WEKA — Israeli-founded AI-native data platform, $1.6B unicorn. Founded 2013 by Liran Zvibel, Omri Palmon, Maor Ben Dayan. Series E $140M led by Valor Equity, backed by NVIDIA. $100M+ ARR. NeuralMesh for BlueField-4. Storage anchor of the Israeli AI infrastructure stack.
WEKA is the Israeli-founded AI-native data platform, the software-defined storage layer built for GPU compute and generative AI training workloads. Founded 2013 in Tel Aviv. Now a US-headquartered unicorn at $1.6 billion, backed by Valor Equity Partners, NVIDIA, and Generation Investment Management. The storage half of the Israeli AI infrastructure stack.
What does WEKA's platform do?
WekaIO Inc., trading as WEKA, is a software-defined data platform designed for accelerated compute. The product replaces traditional NAS and SAN architectures with a GPU-native parallel filesystem that removes I/O as a bottleneck in AI training, inference, and high-performance computing workloads. WEKA's technical claim: eliminate the storage penalty that keeps GPUs idle during large-scale model training.
Headquartered in Campbell, California. Israeli-founded, with the core R&D center in Tel Aviv (approximately 120 employees). Roughly 547 total employees as of mid-2026, up from ~400 during the 2024 unicorn round.
Who founded WEKA?
Liran Zvibel is CEO and co-founder, a prior high-performance storage engineer and veteran of the Israeli deep-tech infrastructure scene, and the public face of the company.
Omri Palmon is a co-founder handling product and architecture leadership.
Maor Ben Dayan is Chief Architect and co-founder, the technical anchor of the parallel-filesystem architecture.
How much funding has WEKA raised?
- Founded 2013. Multiple early rounds through Series B (Qualcomm Ventures, Celesta Capital, Gemini) starting from 2016.
- Series D, November 2022. $135M at a $750M valuation. NVIDIA joined as strategic investor. Hitachi Ventures and Atreides Management participated.
- Series E, May 2024. $140M at $1.6B valuation, more than 2x the prior mark. Led by Valor Equity Partners, whose founder Antonio Gracias joined the board. Participation from NVIDIA, Generation Investment Management, Atreides, 10D, Hitachi Ventures, Ibex, Key1 Capital, Lumir, MoreTech Ventures, Qualcomm Ventures, Micron Ventures.
- Total raised: approximately $372 to $375 million to date.
- Israeli unicorn: 2024, the Israeli infrastructure-software unicorn of the AI-storage category.
How much revenue does WEKA generate?
- Annual recurring revenue: crossed $100M in 2024, the eight-figure enterprise-contract cadence WEKA leadership pointed to as the acceleration signal that closed the Series E.
- Historical revenue: approximately $62.9M as of December 2023 before the AI training buildout compounded.
- Customers: hyperscalers, enterprise AI teams, national laboratories, and financial-services HPC operators. NVIDIA is both an investor and a strategic partner, and the WEKA data platform is validated across NVIDIA GPU reference architectures.
What is NeuralMesh, and where is WEKA's product headed?
In October 2025, WEKA announced NeuralMesh Architecture, built for NVIDIA BlueField-4 DPUs. The direction is toward disaggregated data plane execution, offloading storage protocol handling from the CPU to the DPU inside the same GPU server. That roadmap ties WEKA to the NVIDIA-defined generative-AI infrastructure stack, and separates WEKA from generalist enterprise storage incumbents (Dell PowerScale, NetApp, Pure Storage).
Where does WEKA sit in the Israeli AI infrastructure stack?
The Israeli AI infrastructure layer, taken together, forms one of the most concentrated national footprints in the category:
- WEKA, GPU-native storage / parallel filesystem. The category WEKA defines.
- NextSilicon, alternative compute architecture. Custom silicon for HPC and AI acceleration.
- Run:ai, GPU orchestration. Acquired by NVIDIA 2024.
- Pinecone, vector databases. Israeli-founded, US-headquartered.
- Gambit, inference optimization. The newer entrant.
WEKA is the most heavily capitalized private Israeli operator in the storage sub-layer. Competitors globally include DDN (still private, older category), Vast Data (US-founded but heavily Israeli-staffed), Panasas, and the legacy NAS incumbents.
Why does WEKA matter?
The storage bottleneck. Every large-model training run is I/O-bound at some threshold. WEKA's technical bet is that traditional storage architectures cannot serve GPU-scale training economically. If correct, the storage market for AI training splits from the historical enterprise NAS market entirely, and WEKA holds anchor position in the new one.
The NVIDIA alignment. NVIDIA is both investor and strategic partner. Every incremental NVIDIA GPU shipment expands WEKA's TAM proportionally. The BlueField-4 architecture WEKA is building for is the DPU category NVIDIA is now standardizing.
The Israeli infrastructure export. Alongside NextSilicon on the compute side, WEKA is the storage-side proof point that Israeli deep-tech infrastructure can hold anchor category positions inside the global AI economy, not just the classic Israeli categories of cyber and semi-analog.
What is the one-line summary?
WEKA is the Tel Aviv-founded, California-headquartered AI-native data platform, a $1.6B unicorn with $100M+ ARR, backed by NVIDIA and Valor Equity, building the GPU-native storage layer that separates AI-training infrastructure from legacy enterprise storage. The storage anchor of the Israeli AI infrastructure stack.
How we report: Olam stories draw on public filings, company disclosures and named sources, checked by our editors. Read our methodology · Corrections
