WEKA

What a week at #GITEXGLOBAL in Dubai! The excitement, innovation, and collaboration around the future of #AI infrastructure was off the charts — from conversations on maximizing GPU performance to scaling agentic AI with NeuralMesh. A huge thank you to our incredible WEKA team for bringing the energy every day, to our partners and customers who continue to inspire us with their innovation, and to the GITEX GLOBAL team for putting on another world-class experience. 👋 Until next time

⏪ At World Summit AI Amsterdam last week, WEKA’s Val Bercovici, met with Marzieh Fadaee from Cohere to unpack a critical topic: why the next evolution of AI performance depends on better evals — ones grounded in real-world outcomes, not just saturated benchmarks. As AI moves from the lab to everyday applications, how we measure “intelligence” and “value” needs to evolve too. This conversation dives into what that shift looks like — and why it matters for building truly reliable, capable AI systems. #WorldSummitAI

Join WEKA's Val Bercovici, alongside Ben Mabey (Recursion) and James Chen (Wonder) for The AI Capacity Crunch: A Builder’s Playbook for Latency, Lock-in & Cost, hosted by VentureBeat — a deep dive into the real-world decisions shaping modern AI infrastructure. They’ll explore what it takes to keep systems fast, portable, and affordable when scaling agentic workloads — from model routing and memory efficiency to cost controls and hybrid architecture strategies. This isn’t a panel or a pitch — it’s a working session for the people making AI actually run. 📅 https://weka.ly/4o3cfec

🚀 File system scanning just got a turbo boost! Together, Starfish Storage and WEKA are redefining what’s possible at scale. In lab tests, the combination hit 1.6 billion files scanned per hour—and in production, a leading semiconductor manufacturer nearly doubled that, scanning 3+ PB in just one hour. With WEKA’s metadata performance and Starfish’s multithreaded scanning, large-scale environments are no longer limited by file counts. The future of high-speed, non-disruptive scanning is here. 👉 Read more: https://weka.ly/48v8TLZ

💥 As #AI models get bigger and smarter, they’re slamming into an invisible wall — memory. In a recent interview with The Information, WEKA’s Liran Zvibel breaks down the “AI memory crisis” — a hidden bottleneck slowing inference, wasting GPUs, and driving up infrastructure costs across the industry. He also shares how industry leaders are rethinking efficiency through memory — and why solving this challenge is key to making AI truly scalable and sustainable. 🗞️ Read the full article: https://weka.ly/4ojSL4p

Token prices keep dropping, but somehow your infrastructure bills keep climbing — eating away at your #agenticAI ROI. There’s a better way. Recent production-grade benchmarking on CoreWeave AI cloud shows that WEKA’s Augmented Memory Grid can deliver 4x more token throughput with no new storage hardware required. That means: ⚡ 4.2x more capacity without extra capex and opex. ⏱️ 6x lower latency. 💰 Lower cost per workload. 🚀 Smoother, faster AI experiences that AI researchers & developers actually love. The real problem isn’t your agentic workloads — it’s inefficient system design. See the results and what they mean for your bottom line. 🔗 https://weka.ly/3KYmvWa

🌟 The energy at #GITEXGLOBAL in Dubai this week is electric! From conversations on maximizing GPU performance to scaling agentic #AI with NeuralMesh, it’s been incredible connecting with customers, partners, and fellow innovators shaping the future of AI infrastructure. A big shoutout to our amazing WEKA team on the ground — you make these connections possible!

⚡ The next wave of enterprise #AI is here. We’re proud to collaborate with Red Hat and an incredible community of innovators to power open, production-ready AI at scale. With the launch of Red Hat AI 3, WEKA’s NeuralMesh, NeuralMesh Axon, and Augmented Memory Grid are helping organizations eliminate data bottlenecks, extend GPU memory, and accelerate inference workloads across hybrid environments. Learn more about how Red Hat and its partners are driving the future of open, hybrid AI. 🔗 https://weka.ly/47rLyJZ

📅 Join us at NVIDIA GTC DC and discover how the next-gen #AI stack is driving breakthroughs across government, healthcare, life sciences, fintech, and beyond. From cloud AI to agentic AI to physical AI, see how NeuralMesh + NVIDIA’s advanced platforms eliminate bottlenecks, maximize GPU utilization, and accelerate innovation at scale. Wally MacDermid, WEKA’s Sr. Director, Strategic Alliances – NVIDIA, and our amazing team will be on-site connecting with partners, sharing insights, and showcasing what’s next in AI infrastructure. 👉 Book a meeting with us: https://lnkd.in/gtYkdpeA #NVIDIAGTCDC

🇦🇺 This week, WEKA joined NVIDIA #AI Day at SXSW Sydney — connecting with innovators, customers, and partners who are shaping the next era of AI. During a fireside chat with speakers from NVIDIA and Monash University, WEKA’s Val Bercovici shared how WEKA is helping organizations accelerate inference, reduce costs, and maximize tokens — on-prem or in the cloud. Events like these remind us that AI progress doesn’t happen in isolation — it’s powered by community, collaboration, and the exchange of ideas. #NVIDIAAIDay