MaxLinear unveils Carmel as AI servers push console speeds higher. MaxLinear, Inc. (NASDAQ: MXL) has introduced Carmel, a high-speed USB-to-UART device designed to support console access, debugging, monitoring and infrastructure management across AI servers and hyperscale data-centre platforms. The MxL81434 combines 480Mbps High-Speed USB connectivity with four UART channels capable of operating at up to 15Mbps each. It also integrates an I[2]C master and 32 general-purpose input/output (GPIO) channels, allowing several platform-management functions to be consolidated into a single device. The launch comes as AI data-centre architectures become increasingly complex, with server and rack-scale systems requiring faster access for system bring-up, diagnostics, provisioning and recovery. MaxLinear said the market opportunity for USB-to-UART devices is significant, estimating a serviceable available market of about 200 million units driven by AI data-centre deployments. Carmel expands the company's existing USB-to-UART portfolio, which includes Coronado and Laguna, into higher-performance console-access applications. The company is targeting AI servers, accelerator platforms, universal baseboards (UBBs), rack-scale AI systems, hyperscale cloud infrastructure and enterprise data centres. A key feature of Carmel is its ability to provide four independent UART connections at up to 15Mbps per channel. MaxLinear said the configuration is intended to support data-intensive debugging, provisioning, diagnostics and recovery workflows where higher-speed console access can reduce system bring-up and troubleshooting times. The device also incorporates 1,024-byte transmit and receive first-in, first-out (FIFO) buffers. These are designed to maintain data throughput while reducing the processing burden on the host system. Beyond serial connectivity, the integrated I[2]C master and 32 GPIOs allow the device to support additional management functions, including platform monitoring, reset control and debug access. The integration could help server designers reduce the number of discrete components required for these functions. Carmel is also specified with +/-15 kV human-body-model electrostatic discharge (ESD) protection, addressing the electrical robustness requirements of server and data-centre environments. "As AI infrastructure scales, the performance requirements for console access, debug, and platform management continue to increase," said Amit D. Bavisi, senior vice-president and general manager of MaxLinear's Analog Mixed-Signal Business Unit. The product is being introduced as AI infrastructure operators increase compute density and deploy larger accelerator-based systems. In such environments, management and serviceability interfaces remain important for bringing systems online, diagnosing failures and recovering platforms without relying solely on application-level software. WT Microelectronics is supporting distribution of the new product. Johnny Hsu, vice-president of marketing at WT Microelectronics, said the distributor expects Carmel to address increasingly diverse console-access, management and debugging requirements among cloud service providers. MaxLinear's move comes as semiconductor suppliers increasingly target the supporting infrastructure around AI compute, extending beyond processors, memory and networking into components responsible for system management and serviceability. The company expects Carmel samples to become available in the fourth quarter of 2026.

MaxLinear launches RackCommander control-plane portfolio for next-generation AI infrastructure. MaxLinear, Inc. (NASDAQ: MXL), a leader in high-performance connectivity, mixed-signal, and power-management semiconductor solutions, today introduced RackCommander, a portfolio of connectivity, monitoring, control, and power-management products that help enable modern rack-level control-plane architectures for AI, cloud, enterprise, and edge infrastructure. Building on MaxLinear's recently announced Coronado and Laguna USB-UART solutions for AI data center control-plane connectivity, RackCommander brings together products across the key functions needed to manage increasingly complex rack-scale infrastructure. With Full- and High-Speed USB UARTs, RS-485 transceivers, GPIO expanders, power management, and power protection technologies, MaxLinear offers a comprehensive silicon portfolio for rack management and console-port applications. These solutions are designed to support control-plane and console-access requirements for AI infrastructure, including Universal Baseboards (UBBs), multi-node systems, and rack-scale architectures. As AI infrastructure transitions from server-centric architectures to rack-scale systems, the management challenge is changing dramatically. A modern AI rack may include multiple compute trays, dozens of accelerators, power shelves, liquid-cooling subsystems, sensors, service processors, and multiple management controllers. Traditional server-centric management approaches were not designed to efficiently manage these increasingly complex rack-scale environments. Industry analysts project datacenter semiconductor revenue to reach approximately USD 843 billion by 2030. Independent market surveys also project sustained growth for rack-controller and rack-management infrastructure as operators deploy increasingly dense AI systems. Future AI racks require a control plane that spans console access, infrastructure communications, monitoring, telemetry, protection, and rack-level serviceability. RackCommander helps customers address these requirements with MaxLinear products that can support: * Scalable console connectivity across compute, networking, and storage assets * High-speed sideband connectivity for service access and infrastructure control * Rack-level aggregation of management and telemetry signals * Robust communications between power, cooling, sensors, and infrastructure systems * Expanded monitoring and control across distributed rack resources * Power visibility, protection, and resiliency for critical infrastructure Together, these technologies help customers: * Accelerate deployment of rack-scale AI infrastructure * Simplify console and service access across compute, networking, and storage assets * Aggregate telemetry across the rack * Improve operational visibility and reduce infrastructure downtime * Enable scalable rack-management architectures * Lower system complexity and development effort * Support a resilient management plane that remains available independently of the production network For hyperscalers, OEMs, ODMs, and infrastructure providers, RackCommander highlights MaxLinear products that help enable next-generation AI rack-control systems. This differentiation is increasingly important as AI infrastructure evolves toward rack-level architectures that require a more capable control plane for visibility, connectivity, monitoring, and control across the rack. "AI infrastructure is creating a new control-plane architecture that extends far beyond the traditional BMC," said Amit D. Bavisi, Ph.D., Senior Vice President and General Manager of MaxLinear's Analog Mixed-Signal Business Unit. "Future AI racks must securely connect compute nodes, power shelves, cooling systems, sensors, and service processors into a unified management domain. MaxLinear has assembled a comprehensive silicon portfolio for this transition, spanning full- and high-speed USB-UARTs, RS-485 communications, GPIO expansion, eFuse, monitoring, telemetry, power management, and protection. With RackCommander, MaxLinear is helping customers build the control-plane foundation required for next-generation rack-scale AI infrastructure." "We are excited to work with MaxLinear and a top-tier cloud service provider on RackCommander," said Jack Tu, Senior Vice President and Chief Product Marketing Officer of WPI Group. "As AI infrastructure evolves toward rack-scale architectures, customers are looking for proven technologies that accelerate development and deployment while simplifying management of increasingly complex AI infrastructure. RackCommander brings together a broad range of MaxLinear products that help simplify rack management, improve system visibility, and bring next-generation AI infrastructure solutions to market faster."

MaxLinear has launched two intelligent electronic fuse (eFuse) solutions for AI infrastructure power protection. The MxL745950 is a 16V, 50A eFuse with 1mΩ on-resistance, whilst the MxL745951P offers 16V, 60A capacity with 0.79mΩ on-resistance and PMBus/I²C interface support. Both devices are designed for hyperscalers, OEMs, and ODMs to protect and monitor power paths across AI servers, accelerator trays, and data centre switches. They feature integrated hot-swap protection, current sensing, and fault isolation capabilities. The products are currently sampling, with volume production expected in Q4 2026. MaxLinear will showcase them at the OCP APAC Summit in Taipei on 11-12 August 2026. Market research estimates data centre semiconductor revenue will reach approximately $843 billion by 2030, driven by expanding AI compute capacity.