AI data center "power engine" has not yet exploded. ON Semiconductor Corporation (ON.US) Q4 revenue declined by 11%, performance outlook below expectations.

ON Semiconductor Corporation, a chip giant focusing on the automotive and industrial sectors, released its latest quarterly performance data and future outlook on Tuesday morning Beijing time. The data showed that ON Semiconductor Corporation's fourth-quarter performance as of December 31, 2025, exceeded the average expectations of Wall Street analysts, but the company's latest outlook for the current quarter (first quarter) fell short of Wall Street expectations, causing the chip giant's stock price to drop more than 8% in after-hours trading. Since the beginning of this year, ON Semiconductor Corporation's stock price has continued to rise sharply, driven by the unprecedented wave of AI related to analog chips/data center power chains. The increase since 2026 has been as high as 20%. In comparison, the stock price of this chip giant for the whole of 2025 fell by 15%. According to the financial report data, ON Semiconductor Corporation's fourth-quarter total revenue was approximately $15.3 billion, representing an 11% decrease year-on-year. Adjusted earnings per share under Non-GAAP guidelines were $0.64, significantly lower than the approximately $0.95 in the same period last year. However, both performance data exceeded the average expectations of Wall Street analysts of approximately $15.2 billion and $0.62 per share. ON Semiconductor Corporation's Q4 operating profit margin was approximately 13.1%, lower than the approximately 23.7% in the same period last year. The operating profit margin under Non-GAAP guidelines was approximately 19.8%, also lower than the approximately 26.7% in the same period last year. By business segment, the fourth-quarter revenue of Analog Solutions Group (PSG) was $724 million, down 11% year-on-year and 2% compared to the previous quarter; Analog and Mixed Signal Technology Group (AMG) revenue was approximately $556 million, down 9% year-on-year and 5% compared to the previous quarter; and Intelligent Sensing Group (ISG) revenue was approximately $250 million, down 17% year-on-year and up 9% compared to the previous quarter. In terms of performance outlook that the market is focusing on, according to the midpoint of the latest performance outlook range provided by ON Semiconductor Corporation management, the company expects total revenue of approximately $1.49 billion for the current quarter (outlook range is $1.435 billion to $1.535 billion), with adjusted earnings per share expectations at $0.61 per share (outlook range is $0.56 to $0.66). In comparison, Wall Street analysts expect total revenue for the first quarter to be approximately $1.51 billion and earnings per share to be approximately $0.63. For comparison, in the same period last year, ON Semiconductor Corporation's total revenue was approximately $1.45 billion and adjusted earnings per share were approximately $0.55. This also indicates that after experiencing a year-over-year decline for more than two years, ON Semiconductor Corporation's total revenue and earnings per share for the first quarter are expected to resume growth. Why has ON Semiconductor Corporation's stock price risen sharply this year? Since the second half of 2025, the performance of this chip giant focusing on analog chips and power semiconductors has been continuously benefiting from the strong demand for power chain-related chips driven by the AI data center construction process, especially the growth of its power management products focused on AI infrastructure. This growth has offset the more cautious spending on electric car (especially SiC and EV-related) end and the continued sluggish demand for electric vehicles since the end of 2022. This indicates that the market is not simply imagining "AI driving ON Semiconductor Corporation's performance", but rather it is a reflection of existing orders/revenues. The three major divisions of ON Semiconductor Corporation's main business determine that the company is "more like a power/efficiency semiconductor technology company" rather than a traditional pure analog IC chip company. The company's stock price has risen by more than 20% since the beginning of this year, significantly outperforming the S&P 500 index and the Nasdaq 100 index, which is a core logic of the company benefiting significantly from the "AI data center power chain" - especially when the "AI chip hegemon" NVIDIA Corporation (NVDA.US) announced in the second half of 2025 that it would cooperate with the company to accelerate the next generation of artificial intelligence data centers with an 800-volt direct current (800V DC) power solution. From a fundamental technological perspective, ON Semiconductor Corporation, as a core connection point for AI data centers such as the "$500 billion interstellar gate", is in the "power conversion chain from grid to GPU (grid-to-GPU)" - this is one of the most certain demands in the AI computing power era. Another core connection point is in "analog/power management". The power of AI racks is transitioning from traditional 10-20kW to higher (even hundreds of kW) super trends, turning "power efficiency" into a primary