Ubitium Raised $14 Million for New AI Processor
The startup aims to consolidate sensing and AI inference on a single chip, with evaluation units expected in 2027.
Updated on Oct. 13, 2026 in Semiconductors
Live Poll
Will advancements in universal processors for physical AI ultimately make intelligent machines more affordable?
Ubitium has raised $14 million in new seed funding to accelerate the development of its UB410 processor. The company, founded in 2024, intends to create a single-chip architecture that integrates sensing, AI inference, and control functions.
Why it matters
The company aims to replace current systems that rely on separate processors, a design that adds complexity and latency due to required data transfers between distinct hardware units.
The firm's current UB108 chip uses an 8nm manufacturing process to manage 256 processing elements, while the upcoming UB410 is slated to scale to 4,096 processing elements.
The players
Ubitium
A 2024-founded startup developing reconfigurable processors that unify sensing, AI inference, and control functions.
Ibex Investors
A venture capital firm that led the recent $14 million seed funding round for Ubitium.
The details
Ubitium utilizes a reconfigurable processor architecture designed to handle sensing, AI inference, and real-time control functions on a single piece of silicon. This approach relies on unified memory and standard RISC-V development tools—an open-source instruction set architecture that provides the basic machine language commands for a processor. By consolidating these functions, the design reduces the need for data transfers between multiple specialized chips, which typically increases latency and overall system complexity.
Timeline
2024: Ubitium was founded.
October 13, 2026: Ubitium raised $14 million in seed funding.
2027: Ubitium plans the tape-out and delivery of UB410 evaluation units.
The Tech Race
Ubitium is building its roadmap on the open-standard RISC-V ecosystem to challenge vertically integrated proprietary processor designs. The firm aims to scale its architecture from 256 processing elements to 4,096 in its next-generation chip.
The technology remains in development, with no hardware currently available for commercial or industrial deployment. Users and developers should track the 2027 delivery timeline to see if the unified memory approach simplifies integration for edge AI and real-time robotics workflows.
The takeaway
Ubitium is betting that consolidating disparate processing tasks onto one reconfigurable chip can outperform existing multi-processor system designs. Watch for the 2027 evaluation unit release as the first verifiable benchmark for whether the UB410 can match its theoretical density gains.
What happens next
The company is scheduled to initiate the tape-out process for the UB410 and deliver evaluation units to partners in 2027.
Further reading
For more on the current industry trends in hardware development, explore our coverage of Semiconductors.
Live Poll
Will advancements in universal processors for physical AI ultimately make intelligent machines more affordable?
