Data center "power engine" experiences explosive growth! The AI surge spills over to the power supply chain; ON Semiconductor (ON.US) Q2 cash flow quadruples and outlook beats expectations

According to Zhihui Finance APP, Onsemi (ON.US), the semiconductor giant focused on chips for the automotive and industrial sectors, released its earnings and guidance after the US market closed on Monday (Beijing time Tuesday morning). The company's Q3 revenue guidance range exceeded the consensus expectations of Wall Street analysts, highlighting the explosive growth of its AI data center "power engine" business—the surging demand for power management chips for AI data centers. This drove the company's stock price up by more than 7% in after-hours trading. Amid the AI infrastructure boom, the company's performance for the second quarter ended July 3 also remained robust, especially the Power Solutions Group (PSG), covering automotive, industrial, and AI data center businesses, which achieved the strongest growth momentum.

Since the beginning of this year, Onsemi’s stock price has surged sharply with the unprecedented wave of AI data center construction driving strong recovery expectations for analog/data center power supply chain-related chip demand. Although the stock price has pulled back since July following global AI computing power themes’ de-leveraging and crowded positioning unwinds, the increase since 2026 remains as high as 50%. In contrast, the chip giant’s stock price declined by as much as 15% for the full year in 2025.

The chip demand frenzy brought by AI is spilling over from "computing chips themselves (GPU/ASIC/HBM)" to the broader "data center power and signal chain," with the spillover effect accelerating significantly, further validating the so-called "the end of AI is power" logic related to the data center power chain bull market. The seemingly "endless" chip demand brought by AI training/inference is shifting smoothly from AI chips and memory chips to analog and power semiconductors, strongly driving leading analog chip makers such as Texas Instruments, Infineon, and Onsemi toward a strong recovery trajectory. The market interprets these robust results as the "analog chain beginning to reap the super dividends of AI infrastructure."

$7 Billion Acquisition and AI Data Centers Ignite Power Chip Demand, Onsemi Launches a New Growth Cycle

While announcing strong results and guidance, Onsemi is also working to capture the growing demand in the AI devices and robotics sectors through its largest acquisition to date—a $7 billion all-stock deal announced in June to acquire Synaptics, a subsidiary of EDA chip design software titan Synopsys.

Onsemi management expects third-quarter revenue to be between $1.65 and $1.75 billion. According to LSEG compiled data, the mid-point of its guidance range exceeds the consensus estimate of $1.67 billion from Wall Street analysts.

Onsemi CEO Hassane El-Khoury stated in the earnings release, "AI data center-related business continues to be our fastest-growing segment. We currently expect this business segment to at least double its revenue by 2026, which reflects the strong capabilities of our intelligent power portfolio and the growing adoption of Onsemi solutions by customers across the entire power architecture."

For the second quarter ended July 3, the company’s revenue was $1.6 billion, up 9.2% year-on-year, slightly above the consensus estimate of about $1.59 billion among Wall Street analysts. Adjusted earnings per share were $0.74, up roughly 40% year-on-year and higher than the market estimate of $0.71.

Management expects third-quarter adjusted EPS to be between $0.81 and $0.93, with the mid-point notably above the consensus estimate of about $0.83 from Wall Street analysts.

Onsemi is also continuing its “Fab Right” strategy to significantly cut costs and greatly improve operational efficiency. In July, as part of this strategy, the company sold two manufacturing plants.

For other Q2 performance data, Onsemi's GAAP gross margin increased from 37.6% to 38.4%, and non-GAAP gross margin rose to 39.3%. GAAP operating margin rose from 13.2% to 16.1%, and non-GAAP operating margin increased from 17.3% to 20.8%. Net income attributable to shareholders rose from $170.3 million to $226.8 million, up a substantial 33.2% year-over-year, with GAAP EPS rising from $0.41 to $0.56, a 36.6% increase.

The company’s operating cash flow as of July 3 grew 149.4% year-over-year to $459.7 million. Free cash flow soared from $106.1 million to $425.4 million, nearly quadrupling year-on-year, with free cash flow margin rising from about 7% to 27%. The company repurchased $332 million in stock this quarter. In other words, although revenue grew by just about 9%, profit, cash flow, and shareholder returns expanded by several times, reflecting that Onsemi’s Fab Right capacity optimization, cost discipline, and product portfolio upgrades are already delivering real operational leverage.

