Deep-tech semiconductor startup Staera Semiconductor has exited stealth mode with $4.6 million in pre-seed funding to commercialize Lightwire, an in-wafer optical interconnect technology engineered to replace copper wiring in artificial intelligence computing clusters. Spun out from the Singapore-MIT Alliance for Research and Technology (SMART) Low Energy Electronic Systems (LEES) research group with dual operations in Singapore and San Francisco, the venture integrates gallium nitride (GaN) LEDs and silicon nitride (SiN) waveguides directly onto standard commercial silicon wafers, targeting a tenfold reduction in interconnect power consumption for academic supercomputing facilities and hyperscale AI infrastructure.
Funding Details and the Stealth Exit
Staera Semiconductor announced its stealth exit on November 10, 2026, closing a $4.6 million pre-seed funding round led by Openspace Capital. Co-investors included SGInnovate, Black Kite Capital, 28 Holdings, In Group, and Decimus Technologies. The company was originally established as New Silicon Corporation (NSC) before adopting the Staera Semiconductor name to mark its transition from academic research to commercial product delivery.
The startup will deploy the capital to expand engineering across Singapore and San Francisco, accelerate prototyping with tier-one semiconductor foundries, and deepen engagement with hardware vendors. Staera previously earned early industry recognition by winning the 3rd ECTC Start-up Competition at the IEEE Electronic Components and Technology Conference and finishing as a runner-up in the Lam Capital Venture Competition in mid-2026.
The Lightwire Architecture: In-Wafer Optical Generation
As deep learning models scale, moving data between graphics processing units, neural processors, and high-bandwidth memory has become the primary bottleneck in system performance. Traditional copper interconnects encounter severe signal attenuation, heat generation, and capacitive delays across millimeter-to-meter distances, leaving processors stalled while awaiting memory operands.
Co-packaged optics (CPO) attempts to mitigate distance-dependent losses by mounting discrete optical transceivers alongside processors on specialized substrates. However, these systems rely on external or bolt-on indium phosphide (InP) or gallium arsenide (GaAs) lasers. Discrete modules require delicate fiber coupling alignments, strict thermal management, and expensive packaging assembly flows.
Staera’s proprietary Lightwire architecture embeds the optical link directly inside commercial silicon wafers. By heterogeneously integrating micro-scale gallium nitride (GaN) LEDs, silicon nitride (SiN) waveguides, and standard silicon complementary metal-oxide-semiconductor (CMOS) logic on a single substrate, Lightwire produces light emitters and routing channels within established commercial foundries. The platform builds on foundational heterogeneous integration research developed by the SMART LEES interdisciplinary team, which has fabricated integrated GaN-on-silicon devices in commercial foundries since 2016.
Performance Targets and Specifications
According to specifications released by Staera Semiconductor, the Lightwire platform targets notable improvements in energy consumption, data throughput, and routing density over traditional electrical traces and co-packaged optical assemblies. All performance figures and milestone timelines represent company-reported targets:
- Energy Efficiency: Staera targets interconnect energy consumption under 0.2 picojoules per bit (less than 0.2 pJ/bit), representing less than one-tenth the power required by conventional co-packaged optics (typically drawing roughly 5 pJ/bit) and vastly improving upon electrical copper at high frequencies.
- Total Stack Throughput: When deployed in high-bandwidth memory interfaces, Lightwire is designed to move over 24 terabytes per second (greater than 24 TB/s) per memory stack while drawing less than 50 watts of power, compared to roughly 2.8 TB/s in emerging standard HBM4 interfaces.
- Shoreline Density: The micro-optical architecture delivers a linear shoreline interconnect density exceeding 600 optical links per millimeter of chip edge (more than 600 links/mm), reducing the physical die perimeter required for high-speed I/O.
- Per-Link Data Rates: Individual optical links operate between 4 and 8 Gbit/s, enabling fine-grained parallel data paths without requiring complex, power-hungry multi-level digital signal processors.
- Physical Reach: Lightwire links are engineered to span distances exceeding 50 millimeters (greater than 50 mm), bridging inter-die and board-level distances where advanced packaging standards like Universal Chiplet Interconnect Express (UCIe-A) typically degrade beyond 2 millimeters.
| Metric / Feature | Standard Electrical Copper (UCIe-A) | Co-Packaged Optics (CPO) | Staera Lightwire (Targeted) |
|---|---|---|---|
| Light Source Integration | None (Pure Electrical) | Discrete / Bolt-On InP Lasers | In-Wafer Monolithic GaN LEDs |
| Energy Consumption | 0.5 to 2.0 pJ/bit (Short Reach) | Approximately 5.0 pJ/bit | Under 0.2 pJ/bit |
| Physical Interconnect Reach | Up to 2 mm (Die-to-Die) | Meters (Fiber-Coupled) | Over 50 mm (In-Wafer / Inter-Die) |
| Shoreline Density | Moderate (Pin-Pitch Limited) | Moderate (Fiber-Array Limited) | Exceeding 600 links / mm |
| Thermal & Assembly Overhead | Low Packaging Complexity; High Heat | High Packaging Cost; Thermal Sensitivity | Standard Foundry Flow; Low Heat |
Industry Context: The AI Data Movement Bottleneck
Modern AI supercomputing workloads demand massive aggregate bandwidth across distributed memory and accelerator networks. However, electrical copper wiring exhibits exponential transmission loss at multi-gigahertz frequencies. In dense server racks, copper traces require retimers and active cooling, inflating the operational compute carbon footprint of institutional research clusters.