constraint: every 1% increase in efficiency significantly reduces waste heat, cooling, and electricity costs. The senior management of ON Semiconductor Corporation directly discusses the power and conversion efficiency of AI data center racks in their own technological content, and points out that the industry is transitioning to higher power density and 48V architectures. In the data center-related chain mentioned above, the core devices/solutions commonly used include: front-end AC-DC/PFC + high-frequency DC-DC (increasing efficiency, reducing size, adapting to higher power density); 48V distribution and intermediate bus conversion (reducing IR losses to facilitate high-current power supply); point-of-load (POL) power supply (efficiently reducing 48V/12V to sub-1V voltage required by GPUs/CPUs, requiring low-loss switch devices and drivers/control). In data center solutions, ON Semiconductor Corporation focuses on power devices (MOSFET/power modules/SiC, etc.) + power conversion and driver control, which are the "hardware foundation." The company naturally benefits from the expansion and efficiency upgrade of AI data centers. Therefore, the power semiconductors closely related to the "AI data center power chain" and efficiency upgrades (especially data center 48V, PSU, distribution, and efficient switch devices), along with the market's optimistic expectations for the automotive/industrial cycle and inventory inflection point, have jointly driven ON Semiconductor Corporation's stock price to rise significantly since the beginning of this year. However, as the latest performance outlook fails to beat Wall Street's average expectations, the market now believes that the "super cycle" of strong revenue and profit growth driven by AI data centers has not yet arrived. Therefore, after announcing the latest performance and future outlook, ON Semiconductor Corporation's stock price plunged significantly in after-hours trading. The "chip demand frenzy" triggered by AI has finally spread from AI chips and storage chips to analog chips and power semiconductors. The analog/power chip sector where ON Semiconductor Corporation is located, as well as the power semiconductor/discrete device sector, have seen a strong recovery driven by the unprecedented wave of AI. The recent strong performance of analog chip giants such as Texas Instruments Incorporated, STMicroelectronics NV ADR RegS, and the positive outlook for AI data center-related revenue, along with the recent strong performance data from Infineon, all indicate that the market's expectations of "strong recovery in analog chip demand driven by the AI data center construction boom" are unfolding in the chip industry. Under the unprecedented AI wave, the "endless" chip demand brought by AI training/inference is smoothly transitioning from AI chips and storage chips to the analog chip side, thereby powerfully driving leading analog chip companies such as Texas Instruments Incorporated and Infineon towards a path of strong recovery. The chip demand frenzy brought by AI is transitioning from "computational chip entities (GPU/ASIC/HBM)" to a more generalized "power and signal chain (power + analog/mixed-signal)", and this transition is accelerating significantly. Infineon recently disclosed that it has increased its investment plan for this fiscal year from approximately 22 billion euros to approximately 27 billion, and it expects AI data center-related revenue to increase from approximately 15 billion euros to 25 billion euros (by 2027), with the core logic being that AI data center demand provides a strong "tailwind" in a weak automotive cycle. At the same time, Texas Instruments Incorporated gave a stronger than expected outlook for the first quarter, explicitly listing demand driven by AI data center investments as one of the driving factors. The market also interprets these data as "the analog chain beginning to reap the benefits of the AI infrastructure super boom". The underlying logic of analog chips hitching a ride on the crazy bottom-line engineering of AI computing infrastructure construction is straightforward: the AI training/inference system has pushed "energy consumption per rack" to a historical new level, forcing the upgrade of power architecture (from 48V to higher voltage HVDC), resulting in a "non-linear uplift" in the demand for power devices and power management in the entire chain. OCP has publicly suggested that AI rack power will soon exceed 500 kW, and NVIDIA Corporation is also advancing a high-voltage DC architecture for "AI factories", aimed at covering rack volumes ranging from 100 kW to over 1 MW. The increase in power consumption does not just mean "spending more to buy a few more MOSFETs", but it brings about a stack of devices across the entire chain: the efficiency and thermal design constraints of AC/DC and DC/DC become stricter, multi-stage transformations from 48V (or higher voltage) to point loads (<1V) become more complex, and GPU/CPU transient currents become steeper - all of which significantly increase the demand for power devices (FET/power modules/drivers), multi-phase controllers, hot-swappable/electronic fuses, isolation and current/voltage detection, clocks, and high-speed signal conditioning in niche areas.