All? Who's the dumbfuck not knowing? Scaling DL: continuous, highly randomized, ultra-high-concurrency read access to billions of small unstructured files. Leonardo: Lustre-based DDN EXAScaler storage cluster. Optimized for HPC ops: massive, sequential, parallel block writes

液冷ではExaScalerやリジッドなどの新興も注目。既存重工も負けていません。

The most expensive thing in AI? Forgetting. 🧠 Every time a user steps away from a long AI session — a 100-page contract analysis, a multimodal deep dive — that hard-won context gets evicted from GPU memory. When they return, the model doesn't just reload. It recalculates. Every logical relationship. From scratch. Google Cloud Managed Lustre, powered by DDN EXAScaler, fixes this. Rather than evicting the KV cache, Managed Lustre captures the brain state of the conversation and persists it to storage. When the user returns, the GPU doesn't rethink — it just loads. Sub-second response. Zero wasted compute. At fleet scale, the impact is staggering. Standard architectures leave up to half your GPUs burning cycles on redundant math. With Managed Lustre, that waste disappears — every active GPU is doing productive work, slashing inference total cost of ownership by up to 50%. The customers already running on it are seeing it clearly: → Salesforce Research cut AI training costs by 42% and achieved 70% higher GPU utilization → Resemble AI achieved 100% GPU utilization and called Managed Lustre 6x faster than every other storage solution they evaluated → Google customers are seeing up to 95% TPU utilization and ~25% cost savings at scale — roughly $25M saved for every $100M in infrastructure spend This is what it looks like when storage stops being the bottleneck. See it in action → bit.ly/491Z5bI

Google TPU被视作未来云业务最强的增长点之一，2026年出货量预计为300 万片。那这么大的需求，参与生产的厂商和环节有哪些呢？ 📕芯片设计（ASIC Design） Broadcom $AVGO — TPU 8t 训练芯片 MediaTek 联发科— TPU 8i 推理芯片 $GOOG — 架构定义 软件协同 📗 晶圆制造 $TSMC （N3 现役 / N2 规划中） 📦 先进封装 $TSMC CoWoS-L（主力） ASE/SPIL（补充） 📔高带宽内存（HBM） Samsung — HBM4 主力供应商 SK Hynix — HBM3E 现有供应 Micron $MU — HBM4 爬坡中 📒网络与互联（Networking） Google 自研 — Virgo switch ASIC （核心） Marvell $MRVL — 网络 ASIC Lumentum $LITE — OCS 候选供应商（采样中，未确认） $AAOI / $COHR — 光收发器潜在受益方（无直接 TPU 供应关系披露） 📁 存储（Storage） DDN — Managed Lustre 10T 后端（EXAScaler） Google — TPUDirect Storage / Hyperdisk

Neoclouds $NBIS $CRWV ROIC ~4-6x with "storage-as-memory tier" AI infra startups (WEKA, VAST Data/DASE, DDN ExaScaler/Lustre, PureStorage) that move KV cache to NVMe/SSD fading the recompute cost for stateful AI.

Your EXAScaler is AI-ready. Join us to learn how to unlock it. 🔐 In this recent webinar, DDN's Joel Kaufman and Stephanie Giard break down what KV Cache actually is, why it matters, and how it impacts real-world performance as teams move from training → inference → agentic AI. Watch the replay: bit.ly/4aU60oG #Webinar

From dialogue to demonstrable impact 🦾 This week, the India AI Impact Summit has brought together government, industry, startups, and researchers to turn AI ambition into real-world execution across compute, data pipelines, and AI Infrastructure. DDN’s Atul Vidwansa, took the stage to spotlight what it takes to power India’s Sovereign AI, alongside leaders from Yotta and NVIDIA in the session “Powering India’s Sovereign AI: Infrastructure at National Scale.” It’s not too late to connect with the DDN team. Stop by Booth 5.13 to see how DDN and EXAScaler help accelerate sovereign AI outcomes and keep performance moving from pilot to national scale. Swipe through some highlights of the last few days 📸 IndiaAI #IndiaAIImpactSummit2026 Prem Akella

If you're running large AI workloads, this session could change your infrastructure strategy. 👇 Join Joel Kaufman (Senior Product Manager, DDN) and Peter Bojanic (Principal Customer Solutions Architect, DDN) for a live #webinar exploring how to enable and optimize KV Cache on EXAScaler. This session is focused on practical enablement and operational considerations, building directly on the KV Cache demo video. Reserve your spot now: bit.ly/4kjhpS3

