Company Profile

Building the brain for sovereign edge systems

Sibacus Technologies (a subsidiary of DotOS Holdings Pte. Ltd., Singapore) operates as a pure-IP semiconductor design firm. We develop the architectures, libraries, and compiler runtimes that power sovereign compute networks across ASEAN.

Silicon Sovereignty

We believe digital infrastructure is a defensive community shield. By keeping advanced compute-in-memory IP open to regional partners, we prevent multinational monopolies from capturing digital policy controls.

Sibacus is headquarted in Singapore to access global financial and fabrication networks, with engineering labs in the Philippines training the next generation of analog chip design talent.

Corporate Offices

DotOS Holdings Pte. Ltd.
12 Marina Boulevard, Marina Bay Financial Centre
Singapore 018982

Ecosystem Inquiries

contact@sibacus.me

The Arc: Four Decades of Compute Migration

The workload follows the better economics. It always has.

Sibacus is led by a founder who has benchmarked and optimized compute at every level of the stack — from Cray-class supercomputers in the late 1980s, through engineering workstations, high-performance PCs, and embedded systems. That career traced the same pattern that has repeated across every generation of computing:

1980s
Supercomputers — Centralized. Million-dollar machines in cooled rooms. All data flows to the machine.
1990s
Workstations — Compute moves to the engineer's desk. Sun, SGI, DEC Alpha. Decentralization begins.
2000s
Commodity HPC — PCs become powerful enough for real work. Clusters of commodity hardware outperform supercomputers.
2010s
Cloud Re-centralization — Data flows back to data centers. GPUs need 300-700W and liquid cooling. Bandwidth seems cheap.
2020s+
Edge AI / CIM — Data volume outgrows bandwidth. Power costs become unsustainable. Compute migrates outward again — this time with intelligence.

The pattern is a pendulum — and the driver is always the same physics: when the cost of moving data to the compute exceeds the cost of putting compute at the data, compute migrates outward. Every single time. The cloud was the last cycle of centralization. CIM is the technology that enables the next decentralization — because for the first time, you can run real neural network inference at milliwatts instead of hundreds of watts.

Few people have benchmarked and optimized at every level of this trajectory. Sibacus isn't chasing a trend — it's positioned at the next inflection point of a pattern its founder has watched play out for four decades. The workload follows the better economics. It always has.

Partner With Us

We are building an ecosystem, not a product line.

Sibacus is an IP design layer — we create the architecture, not the end products. Our partners build what matters in their markets. We are actively seeking collaboration with:

Foundries & OSAT

Fabrication partners with mature-node capacity (22nm/28nm DUV) and BEOL RRAM process capability or interest in qualifying one.

Universities & Research Labs

Academic groups working on RRAM device physics, CIM architectures, edge ML, or domain-specific AI models for underserved languages and regions.

Device OEMs & Integrators

Hardware companies building edge devices — drones, satellites, IoT sensors, industrial systems — that need AI inference within strict thermal and power budgets.

Sovereign Technology Funds

National or regional investment bodies seeking to build domestic semiconductor design capability without dependence on EUV-class fabrication.

Domain AI Teams

Groups building narrow AI models for agriculture, maritime, conservation, healthcare, or local languages that need to run at the edge, not in a data center.

Development Finance

Institutions funding capacity building, technology transfer, and infrastructure development in emerging markets across ASEAN and beyond.

Early partners shape the ecosystem. If you see alignment, reach out.

partners@sibacus.me