Atlas
We solve the compute bottleneck holding back real-time, graph-native AI.
- Raising
- Seed
- Round
- Closing
- Cohort
- Autumn 25
- Sector
- Financial Services · Aerospace & Defence
What they do
Today’s CPUs and GPUs are too slow to handle the deep graph data searches required for semantic reasoning at scale. This is most acute in fraud detection, where banks have <100ms to stop crime but lose $80B annually because current systems can’t go deep or fast enough.
Product & technology
Atlas is building a graph-native ASIC: a custom processor engineered to store semantic data and accelerate graph traversal and reasoning tasks 100x faster than GPUs or CPUs. Packaged as a server-ready accelerator card using standard PCIe protocols, it will integrate seamlessly into existing infrastructure. Our first product is this accelerator card bundled with an SDK for frictionless graph database integration, sold as hardware with recurring software licensing and future cloud consumption options.
Why we invested
- →Step-change hardware: graph-native ASIC designed for deep traversal and reasoning, targeting ~100x speed-up vs GPUs/CPUs.
- →Easy to adopt: delivered as a PCIe accelerator card plus SDK, plugging straight into existing servers and graph databases.
- →Attractive model: hardware sale with recurring software licence and future cloud consumption, giving strong margin and expansion potential
