ON Semiconductor (ON.US) reported its latest quarterly results and forward guidance early Tuesday. While the company's fourth-quarter performance, ending December 31, 2025, exceeded average Wall Street analyst estimates, its outlook for the current quarter fell short of expectations. This disappointing forecast caused the chipmaker's stock to drop more than 8% in after-hours trading.
Despite a challenging 2025, during which ON Semiconductor's stock fell 15% for the full year, its shares have surged approximately 20% since the start of 2026. This rally has been fueled by market optimism surrounding unprecedented AI-driven demand for analog chips and power management semiconductors essential for data centers.
Financial results for the fourth quarter showed total revenue of approximately $1.53 billion, representing an 11% year-over-year decline. Non-GAAP adjusted earnings per share were $0.64, significantly lower than the $0.95 reported for the same period last year. However, both figures surpassed analyst expectations of roughly $1.52 billion in revenue and $0.62 EPS.
The company's operating margin for Q4 was about 13.1%, down from 23.7% a year earlier. The non-GAAP operating margin stood at 19.8%, compared to 26.7% in the prior-year period.
By business segment, revenue from the Power Solutions Group (PSG) was $724 million, down 11% year-over-year and 2% sequentially. The Analog and Mixed-Signal Group (AMG) generated $556 million, a 9% annual decrease and a 5% quarterly decline. The Intelligent Sensing Group (ISG) reported revenue of $250 million, falling 17% compared to the previous year but increasing 9% from the prior quarter.
For the current quarter, ON Semiconductor's management provided guidance with a midpoint projection of $1.49 billion in revenue (range: $1.435 billion to $1.535 billion) and non-GAAP EPS of $0.61 (range: $0.56 to $0.66). Wall Street analysts had anticipated revenue of about $1.51 billion and EPS of $0.63. Compared to the year-ago quarter's revenue of $1.45 billion and adjusted EPS of $0.55, the new guidance suggests a potential return to growth after more than two years of annual declines.
The significant stock appreciation this year is largely attributed to ON Semiconductor's positioning within the AI data center power chain. Since the second half of 2025, the company has benefited from robust demand for power-related semiconductors driven by intensive AI data center construction. Growth in its AI infrastructure-focused power management products has helped offset weaker spending in the electric vehicle sector, particularly for silicon carbide and EV-related components, amid persistently soft demand since late 2022. This indicates that market expectations of AI-driven performance are grounded in actual orders and revenue, not mere speculation.
ON Semiconductor's three main business segments position it as a power and energy efficiency semiconductor technology company rather than a traditional analog IC-focused firm. The core thesis behind its stock outperforming the S&P 500 and Nasdaq 100 indices this year is its strategic leverage in the "AI data center power chain." A key catalyst was the announcement in late 2025 that the company would collaborate with AI chip leader NVIDIA (NVDA.US) to advance 800V DC power solutions for next-generation AI data centers.
From a technical perspective, ON Semiconductor's critical connection to large-scale AI data centers, including projects like the $500 billion "Stargate," lies in the "grid-to-GPU" power conversion chain—a fundamental requirement in the AI computing era. Another vital link is analog and power management technology. As AI cabinet power scales from traditional 10–20 kW levels to 100 kW or higher, power efficiency becomes a primary constraint. Even a 1% efficiency gain can significantly reduce thermal losses, cooling needs, and electricity costs.
The company's senior management has directly addressed power and thermal management challenges arising from rising AI data center cabinet power and conversion efficiency, noting the industry's shift toward higher power density and 48V architectures. Key components in this data center power chain typically include front-end AC-DC/PFC units, high-frequency DC-DC converters for improved efficiency and smaller size, 48V power distribution and intermediate bus conversion to minimize resistive losses, and point-of-load (POL) power delivery to efficiently step down voltage to sub-1V levels required by GPUs and CPUs.
ON Semiconductor's core offerings in this domain—power devices (MOSFETs, power modules, SiC), power conversion, and driver control ICs—form the essential hardware foundation, positioning the company to naturally benefit from AI data center expansion and energy efficiency upgrades.
Therefore, the combination of strong demand for power semiconductors and efficiency upgrades related to AI data centers—particularly in 48V systems, power supply units, distribution, and high-efficiency switching devices—coupled with optimistic market expectations for a recovery in automotive and industrial cycles and inventory normalization, has driven ON Semiconductor's stock higher this year.
However, the weaker-than-expected guidance has led investors to question whether the anticipated "super-cycle" of AI data center-driven revenue and profit growth has truly arrived. Consequently, the stock declined sharply following the earnings release.
The AI-driven surge in chip demand, initially concentrated in AI processors and memory, is now spreading to analog chips and power semiconductors. Recent strong results and positive AI data center revenue outlooks from analog chip leaders like Texas Instruments and STMicroelectronics, alongside robust earnings from Infineon, indicate that the long-awaited recovery in analog chip demand, fueled by intensive AI data center build-out, is underway.
The seemingly insatiable chip demand generated by AI training and inference is cascading from computing and memory chips to the analog segment, powerfully boosting the performance of industry leaders like Texas Instruments and Infineon. The AI-driven demand wave is expanding from "compute chips themselves (GPU/ASIC/HBM)" to the broader "power and signal chain (power + analog/mixed-signal)," with the spillover effect accelerating noticeably.
Infineon recently raised its current fiscal year investment plan from approximately €2.2 billion to €2.7 billion, projecting AI data center-related revenue to grow from about €1.5 billion to €2.5 billion by 2027. The core rationale is that AI data center demand provides a strong tailwind amid a weaker automotive cycle.
Similarly, Texas Instruments issued a better-than-expected Q1 outlook, explicitly citing demand from AI data center investments as a contributing factor. The market interprets these developments as the analog chain beginning to capture substantial benefits from AI infrastructure spending.
The fundamental engineering logic behind analog chips benefiting from AI compute infrastructure is straightforward: AI training and inference systems are pushing power consumption per cabinet or rack to unprecedented levels, forcing power architecture upgrades (e.g., to 48V or higher voltage HVDC). This leads to a non-linear increase in the semiconductor content of power devices and power management systems.
The Open Compute Project has indicated that AI rack power will "soon exceed 500 kW," while NVIDIA is advancing high-voltage DC architectures for "AI factories," targeting power levels from 100 kW to over 1 MW per rack. Rising power demands necessitate more than just additional MOSFETs; they drive increased complexity and higher specifications across the entire power chain. This includes stricter efficiency and thermal design for AC-DC and DC-DC conversion, more complex multi-stage conversion from 48V (or higher) to point-of-load levels below 1V, and steeper transient current demands from GPUs/CPUs. These factors significantly boost the usage and performance requirements for power stages (FETs/power modules/drivers), multi-phase controllers, hot-swap controllers/e-fuses, isolation and current/voltage sensing, clocks, and high-speed signal conditioning components.