Loading article…
Mixx Technologies unveiled the SxC connector, which supports 24,576 fibers per rack unit—a 4x density increase over current industry-standard MMC-VSFF hardware.
San Jose-based Mixx Technologies has launched the SxC™ Connector, an optical interface designed to support multi-petabit connectivity for hyperscale AI infrastructure by terminating up to 24,576 fibers in a single rack unit [1]. This hardware aims to resolve the "inference challenge" by increasing radix—the number of endpoints a single node can reach—to reduce latency and power consumption in large-scale AI compute clusters [1, 2].
| At a glance | |
|---|---|
| Product | SxC™ Connector |
| Fiber Density | 24,576 fibers per rack unit |
| Density Improvement | 4x vs. MMC-VSFF |
| Launch Date | September 2, 2026 |
The SxC connector is the first product from the HBxIO™ platform, an architecture Mixx is developing to facilitate the industry's transition toward co-packaged optics (CPO) [1, 2]. Current data center designs are increasingly limited by fiber density at the box edge, as advanced cooling and packaging technologies have allowed silicon to scale faster than the optical I/O required to move data between nodes [1].
By utilizing a wafer-scale integrated process to form optical structures, Mixx claims its connector can terminate 1,024 fibers in less than 800 square millimeters [1, 2]. This density allows a single rack unit to house the same number of connections that previously required a four-rack-unit enclosure using MMC-VSFF, the current highest-density platform qualified for deployment [1, 2]. The connector supports both single-mode (SMF) and polarization-maintaining fiber (PMF) within the same form factor, allowing it to serve both front-plane and backplane applications [1, 2].
Mixx, which recently acquired Sophic Silicon to bolster its analog mixed-signal capabilities, is positioning its technology to replace the fragmented connector ecosystem currently used in data centers [2]. CEO Vivek Raghuraman stated that the industry has historically relied on different connectors for different generations, such as MPO and VSFF, but that the demands of mixture-of-experts and agentic AI workloads necessitate a unified ferrule technology that can handle the entire signal chain—from the package to the laser and the backplane [1, 2].
To address manufacturing bottlenecks, Mixx is automating cable assembly through robotic fiber assembly and automated visual inspection [1]. This approach seeks to eliminate the manual, active-alignment steps that have historically limited yield and throughput in optical connector production [1].
Whether this connector becomes the new standard for AI-scale networking depends on its ability to move from bench-scale demonstrations to mass-market deployment in power-intensive, high-radix environments.
Coverage is mostly measured — 112 of 112 reports stay neutral.
Every Monday — the token unlocks, Fed dates & catalysts set to move crypto and markets this week. So you’re never blindsided.
Free · 3-min read · one-click unsubscribe
AI-assisted synthesis by the TrendWatcher Editorial Desk · sourced from 3 outlets · Sep 2, 2026 · How we report
Layer 2 Scaling is necessary because blockchains possess fundamental limitations that prevent them from facilitating global-scale transactions without sacrificing decentralization and verifiability. By moving frequent, smaller transactions to a secondary layer, the main blockchain can function as a secure settlement layer.
Examples of Layer 2 Scaling systems include the Lightning Network, Ark, sidechains, statechains, rollups, and client-side validated systems. These architectures are designed to facilitate higher transactional volumes while maintaining varying degrees of independence from the main chain.
Layer 2 Scaling systems attempt to mitigate the need for trusted third parties by developing mechanisms that allow users to unilaterally exit back to the mainchain. While early sidechain designs often required some form of trusted third party, newer schemes focus on permissionless exit capabilities.