Software Engineer

Access IPO date, IPO status, valuation, and trade in proven Computing Hardware pre-IPO companies like Alif Semiconductor on Forge marketplace.

Alif Semiconductor Bets on Edge AI Leadership with Next-gen AI MCUs. Article by: stephen las marias. At COMPUTEX 2026, Alif Semiconductor highlighted how AI-enabled MCUs are bringing generative AI to battery-powered edge devices. Alif Semiconductor believes the next phase of edge AI will be defined by how efficiently developers can run increasingly sophisticated neural networks within the tight power, memory, and cost constraints of embedded systems. At COMPUTEX 2026, Alif showcased how its latest AI-enabled microcontrollers (MCUs) are designed to address those challenges while positioning the company for emerging applications ranging from wearables and industrial sensing to smart cameras, robotics and AR glasses. Founded in 2019, Alif may be one of the industry's younger semiconductor companies, but its engineering pedigree is anything but new. The company was co-founded by President Reza Kazerounian, widely known for leading the development of the highly successful STM32 microcontroller family during his tenure at STMicroelectronics, following earlier leadership roles at Freescale Semiconductor. That experience, according to Sampan Chen, director of field marketing for Asia at Alif, helped shape the company's vision from the beginning. "We are a young company, but our experience and technology are definitely not young," Chen said during an interview with EE Times Asia at COMPUTEX 2026. Rather than adapting conventional MCU architectures for AI workloads, Alif built its products specifically for AI at the edge. The company's Ensemble family was the first commercial MCU platform to combine Arm Cortex-M55 processors with the Arm Ethos-U55 neural processing unit (NPU), enabling machine learning inference while maintaining ultra-low power consumption. At COMPUTEX 2026, the company expanded that vision with its next-generation Ensemble series, comprising the E4, E6 and E8 devices. The new family integrates Arm's latest Ethos-U85 NPU alongside Cortex-M55 CPUs, making it one of the earliest MCU platforms capable of supporting transformer-based neural network operators required by small language models and other generative AI workloads. According to Chen, this capability distinguishes the company from conventional MCU suppliers. "We are the first company to adopt Arm's latest NPU technology that supports transformer network operators," he said. "That is the key to enabling generative AI applications such as small language models on microcontrollers." During COMPUTEX demonstrations, Alif showed transformer-based inference running directly on its AI MCU platform, highlighting use cases including intelligent vision, voice processing and multimodal edge AI without requiring continuous cloud connectivity. Solving two persistent edge AI challenges Chen believes customers continue to face two major obstacles when designing edge AI systems. The first is hardware efficiency. Until recently, many microcontrollers simply lacked the processing capability and energy efficiency needed to execute machine learning inference for extended periods on battery-powered devices. The second challenge has shifted toward software. As large AI models continue growing inside cloud data centers, embedded developers must determine how to compress, optimize and deploy those models onto devices with limited memory and compute resources. Chen said the challenge now is no longer only hardware, but how developers reduce those AI models so they can fit into edge devices with limited memory, limited storage and limited cost. Instead of attempting to build every software component internally, Alif has assembled an ecosystem of AI software partners. At COMPUTEX, the company demonstrated collaborations with Nota AI for computer vision optimization, Sensory Inc. for voice AI, Edge Impulse for embedded machine learning development, and EmbedUR Systems for AI software deployment and optimization. The strategy allows customers to combine Alif's hardware with domain-specific AI expertise for applications such as driver monitoring, industrial fault detection, speech recognition and predictive maintenance. "We partner with experts in different AI categories because every customer has different requirements," Chen said. Building AI without sacrificing battery life Power consumption remains the company's primary engineering priority. Unlike many AI processors that rely on large batteries or continuous external power, Alif targets always-on embedded devices expected to operate for months or even years. Its AI MCUs incorporate an integrated power management architecture that partitions silicon into multiple independently controlled power domains. Unused sections of the chip can be completely shut down, minimizing leakage current during standby operation. Chen said deep-sleep current can be reduced to approximately 1μA, enabling significantly longer battery life for sensors, wearables and other battery-operated products. Memory architecture is another differentiator. The Ensemble devices integrate up to 5.5MB of MRAM and 13.5MB of SRAM, allowing many AI models and real-time operating systems to reside entirely on-chip without external memory. Eliminating external DRAM not only reduces system cost but also lowers overall power consumption. Equally important for developers is scalability. The Ensemble family maintains software compatibility across multiple devices while offering pin-to-pin compatibility between package variants, allowing designers to migrate across performance levels without major hardware redesign or software rewrites. "You develop the software once," Chen explained. "Then you can move across the product family with minimal modification." Beyond the cloud Throughout COMPUTEX, edge AI emerged as one of the dominant themes across the exhibition floor. While cloud AI continues driving large language models and data center investments, embedded intelligence is becoming increasingly important for applications requiring low latency, privacy and lower energy consumption. Chen sees edge AI complementing rather than replacing cloud computing. "Our products cannot replace cloud AI," he said. "They are part of the ecosystem. Cloud AI still needs intelligent devices, and intelligent devices also benefit from cloud AI." That philosophy has influenced Alif's product roadmap. Rather than targeting a single vertical market, the company is designing general-purpose AI MCUs capable of serving applications ranging from industrial automation and smart home systems to drones, robotics, wearables and AR glasses. As those products evolve, customer requirements continue expanding beyond AI performance alone. Developers increasingly demand support for multiple image sensors, higher-resolution cameras, larger AI models and longer battery life - all while maintaining competitive system costs. "The future is still about reducing power while increasing performance," Chen said. Improving industries, and lives Despite ongoing concerns that AI could replace human workers, Chen believes edge AI will primarily augment human capabilities by reducing repetitive tasks and minimizing human error. "AI should be a tool that helps people improve their lives, not replace them," he said. "If we use the technology properly, whether in the cloud or at the edge, it can make people's lives and industries much better." That perspective reflects Alif's broader strategy. As AI workloads continue migrating toward endpoint devices, the company is betting that future embedded systems will require more than faster processors. They will need tightly integrated compute, memory, power management and AI acceleration packaged into highly efficient microcontrollers capable of delivering intelligent experiences where they matter most - at the edge.