Your EXAScaler is AI-ready. Join us to learn how to unlock it. 🔑 In this session, Joel Kaufman (Senior Product Manager, DDN) and Stephanie Giard (Senior Product Marketing Manager, DDN) will break down how KV Cache works with EXAScaler to improve performance and GPU efficiency. The session focuses on architecture, impact, and when KV Cache delivers the most value, without requiring new infrastructure. What you’ll learn: • How KV Cache reduces time to first token and improves GPU efficiency • Why storage performance and capacity become critical for large context windows • How teams typically introduce and utilize KV Cache on existing EXAScaler systems 🔗 Register now: bit.ly/4rfcvaW

Tintri's Brock Mowry on Kubernetes Data Management & CSI Integration - KubeCon NA 2025. @vmblog bit.ly/4oIqPaY @Tintri @DDNintelligence #MultiCloud #FileStorage #ObjectStorage #AI #HPC #Exascaler #Infinia #GPU #S3 #FastIO #ExaScale #ITPT

🚀 DDN & Yotta unleash Asia's largest sovereign AI platform: EXAScaler powers 8K NVIDIA B200 GPUs in India's Shakti Cloud—multi-million $ boost for secure AI under IndiaAI Mission! Sovereign compute for govt, startups & enterprises, all data-local.

Build what powers the world’s fastest AI and HPC. ⚡ In this short video, our engineers share why they love building at DDN: deep distributed-systems challenges, multi-petabyte scale with EXAScaler® Lustre, and tech that fuels breakthroughs in healthcare, finance, energy, cloud, and research. Ready to grow quickly, both technically and personally, at the intersection of AI, HPC, and large-scale data? Explore open roles👉 bit.ly/4n3AnMB Thank you: Peter Jones, Andreas Dilger, Matthew Raso-Barnett, Colin Faber, Santosh Mahantshetti, Manraj Singh

At DDN, we’re setting new standards for AI and HPC performance in the cloud. Google Cloud Managed Lustre, powered by our EXAScaler® Lustre file system, delivers petabyte-scale storage with sub-millisecond latency, millions of IOPS, and up to 1 TB/s aggregate throughput—purpose-built for the most demanding AI and HPC workloads. And that’s just the start. Google Cloud innovations like Rapid Storage (< 1 ms latency, 20× faster data access, up to 6 TB/s throughput) and Anywhere Cache (SSD-backed zonal read cache for Cloud Storage delivering up to 2.5 TB/s throughput and up to 70% lower latency) work seamlessly with DDN’s technology to accelerate every stage of the AI pipeline—from ingest and preprocessing to training, inference, and discovery—at unprecedented scale. When organizations demand uncompromising performance for their AI and HPC workloads, they choose DDN. Learn more: bit.ly/45bYx1t

📢 Heading to PEARC25 in Columbus, OH? Come connect with DDN and discover how we’re driving performance at scale — and transforming AI support in research computing. 🔍 Workshop The Art of the Stripe 📅 July 21 | 🕘 9:00 AM – 12:30 PM Unpack how Lustre striping impacts I/O performance at scale with DDN experts: 👨‍💻 Morris Skupinsky, Systems Engineer 🤝 Rob Leetun, Account Executive 🤖 Presentation DDN’s AI-Assisted Support Transformation 📅 July 23 | 🕓 3:45 PM – 4:00 PM Explore how DDN is using AI to deliver faster, smarter support for data-intensive workloads 🎤 Kevin Hogerty, Sr. Director, Services and Support We’re proud to support the PEARC community with technologies and services that keep research moving forward. See you in Columbus! #PEARC25 #HPC #Lustre #AIinResearch #DDN #DataIntelligence #ResearchComputing #HPCStorage #AITransformation #ScientificInnovation #EXAScaler

DDN is in the spotlight — and for good reason. From powering the world’s fastest supercomputers to scaling the most demanding AI workloads, DDN has been building at the edge of what’s possible for over 25 years. This new feature in StorageNewsletter dives into the journey of Alex Bouzari and Paul Bloch — from CalTech classmates to co-founders of a global AI infrastructure leader valued at $5B . Here’s a quick snapshot: 🔹 $800M revenue, fueled by AI growth 🔹 $500M invested in R&D 🔹 Trusted by NVIDIA, xAI, NASA, MLB, and more 🔹 Built for real-world breakthroughs in LLMs, RAG, genomics, autonomous systems, and beyond With EXAScaler, Infinia, and xFusionAI, DDN isn’t just delivering infrastructure — it’s delivering advantage. 📖 Read the full spotlight here: storagenewsletter.com/2025/0… #AIInfrastructure #DataIntelligence #HPC #DDN #Infinia #EXAScaler #LifeSciences #AutonomousVehicles #Genomics #DrugDiscovery #RAG #LLM #StorageNewsletter #InnovationLeadership

