WEKA
Category: AI Infrastructure
WEKA (WekaIO) builds NeuralMesh, an AI-native data and memory platform, and Augmented Memory Grid, which extends GPU memory by persisting LLM KV-cache on flash to cut recomputation in inference. WEKA was founded in 2013. The company is led by Liran Zvibel. Based in Campbell, CA, United States. Team size: 500+. Total funding raised: $415M. Latest round: Series E. Key investors include Valor Equity Partners, Generation Investment Management, Atreides Management, Norwest Venture Partners, NVIDIA, Hitachi Ventures, Qualcomm Ventures, Micron Ventures, Mirae Asset Capital, Celesta Capital, Seagate Technology, Western Digital.
- Founded
- 2013
- Headquarters
- Campbell, CA, United States
- Team size
- 500+
- Total funding
- $415M
Value proposition
Makes production AI faster to train, cheaper to serve and easier to scale. It removes data and memory bottlenecks in GPU clusters, and it reuses LLM context (KV-cache) across GPU servers so GPUs spend less compute on prefill.
Products and solutions
NeuralMesh (software-defined distributed AI storage/data platform, formerly WEKA Data Platform), NeuralMesh Axon (converged deployment inside GPU servers), Augmented Memory Grid (KV-cache memory extension / 'token warehouse' for inference), NeuralMesh AI Data Platform (AIDP, based on the NVIDIA AI Data Platform design), WEKApod appliances
Unique value
Software-defined flash storage that acts as a memory tier for AI. WEKA claims Augmented Memory Grid delivers 1000x more effective memory than DRAM and up to 20x faster time-to-first-token, with microsecond-latency parallel file access.
Target customer
Frontier AI labs, GPU/neocloud providers, hyperscale cloud providers, Fortune 500 enterprises, and HPC and research institutions running large-scale AI training and inference
Industries served
Generative AI / foundation model labs, AI cloud (neoclouds), life sciences, financial services, autonomous vehicles, media & entertainment, government/defense, HPC research
Technology advantage
Parallel, distributed NVMe file system architecture built from scratch for low latency and high throughput. Integrates with NVIDIA Dynamo/NIXL and GPUDirect Storage for KV-cache offload. Certified on NVIDIA GB200 NVL72. Runs on-prem and across AWS/GCP/Azure/OCI.
How they differentiate
Ships a production-validated, packaged KV-cache memory-extension tier (Augmented Memory Grid) on top of a software-defined parallel file system. WEKA positions this as native to the platform rather than bolted on. It claims class-leading density with WEKApod and offers flexible deployment, including converged deployment inside GPU servers (Axon).
Main competitors
VAST Data, DDN, NetApp, Everpure (formerly Pure Storage)
Key partnerships
NVIDIA (investor, NVIDIA AI Data Platform and Dynamo integration, GB200 certification), Oracle Cloud Infrastructure (Augmented Memory Grid validation), CoreWeave, Nebius, HPE, AWS, Google Cloud, Microsoft Azure
Notable customers
Stability AI, Midjourney, Cohere, ElevenLabs, Hugging Face, Together AI, CoreWeave, Nebius, WeRide, Contextual AI, Center for AI Safety
Major milestones
2013: founded by Liran Zvibel, Maor Ben-Dayan and Omri Palmon. 2019: $31.7M Series C. 2022: $135M Series D at $750M valuation. May 2024: $140M Series E at $1.6B valuation (unicorn). 2025: rebranded the platform as NeuralMesh, Augmented Memory Grid GA at NVIDIA GTC, validated on OCI (Nov 2025). Mar 2026: NeuralMesh AI Data Platform GA at GTC 2026. Jul 2026: NeuralMesh 6 and WEKApod 3 launched.
Growth metrics
About $100M revenue in 2025 (Latka estimate). About 511 employees (Nov 2025). 300+ customers, including 12 of the Fortune 50. $1.6B post-money valuation (May 2024).
Market positioning
High-end AI storage and inference-memory infrastructure unicorn. It sits in the premium AI-factory storage tier alongside VAST Data and DDN, and positions itself for the 'inference economy' and agentic AI.
Geographic focus
Global: North America, Europe, Middle East and Asia-Pacific. R&D in Israel and India.
Patents and IP
Holds a portfolio of awarded patents on distributed storage and data-platform technology (exact count not verified).
About Liran Zvibel
Co-founded WEKA in 2013 and previously served as its CTO. Co-founded Fusic and led R&D. Previously principal software architect and founding technical-team member at XIV Storage Systems; subsequently managed XIV hardware and platform teams at IBM. BSc in Mathematics and Computer Science, Tel Aviv University.
Latest news about WEKA
More AI Infrastructure companies
Official website: https://www.weka.io