The chipmaker presented a roadmap built around high-density power technologies spanning AI data centres, electric vehicles, industrial applications and Physical AI, alongside financial targets for revenue, margins and free cash flow extending to 2030. Speaking at the firm's analyst event, President and CEO Hassane El-Khoury noted that AI infrastructure faces growing constraints not just from compute and memory demands, but also from the challenge of delivering and managing power efficiently. Upcoming infrastructure, he noted, would need power supplies that are dispersed, energy-efficient, and capable of sustaining prolonged peak demand. The firm launched its Embedded Power Platform — known as EPP — positioning it as a broader ecosystem rather than a standalone offering. The technology employs a silicon wafer as an active substrate, integrating diverse semiconductor materials — among them silicon carbide and gallium nitride — within a wafer-level architecture; onsemi noted that the approach targets concurrent gains in electrical, thermal, and mechanical performance while shortening the time needed for system development. According to El-Khoury, EPP is currently undergoing sampling trials for automotive, industrial, and AI data-centre applications through the firm's established 12-inch fabrication process. The executive noted that the platform is compatible with onsemi's high-voltage product lines and its Treo intelligence platform, a system built to enable modular power-control functionality across a range of voltage levels and end markets. Vertical gallium nitride, known as vGaN, was among the technologies the firm drew attention to. According to El-Khoury, onsemi has taken vertical GaN devices through the full development and production cycle, with units now deployed on customer boards across automotive, industrial, and AI data-centre sectors. Manufacturing output would be backed by the firm's existing Syracuse plant, eliminating the need for a new greenfield fabrication facility. The automaker has indicated it is assessing onsemi's integrated power module technology for use in upcoming vehicle platforms. According to a Subaru spokesperson, the technology holds potential for greater power density, cooling on both sides, enhanced electrical management, and lower costs at the system level, though the manufacturer noted it is still assessing where the technology might fit within its lineup. Achyut Shah, who leads onsemi's Power Solutions Group, noted that AI processor energy demands are scaling from the hundreds-of-watts range toward kilowatt territory, with rack-level consumption projected to climb from tens of kilowatts into the megawatt range. The statement outlined that the changeover would involve moving away from conventional alternating-current distribution frameworks toward high-voltage direct-current infrastructure spanning from the power grid through to computing hardware. According to Shah, the move toward higher-voltage rack configurations and sidecar architectures is forecast to gather momentum over the coming year to year and a half. Wider uptake of solid-state transformers was projected to get under way toward the end of 2028 or into early 2029. Shah projected that onsemi's addressable semiconductor content per rack could rise from roughly $15,000 to beyond $115,000 as data-centre infrastructure shifts toward higher-voltage architectures. The executive projected that AI data-centre revenue would surpass $500 million in 2026, representing a year-on-year increase of more than 100 percent, with a further doubling anticipated by 2027. According to its long-range projections, the firm anticipates AI data-centre income will outpace the broader market by a minimum of ten percentage points annually through to 2030. Applying a hypothetical 40% compound annual growth rate for the broader market, Shah indicated that onsemi anticipates its AI data-center revenues expanding by at least 50% per year, rising from a $500 million base in 2026 to above $2.5 billion by 2030. Among the performance figures put forward was an 800-volt-to-6-volt or 12-volt conversion architecture built on vGaN technology, which the firm stated delivers a 20% gain in efficiency alongside a 20% reduction in footprint relative to rival products. The company also outlined an EPP-based silicon carbide JFET solid-state circuit-breaker architecture that, according to its specifications, achieves a 50% reduction in footprint and runs 20% cooler than comparable products on the market. Revenue from a Vcore socket awarded to onsemi is projected to start flowing before the close of 2026, Shah noted. Onsemi's Intelligent Sensing and Analog and Mixed-Signal Group president Sudhir Gopalswamy projected automotive revenue expansion of roughly 9% per year between 2026 and 2030, even under the company's forecast that worldwide light-vehicle production would hold steady across that timeframe. The vehicle roadmap centres on electric powertrains, self-driving technology, and software-defined platforms underpinned by zonal electrical architectures. The executive outlined plans to push forward silicon carbide technology, extend device-voltage capacity to around 1,400 volts for compatibility with 1,000-volt battery systems, and develop intelligent current controllers for use in traction inverters. The company is providing samples of a silicon carbide current-controller product with built-in intelligence capabilities. The firm anticipates that combining sensing and control technologies will deliver a generational leap in performance while lowering certain component costs across the system. Gopalswamy noted that the EPP-based traction inverter architecture could deliver twice the output power within half the physical footprint, achieving a fourfold increase in power density. He estimated that onsemi's addressable market spanning electrification, advanced driver-assistance systems, and zonal architectures had the potential to generate as much as $2,000 worth of semiconductor content in each vehicle. The firm has set a 10% expansion target for industrial markets by 2030, underpinned by investment in grid modernisation, energy infrastructure, factory automation, and rising demand for power systems driven by artificial intelligence. Onsemi's physical AI segment, excluding the automotive sector, could represent a total addressable market of around $6 billion by 2030, according to Gopalswamy. Humanoid robots were projected to generate approximately $900 in potential revenue per unit for the company by that point, according to the statement. According to CFO Thad Trent, onsemi has deployed a total of $8.7 billion over the past five years, comprising $3.6 billion directed toward research and development, $800 million allocated across five tuck-in acquisitions, and $4.3 billion in capital expenditures. Funding was allocated across power platforms, sensing technologies, silicon carbide, vertical GaN, and expanded manufacturing capacity, according to the statement. The company projected its total addressable market would grow by $81 billion, with $45 billion of that figure tied to AI data centres, $20 billion attributed to automotive, and $16 billion linked to industrial sectors, Trent noted. Among its financial objectives for the period spanning 2026 to 2030, the firm has outlined annual revenue expansion of between 12% and 14%, total revenues approaching $11 billion by the decade's end, a gross margin of 53%, an operating margin of 38%, capital expenditure equivalent to roughly 5% of revenues, and a free-cash-flow margin of 30% to 35%, with free cash flow projected to surpass $3.5 billion by 2030. According to Trent, the proposed Synaptics deal falls outside the scope of the financial model. According to the executive, onsemi anticipates that Synaptics — pending completion of the transaction — will achieve annual revenue growth exceeding 15% over the longer term, maintain gross margins of more than 55%, and deliver upwards of $200 million in combined synergies within the first 18 months. The firm outlined a strategy to pursue organic expansion and targeted acquisitions, preserve financial flexibility, and distribute all remaining free cash flow to shareholders once those objectives are met. The company disclosed that approximately $5.3 billion remains available under its current share-buyback programme, according to Trent.