🚀 DDN Dominates IO500 Again – But the Real Story Is AI Performance at Scale At ISC 2025, the latest IO500 10-Node Production results speak volumes: ✅ 7 of the top 10 systems ✅ 10 of the top 24 All powered by DDN EXAScaler®. But this isn’t just benchmark bragging rights—it’s proof of how DDN is enabling real-world AI success. While others tune for synthetic tests, DDN delivers balanced, high-throughput performance in production environments powering AI, HPC, and enterprise workloads. 📊 Up to 3x faster than WekaFS. 💥 Over 10x faster than VAST Data. 💸 Hundreds of millions in TCO savings. 🧠 15% better GPU efficiency = more results, faster. From national labs to autonomous vehicle research, leaders trust DDN to keep AI pipelines flowing—and GPU farms working at full throttle. 📖 Read Moiz Kohari’s full breakdown of why IO500 results really matter: ddn.com/blog/io500-10-node-p… #AI #HPC #IO500 #EXAScaler #Infinia #DDN #AIInfrastructure #DataIntelligence #GPUAcceleration #ISC2025

We're making it even easier to achieve breakthrough performance for your AI/ML workloads with Google Cloud Managed Lustre—now GA! Powered by DDN’s EXAScaler Lustre, this fully managed service offers up to 1000 MiB/s per TiB & scales to 8 PiB. Read more ↓ cloud.google.com/blog/produc…

🚨 Now Available: Google Cloud Managed Lustre, powered by DDN EXAScaler® The world’s most powerful parallel file system is now a fully managed, cloud-native service — delivering up to 1 TB/s throughput, sub-millisecond latency, and seamless scale from 18 TiB to 8 PiB . ✅ Purpose-built for AI, HPC & data-intensive workloads ✅ Native Google Cloud integration (GKE, Vertex AI, VPC SC & more) ✅ 99.9% SLA with Terraform support and CSI for GKE ✅ Perfect for LLM training, genomic analysis, risk modeling & more “With this launch, organizations can harness DDN’s extreme performance without managing infrastructure.” – Paul Bloch, Co-Founder & President 🌐 Learn more & get started: ddn.com/press-releases/googl… #DDN #GoogleCloud #EXAScaler #ManagedLustre #AIInfrastructure #HPC #GenAI #CloudPerformance #DataIntelligence #VertexAI #GKE #LustreFilesystem

💡 Smart AI Needs Smarter Infrastructure. We’re proud to be named a Sample Vendor in Intelligent Platforms in the Gartner® Hype Cycle™ for Infrastructure Platforms, 2025. Why it matters? Because the next era of AI won’t be won with compute alone—it requires infrastructure that can think. 📍 Real-time performance optimization 📍 Unified pipelines from training to inference 📍 Intelligent tiering and automated data movement 📍 Massive gains in speed, efficiency, and insight From accelerating drug discovery to slashing fraud detection time, DDN’s intelligent platforms are turning raw data into real business outcomes. 🏆 Infrastructure that adapts. 💥 Storage that scales with your AI. 🚀 Results that speak for themselves. 🔗 Learn more about why DDN is the intelligent foundation for scalable, real-time AI: ddn.com/blog/ddn-for-its-int… #AIInfrastructure #GartnerHypeCycle #DataIntelligence #DDN #EXAScaler #Infinia #SmartStorage #AIFoundation #GenAI #AIAtScale #DataDrivenInnovation

🏆 DDN Wins 2025 AI Breakthrough Award for AI Platform Innovation! We’re proud to share that DDN has been named the winner of the “AI Platform Innovation Award” by the AI Breakthrough Awards, recognizing our leadership in powering the most demanding AI and GenAI workloads at scale. 🔍 What sets us apart? AI-native architecture built for performance, trust, and speed Infinia: Cloud-native platform optimized for GenAI and edge EXAScaler: Ultra-low latency, petabyte-scale file system Deep integration with NVIDIA NeMo, TensorFlow, PyTorch, and more Trusted across LLM training, RAG, autonomous systems, and scientific discovery 💡 “AI performance is only as powerful as the data platform behind it,” said Sven Oehme, CTO at DDN. “This award validates the innovation of our teams and the mission-critical work of our customers around the world.” With over 5,000 nominations globally, this award is a testament to DDN’s impact at the intersection of infrastructure and intelligence. 🔗 Learn more → ddn.com/press-releases/ddn-w… #AI #AIInfrastructure #GenAI #LLM #RAG #AIPlatform #HPC #AIInnovation #AIBreakthroughAwards #DDN #DataIntelligence #TechAwards #SvenOehme #PerformanceAtScale