. But what if we've been looking at it the wrong way? In his latest article, our CTO and Co-Founder, Yihua Cheng, argues that the next wave of AI infrastructure won't be defined by faster GPUs alone. As AI applications become more complex, model-produced state (especially KV cache) is emerging as a new class of data that requires its own infrastructure. The article explores why AI-native data demands a new software stack, how today's approaches to inference are falling short, and what production-grade AI infrastructure will need to look like in the years ahead. If you're building AI agents, long-context applications, or production LLM systems, it's a perspective worth reading. 📖 Read the full article: https://lnkd.in/gzVPYFTi #AI #ArtificialIntelligence #LLM #Inference #AIInfrastructure #KVCache #MLOps #DistributedSystems #OpenSource