Hitachi Ventures Backs Edgescale AI to Bring AI Solutions from Cloud to Physical World
There is cloud AI. There is factory floor reality. And then there is the awkward silence between them where latency, legacy systems, and operational risk like to hang out and break things. Edgescale AI decided that silence was unacceptable.
This newly formed industrial AI infrastructure company emerged from stealth in Oct 2024 with a clear thesis: if AI is going to matter in factories, plants, hospitals, warehouses, and utilities, it cannot live exclusively in a hyperscale data center sipping oat milk. It has to get its boots dirty. It has to run on site, shoulder to shoulder with machines, sensors, and control systems that actually move the world.
Co-founder and CEO Brian Mengwasser knows something about building in complex environments. He co-founded Aurora Insight, later acquired by Maxar Technologies, and held senior roles at Dish Network working on developer and marketplace initiatives. Co-founder, Chairman, and CTO Marc Rouanne helped build Dish Network’s cloud-based open RAN 5G network and previously drove mobile and cloud-native architecture at Nokia. Telecom-grade distribution meets industrial-grade consequence.
The product is called Virtual Connected Edge. The hardware is The Cube. And no, this is not another shiny box looking for a rack to haunt. The Cube is a 3U appliance powered by an Intel Xeon 6520P with 24 cores, 128 GB DDR5, 12 TB NVMe PCIe 5.0 storage, and an NVIDIA RTX Pro 5000 Blackwell GPU with 48 GB GDDR7 ECC. It runs on a hardened RHEL-based stack with Red Hat CoreOS, OpenShift, vLLM, and Apollo for continuous delivery. Secure boot with TPM 2.0. AES 256 at rest. TLS 1.2 and 1.3 in transit. Default-deny firewall. Micro-segmented networking. Signed updates. Intrusion detection. Audit logging. Zero trust, but make it practical.
It deploys in under 1 hour. It runs fully offline. It synchronizes when connectivity returns. It connects directly to machines, sensors, control systems, and on-floor software, organizing operational data into an AI-native format through its data manifold. No Rube Goldberg pipelines. No praying to the bandwidth gods.
Edgescale AI partnered with Palantir on Live Edge, integrating Apollo and Sensor Inference Platform to push models and actions from centralized systems into physical operations with closed-loop capability. They also integrated Red Hat vLLM, OpenShift, and RHEL to create a secure, enterprise-grade stack. This is what happens when infrastructure companies collaborate instead of posture.
Now Hitachi Ventures has invested, reinforcing the signal. They back AI, energy, manufacturing, and deep tech through its $400M fund. Alumni Ventures is also in the mix. The amount and valuation are not disclosed. Frankly, the real number that matters is operational uptime in a power plant at 2 AM.
The takeaway is simple and not simple at all. Industrial AI is not a slide deck exercise. It is hardware plus software plus security plus resilience, delivered as one cohesive system. Edgescale AI is building the connective tissue between cloud ambition and physical execution.