Segment data further demonstrate that growth is clearly concentrating in data center power semiconductor business. Power Solutions Group (PSG) revenue reached $829 million, up 18.7% year-on-year and 13% sequentially, increasing its share of total revenue from 47.5% in the same period last year to 51.7%, making it the company's main source of income. The Analog and Mixed Signal Group (AMG) achieved revenue of about $545.7 million, down 1.8% year-on-year, while the Intelligent Sensing Group (ISG) booked $228.8 million in revenue, up 6.6% year-on-year.

In the first half of the year, Onsemi posted total revenue of $3.1168 billion, up 6.9% year-on-year. Adjusted EPS rose from $1.08 to $1.38, a 27.8% year-on-year increase, and free cash flow was $642.6 million, up 14.6%. Both semi-annual and segment results indicate that Onsemi has not yet achieved synchronized growth across all business lines. Instead, growth is led first by PSG, driven by AI data centers, high-voltage power, and partial automotive electrification demand, while AMG remains in a relatively moderate recovery phase.

The "Chip Demand Frenzy" Triggered by the AI Infrastructure Boom Is Spilling Over from AI Chips and Memory to Analog and Power Semiconductors

Onsemi’s robust results indicate that AI chip demand is structurally spilling over from GPUs, ASICs, and HBM itself to power semiconductors, analog control, and monitoring signal chains. This isn’t just analog chip companies “hitching a ride on AI”—it’s a natural physical demand due to increased computing density.

Onsemi has already defined AI data center as its fastest-growing sector and expects related revenue to more than double by 2026. Texas Instruments’ data center revenue grew about 90% year-on-year in Q1 2026, with further Q2 growth led collectively by industrial, data center, and automotive. Infineon has joined the Nvidia MGX ecosystem, providing a complete solution for 800V DC power—ranging from high-voltage conversion to GPU core power—driving significant growth in its first-half results. AI capex is thus broadening its beneficiaries from a few advanced process chips to a large number of relatively low-priced, high-quantity, long-lifecycle basic analog devices forming the backbone of the data center power chain that must be certified for reliability.

In the analog/power chip field and the power semiconductor/discrete device sector, the core underlying logic behind the strong demand expansion due to the unprecedented AI wave lies in that AI server racks are leaping from tens of kilowatts in traditional servers to above 100 kilowatts, moving towards 600 kilowatts or even 1 megawatt. Next-generation GPUs/TPUs/ASICs may each consume 2–4 kilowatts, and traditional 48V/54V architectures will bring thousands of amperes of current, with copper loss, heat generation, and busbar size all surging sharply. That is why global AI data centers are pivoting actively to ±400V or 800V DC architectures driven by Onsemi and Nvidia.

Before power from the grid reaches the GPU, it must go through AC/DC rectification, PSU, BBU, high-voltage distribution, 800V to 50V intermediate bus conversion, and then a multiphase VRM steps voltage down to below 1V for the GPU core. Each stage requires Si, SiC or GaN power switches, gate drivers, digital controllers, power management ICs, and electronic fuses. Onsemi, Texas Instruments, and Infineon precisely sell these "per-watt essential" devices. The higher the rack power and the more stages in conversion, the greater the semiconductor value per cabinet.

Onsemi is able to supply, within this "grid-to-GPU" power tree, EliteSiC MOSFET/JFET, high/low voltage silicon MOSFETs, GaN devices, hot-swap intelligent fuses, multiphase controllers, power stages, and PoL regulators—so a single AI rack not only requires more chips but also more expensive, higher voltage and more efficient power devices. Nvidia's MGX platform is essentially Onsemi's key strategic entry into the AI data center market. Onsemi already supplies power FETs, multiphase power, SiC JFETs, and GaN solutions for existing MGX systems, directly serving Nvidia along with PSU, BBU, and future 800V distribution board vendors within the MGX ecosystem. Standardized MGX server and rack designs will also make it easier for a certified power device to be replicated by multiple OEM and ODM partners.