With a Revolutionary Service-Oriented Mesh Architecture, NeuralMesh Optimizes AI Infrastructures to Create Resilient, Efficient, Massively Scalable Token Warehouses and AI Factories That Accelerate Time to First Token and Lower the Cost of AI Innovation
SINGAPORE and CAMPBELL, Calif., June 18, 2025 /PRNewswire/ — From SuperAI 2025: WEKA today unveiled a revolutionary advancement in AI data infrastructure with the debut of NeuralMesh™, a powerful new software-defined storage system featuring a dynamic mesh architecture that provides an intelligent, adaptive foundation for enterprise AI and agentic AI innovation. WEKA’s NeuralMesh is purpose-built to help enterprises rapidly develop and scale AI factories and token warehouses and deploy intelligent AI agents, delivering world-class performance with microsecond latency to support real-time reasoning and response times. Unlike traditional data platforms and storage architectures, which become more fragile as AI environments grow and stall as AI workload performance demands increase, NeuralMesh does the opposite—becoming more powerful and resilient as it scales. When hardware fails, the system rebuilds in minutes, not hours. As data grows to exabytes, performance improves rather than degrades.
With The Rise of Inference, Traditional Data Infrastructure Is Reaching Its Tipping Point
The AI industry is shifting from AI model training to inference and real-time reasoning with unforeseen velocity. As agentic AI proliferates, AI teams require adaptive infrastructure that can respond in microseconds, not milliseconds, drawing insights from multimodal AI models across distributed global networks. These increased performance and scale requirements are straining traditional data architectures and storage, pushing them to their breaking point. As a result, organizations face mounting infrastructure costs and latent performance as their GPUs—the engines of AI innovation—sit idle, waiting for data, burning energy, and slowing token output. Ultimately, many enterprises are compelled to augment their data and GPU infrastructure by continually adding costly compute and memory resources to keep pace with their AI development needs, thereby contributing to unsustainably high innovation costs.
“AI innovation continues to evolve at a blistering pace. The age of reasoning is upon us. The data solutions and architectures we relied on to navigate past technology paradigm shifts cannot support the immense performance density and scale required to support agentic AI and reasoning workloads. Across our customer base, we are seeing petascale customer environments growing to exabyte scale at an incomprehensible rate,” said Liran Zvibel, cofounder and CEO at WEKA. “The future is exascale. Regardless of where you are in your AI journey today, your data architecture must be able to adapt and scale to support this inevitability or risk falling behind.”
NeuralMesh: Purpose-Built to Power Agentic AI Innovation and Dynamic AI Factories
With NeuralMesh, WEKA has completely reimagined data infrastructure for the agentic AI era, providing a fully containerized, mesh-based architecture that seamlessly connects data, storage, compute, and AI services. NeuralMesh is the world’s only intelligent, adaptive storage system purpose-built for accelerating GPUs, TPUs, and AI workloads.
But NeuralMesh is more than just storage. Its software-defined microservices-based architecture doesn’t just adapt to scale—it feeds on it, becoming faster, more efficient, and more resilient as it grows, from petabytes to exabytes and beyond. NeuralMesh is as flexible and composable as modern AI applications themselves, adapting effortlessly to every deployment strategy—from bare metal to multicloud and everything in between. Organizations can start small and scale seamlessly without costly replacements or complex migrations.
NeuralMesh’s architecture delivers five breakthrough capabilities:
- Consistent, lightning-fast data access in microseconds, even with massive datasets
- Self-healing infrastructure that gets stronger as it scales
- Deploy-anywhere flexibility across data center, edge, cloud, hybrid, and multicloud
- Intelligent monitoring that optimizes performance automatically
- Enterprise-grade security with zero-compromise performance
Unlike rigid platforms that force AI teams to work around limitations, NeuralMesh dynamically adapts to the variable needs of AI workflows, providing a flexible and intelligent foundation for enterprise and agentic AI innovation. Whether an organization is building AI factories, token warehouses, or looking to operationalize AI in their enterprise, NeuralMesh unleashes the full power of GPUs and TPUs, dramatically increasing token output while keeping energy, cloud, and AI infrastructure costs under control to deliver real business impact:
- AI Companies can train models faster and deploy agents that reason and respond instantly, gaining a competitive advantage through a superior user experience.
- Hyperscale and Neocloud Service Providers can serve more customers with the same infrastructure while delivering guaranteed performance at scale.
- Enterprises can deploy and scale AI-ready infrastructure and intelligent automation throughout their operations without complexity.
“WEKA delivers exceptional performance density in a compact footprint at a very cost-effective price point, enabling us to customize AI storage solutions for each of our customers’ unique requirements,” said Dave Driggers, CEO and cofounder at Cirrascale Cloud Services. “Whether our clients need S3 compatibility for seamless data migration or the ability to burst to high-performance storage when computational demands spike, WEKA eliminates the data bottlenecks that constrain AI training, inference, and research workloads, enabling them to focus on developing breakthrough innovation rather than managing storage and AI infrastructure complexities.”
“Nebius’ mission is to empower enterprises with the most advanced AI infrastructure available. Our customers’ most demanding workloads require consistent, ultra-low-latency performance and exceptional throughput for training and inference at scale,” said Arkady Volozh, founder and CEO of Nebius. “Our collaboration with WEKA enables us to offer outstanding performance and scalability, so that our clients can harness the full potential of AI to drive innovation and accelerate growth.”
“With WEKA, we now achieve 93% GPU utilization during AI model training and have increased our cloud storage capacity by 1.5x at 80% of the previous cost,” said Chad Wood, HPC Engineering Lead at Stability AI.
Over a Decade In The Making
WEKA’s NeuralMesh system is underpinned by more than 140 patents and over a decade of innovation. What started as a parallel file system for high-performance computing (HPC) and machine learning workloads, before AI applications became mainstream, evolved into a high-performance data platform for AI, a market category WEKA pioneered in 2021. But NeuralMesh is more than just the next evolutionary step in WEKA’s innovation journey. It’s a revolutionary leap to meet the exploding growth and unpredictable demands of the dynamic AI market in the age of reasoning.
“WEKA is not just making storage faster. We’ve created an intelligent foundation for AI innovation that empowers enterprises to operationalize AI into all aspects of their business and enables AI agents to reason and react in real time,” said Ajay Singh, Chief Product Officer at WEKA. “NeuralMesh delivers all the benefits our customers loved about the WEKA Data Platform, but with an adaptable, resilient mesh architecture and intelligent services designed for the variability and low latency requirements of real-world AI systems, while allowing growth to exascale and beyond.”
Availability
NeuralMesh is available in limited release for enterprise and large-scale AI deployments, with general availability scheduled for fall 2025. For more information:
- Watch the NeuralMesh launch video: https://weka.ly/nmvideo
- See ‘How It Works’: https://weka.ly/howitworks
- Visit our blog: https://weka.ly/nmblog
About The Author
WEKA’s 2025 NeuralMesh launch redefines AI data infrastructure for the reasoning age, a software-defined storage system with a service-oriented mesh architecture built for enterprise/agentic AI. Unlike fragile traditional platforms, it grows more resilient and performant at exabyte scale, delivering microsecond latency, fast hardware rebuilds, and five breakthrough capabilities. Backed by 140+ patents, it boosts GPU utilization, cuts AI costs, and is in limited release (general availability fall 2025), with industry partners praising its game-changing scalability.