SHERIDAN, WYOMING - April 14, 2026 - Lightelligence, the optical computing company operating under LIGHTELLIGENCE Co., Ltd., has made its PACE 2 optoelectronic accelerated computing card commercially available, targeting AI infrastructure operators with a PCIe- and CXL-compatible hardware solution. The release marks the company's advancement from photonic development into production-ready AI compute hardware, backed by a platform architecture built around optical multiply-accumulate processing and chip-level optical networking. Alongside PACE 2, Lightelligence's Photowave® optical communications hardware is positioned for disaggregated AI memory deployments in data center environments. The combined product set reflects a concerted push to bring photonics-based performance advantages into mainstream server infrastructure.
PACE 2 and the Photonic Compute Platform
PACE 2 is an optoelectronic accelerated computing card supporting both PCIe and CXL interconnect standards, making it compatible with current and next-generation server configurations. The product forms part of Lightelligence's Photonic Compute product family, which also includes the original PACE accelerator, Optical Fabric, and Lightsphere X. The platform is built on three core technology components: oMAC (Optical Multiply Accumulate), oNOC (Optical Network on Chip), and oNET (Optical Networking). These building blocks are designed to shift matrix-intensive computations - central to AI model inference and training - away from purely electronic pathways and toward optical signal processing, which the company states will deliver exponential increases in computing speed.
Business Impact
The availability of PACE 2 in a PCIe and CXL form factor is commercially significant because it allows data center buyers to integrate photonic acceleration into existing PCIe- and CXL-based server environments without mandatory infrastructure replacement. PCIe 5.0/6.0 and CXL 2.0/3.0 compatibility - specifications explicitly supported by Photowave® - aligns the product line with current high-bandwidth memory disaggregation architectures used in AI data centers. Photowave® enables data center managers to scale compute and memory resources within or across server racks, addressing a core operational constraint in density-sensitive deployments. The latency and energy efficiency advantages attributed to the photonic approach are directly relevant in facilities where power consumption and inference throughput are primary cost drivers.
Photowave and Disaggregated Memory
Photowave® is described as optical communications hardware designed specifically for disaggregated AI memory applications. Its compatibility with both PCIe 5.0/6.0 and CXL 2.0/3.0 positions it to serve environments adopting composable infrastructure models, where compute and memory resources are decoupled and dynamically allocated across racks. The product's photonic interconnect architecture is intended to reduce the latency penalties typically associated with memory disaggregation over conventional electrical interconnects. This makes Photowave® relevant not only to new builds but potentially to operators retrofitting existing high-density AI clusters.
Industry Engagement and Standards Alignment
Lightelligence has maintained an active presence at technical forums directly relevant to its target market. The company appeared at SPIE Photonics West in January 2026, SC25 in November 2025, and the CXL Consortium DevCon in May 2025 - the latter being a key venue for the memory disaggregation standards community. The company also received a Best in Show recognition at Flash Memory Summit in August 2023, an early signal of traction in storage and memory-adjacent markets. These engagements suggest Lightelligence is working within established standards ecosystems rather than pursuing a fully proprietary integration path.
Key facts:
- PACE 2 optoelectronic accelerated computing card is commercially available with PCIe and CXL support for AI workloads
- Photowave® optical communications hardware supports PCIe 5.0/6.0 and CXL 2.0/3.0, targeting disaggregated AI memory applications
- Platform architecture includes oMAC (Optical Multiply Accumulate), oNOC (Optical Network on Chip), and oNET (Optical Networking)
- Lightelligence appeared at SPIE Photonics West (January 2026), SC25 (November 2025), and CXL Consortium DevCon (May 2025)
- Best in Show recognition at Flash Memory Summit (August 2023)