As computing centers pursue systemic energy proportionality (computing), interconnect power has become an architectural bottleneck. In large model training runs, accelerators frequently stall during collective communications while memory buffers synchronize. By embedding optical emitters directly into the silicon substrate, in-wafer photonics eliminates copper trace impedance mismatch without the assembly overhead of discrete external lasers.
Executive and Investor Perspectives
Announcing the stealth exit, Dr. Andrew Kim, Chief Executive Officer of Staera Semiconductor and former member of the Scientific Advisory Board for the SMART LEES program, outlined the challenge facing hardware designers:
“AI’s bottleneck is no longer how fast chips compute, but how fast and far they can move data. Copper performance plummets over the millimeter-to-meter distances that matter most in AI accelerators, and bolt-on optics add power and complexity. We build the light into the chip, the entire optical link inside a chip, so the connection gets smaller, cooler and faster with each generation. This funding takes Lightwire from the lab to customers, and the Staera name marks that move.”
Hian Goh, Co-Founder of lead investor Openspace Capital, underscored the broader infrastructure implications:
“Staera is addressing one of AI infrastructure’s most urgent challenges: moving vast amounts of data between chips without a corresponding rise in power, heat and complexity. Its ability to build optical links directly into silicon offers a compelling path to faster, more efficient, and more scalable AI systems. We backed Staera because of the strength of its technology, the depth of its team and its potential to define a new category in semiconductor connectivity.”
Implications for University Research and Research Offices
The emergence of Staera Semiconductor provides an operational case study in university research commercialization for capital-intensive deep technology. Originating from research supported by the National Research Foundation Singapore and the Massachusetts Institute of Technology, the venture reflects the pathways through which academic consortia translate materials science into market-ready ventures.
For academic research institutions and administrative leaders, the development highlights three operational dimensions:
- University Intellectual Property and Spin-Out Structuring: Developing hardware that bridges compound semiconductors and silicon CMOS requires coordinated IP protection. Technology transfer offices operating under institutional university intellectual property policies must balance sponsored research mandates with venture capital expectations. When academic discoveries spin out via a university technology licensing office (TLO), clear licensing agreements are critical to ensure that foundational patents can be safely licensed into a venture-backed commercial entity.
- Cleanroom Priorities and Interdisciplinary Fabrication: Academic cleanrooms have traditionally segregated silicon CMOS from III-V materials. Heterogeneous integration platforms, uniting GaN emitters with SiN waveguiding, require integrated cleanroom protocols and updated engineering curricula. Research presented at venues such as SPIE Photonics West increasingly emphasizes foundry-compatible materials integration.
- High-Performance Computing Facility Planning: Research computing directors planning multi-year cluster refreshes must monitor shifting interconnect roadmaps. As in-wafer photonics matures toward commercial availability, academic data centers may benefit from lower facility cooling requirements and denser compute topologies for scientific machine learning.
Commercial Roadmap and Next Steps
Staera Semiconductor is targeting an initial commercial release circa 2030. In the interim, the company will use its pre-seed funding to expand engineering teams in Singapore and California, validate silicon test vehicles across commercial packaging lines, and partner with semiconductor vendors to demonstrate multi-die optical interposers.
Over the next several years, the company must verify that monolithic GaN-on-silicon integration maintains reliable manufacturing yields and thermal stability under data center workloads. All commercial milestones and release timelines remain company projections as pilot development continues.
Sources
- PR Newswire — Staera Semiconductor Emerges from Stealth with US$4.6 Million to Build Light into the Chip and Break AI’s Data Bottleneck (Published 2026-11-10; accessed 2026-11-15).
- Semiconductor Today — Staera emerges from stealth with $4.6m in pre-seed funding (Published 2026-11-10; accessed 2026-11-15).
- TechNode Global — Singapore’s Staera Semiconductor raises $4.6M to put optical links inside AI chips (Published 2026-11-11; accessed 2026-11-15).
- Optics and Photonics News (Optica OPN) — Staera Raises US$4.6 Million for On-Chip Optical Interconnects (Published 2026-11-12; accessed 2026-11-15).







