At a glance
| Booth | I2923 |
| Country | TW |
| Website | mixxtech.io |
Company profile
mixxtech.io was incorporated in 2023 by a team that had already taken multiple optical products into volume production, and it builds full-stack optical interconnect for AI data-center infrastructure. The company addresses the gap between rapidly scaling accelerator compute and the movement of data by treating bandwidth density, energy efficiency, and manufacturability as linked design requirements. Its optical path is designed as one integrated system, and products share the 5DS Platform as a common foundation across optics, electronics, packaging, and control software. The broader system-level connectivity approach combines high-density silicon photonics with advanced packaging, system design, and software for AI workloads and compute scaling. The company is headquartered in California, USA, with operations in Taiwan and India, including Hsinchu and Bangalore locations identified in its own company information. Its team describes experience developing and scaling next-generation interconnects from concept to high-volume deployment for hyperscalers. Hyperscalers and ecosystem partners are described as working with mixxtech.io to qualify and drive adoption of next-generation AI interconnect solutions. The company positions its engineering around performance, power, efficiency, scalability, reliability, and practical manufacturability for evolving AI infrastructure. Its stated global approach combines strategic collaborations, research, and ecosystem partnerships across major technology hubs. Company milestones listed on its site include a manufacturing collaboration with Kaynes Semicon and the acquisition of Sophic Silicon Technologies as part of scaling production and semiconductor capability.
Exhibits
The product set includes sXc™, an ultra-high-density EBO connector; HBxIO™, a silicon-level optical engine; and STxIQ™, a software platform for link management, telemetry, calibration, and diagnostics. sXc™ is designed to move a maximum fiber count through a minimum aperture, combining ultra-high fiber density in a compact EBO form factor with Ultra High Radix and Robust Reliability features. One sXc™ connector architecture serves backplane, front-panel, patch-panel, and shuffle-box applications, reducing the need for separate connector architectures across those deployment positions. HBxIO™ brings photonics progressively closer to the compute die to increase bandwidth density and lower energy per bit at the silicon level. STxIQ™ adds operational software functions so production users can monitor links, collect telemetry, calibrate optical behavior, run diagnostics, and tune photonics in service. At the solution level, mixxtech.io describes an architecture that unifies compute, memory, and network as a foundational block for next-generation AI connectivity. The Engine layer integrates an optical engine with core compute, while The Package layer links exaflop-class computing concepts with petabit connectivity. The System layer extends the connectivity fabric to rack scale and combines it with intelligent orchestration and flexible resource allocation so compute resources can be placed where they are needed. Multi-petabit interfaces are intended to support real-time inference and large-scale AI training with low latency and high throughput. The reliability approach integrates siliconized optical interconnects with advanced ASICs to reduce network complexity and support continuous operation for real-time AI workloads. The solution roadmap also emphasizes standards-compliant designs so new connectivity technology can be integrated while infrastructure evolves across generations. Across the stack, the offering connects the physical optical path, high-density packaging, compute-adjacent optical engines, connectors, and control software rather than treating each layer as an isolated component. This full-stack structure is aimed at AI data-center infrastructure where accelerator performance creates increasing demands for bandwidth density, energy efficiency, low-latency data movement, and manufacturable interconnects. The high-density silicon-photonics approach is presented as part of an end-to-end connectivity architecture that links device-level photonics with packaging, system design, and software rather than optimizing only one physical layer. The company describes its connectivity solutions in terms of scalability, density, efficiency, speed, and reliability, with those attributes treated together for the changing mix of AI training and inference workloads. System-level end-to-end connectivity is intended to enable compute scaling while addressing current gaps in performance, power, and efficiency across AI infrastructure. The SxC™ connector launch is associated with high-radix scale-up and multi-petabit connectivity, extending the connector role from dense fiber attachment toward rack-scale AI connectivity requirements. The site also describes an ultra-efficient connectivity fabric for demanding workloads and a system concept that combines rack-scale connectivity with sustainable architecture goals. Taken together, the connector, optical-engine, software, package, and system layers form a path from physical fiber density to managed optical links and large-scale resource connectivity for AI infrastructure.
Capabilities and products
- full-stack optical interconnect for AI data center infrastructure
- ultra-high-density EBO connector
- shared connector architecture for backplane, front panel, patch panel, and shuffle box
- silicon-level optical engine
- optical-link management, telemetry, calibration, and diagnostics software
- silicon-photonics and advanced-packaging integration
- compute, memory, and network integration
- optical-engine integration with core compute
- multi-petabit interfaces for real-time inference and large-scale AI training
- siliconized optical interconnect integration with advanced ASICs
- rack-scale connectivity fabric
- intelligent orchestration and flexible resource allocation