Onsemi's Q2 Beat Shows the AI Power Race Is Accelerating

Onsemi reported second-quarter revenue of $1.60 billion, then raised its 2026 AI data center outlook beyond 100% growth. That combination pushed the chipmaker into google news coverage usually dominated by Nvidia, AMD, and other processor suppliers.

The numbers matter because onsemi does not manufacture the accelerators training or serving AI models. It supplies power semiconductors that convert, regulate, and deliver electricity from the utility connection to processors inside each rack.

This creates a more consequential contest than another earnings beat. Onsemi is betting that rising rack power will shift spending toward complete power-delivery systems. Infineon, STMicroelectronics, Texas Instruments, and Analog Devices are chasing many of the same sockets.

The opportunity is real, but the scale remains unclear. Onsemi does not separately disclose AI data center revenue in its main end-market breakdown. Its fastest-growing narrative therefore sits inside a broader business still led by automotive and industrial customers.

Why Onsemi's Google News Moment Matters

The headline result was not merely a quarterly beat. Onsemi connected improving companywide performance to a much larger AI power opportunity.

Onsemi reported $1.6035 billion in second-quarter revenue for the period ending July 3, 2026. Revenue rose 6% sequentially and approximately 9% from the year-earlier quarter.

The result exceeded the midpoint of management's previous guidance. Non-GAAP gross margin reached 39.3%, an increase of 80 basis points from the first quarter. A basis point equals one-hundredth of a percentage point.

Non-GAAP operating margin rose 170 basis points sequentially to 20.8%. Non-GAAP diluted earnings per share reached $0.74, up from $0.64 in the first quarter.

Those figures appear in onsemi's quarterly presentation. The presentation also says second-quarter free cash flow increased threefold year over year to approximately 27% of revenue.

Onsemi placed one operating claim beside those financial improvements. The company raised its AI data center outlook and now expects related revenue to more than double during 2026.

That outlook extends momentum reported one quarter earlier. In its first-quarter results, onsemi said AI data center revenue had already more than doubled year over year.

Management attributed that earlier increase to broader adoption across the power tree. It also cited business with multiple chip vendors and leading hyperscalers, meaning companies operating very large cloud platforms.

The second-quarter announcement strengthened that narrative because the broader company improved at the same time. Revenue, margin, earnings, and cash generation all moved in the desired direction.

Management also guided third-quarter revenue between $1.65 billion and $1.75 billion. The midpoint represents another sequential increase from the second-quarter result.

Its non-GAAP gross-margin forecast ranges from 40% to 42%. Non-GAAP diluted earnings guidance ranges from $0.81 to $0.93.

That is why the story traveled beyond an ordinary earnings recap. Onsemi presented AI power as a growth engine arriving while its wider automotive and industrial operations recover.

Yet the distinction between acceleration and transformation remains important. AI data center sales can double from a relatively small base without changing the company's overall revenue mix immediately.

The headline establishes momentum. The next question is why AI facilities suddenly need more of the components onsemi sells.

AI Data Centers Need More Than Faster GPUs

Every additional watt consumed by an AI processor must travel through a chain of conversion and protection components before reaching the silicon.

An AI power tree is the complete path that moves electricity from the grid to processors. It includes converters, power supplies, distribution units, backup systems, controllers, and point-of-load components.

Traditional data centers already required this infrastructure. AI changes the engineering problem because accelerator clusters pack far more computing demand into each rack.

Higher density creates two connected pressures. Operators must deliver more electricity through limited physical space, and they must reduce the energy lost during repeated voltage conversions.

Power conversion is never perfectly efficient. Energy lost during the process becomes heat, which adds cooling demand and reduces the facility's useful computing output.

Onsemi's presentation maps its products across both current 54-volt rack designs and future 800-volt direct-current systems. Direct current, or DC, moves electrical charge in one direction.

The company says it can address voltages from 0.8 volts to 800 volts. Its portfolio includes silicon, silicon carbide, gallium nitride, and vertical gallium nitride devices.

Silicon carbide and gallium nitride are wide-bandgap materials. They can support higher voltages, temperatures, and switching frequencies than conventional silicon in suitable applications.

That does not mean one material replaces every other option. Cost, reliability, voltage, switching frequency, packaging, and customer qualification determine which device fits each stage.

Onsemi's strategic argument focuses on coverage. If the company supplies more stages of the power tree, each data center deployment can generate more component revenue.

Its estimates illustrate the potential change. Onsemi puts its current opportunity near $15,000 per rack under a 54-volt design.

For a future rack operating between 600 kilowatts and one megawatt, onsemi estimates an opportunity near $115,000. Those figures are company projections, not independently verified market averages.

The increase depends on new components entering the system as power density rises. It also assumes customers select onsemi across several conversion stages instead of buying isolated parts.

Industry evidence supports the broader need for better power delivery. A power-chip analysis cited Omdia estimates that global power-semiconductor revenue will reach $100 billion by 2029.

That report placed 2026 industry revenue near $80 billion. It also identified Infineon, onsemi, STMicroelectronics, Texas Instruments, and Analog Devices among major suppliers.

The processor remains the visible center of an AI server. However, usable compute depends on supplying that processor with stable power at the correct voltage.

That makes power efficiency a capacity question, not merely an electricity-bill issue. Lower conversion losses can leave more of a site's constrained power budget available for computing.

The shift pressures data center designers to consider power architecture earlier. It also gives component vendors a chance to capture value that previously sat behind the processor headline.

The Contest Is for the Entire Power Tree

Onsemi is competing to own more conversion stages, while rivals defend their positions with equally broad power portfolios.

The primary contest is not onsemi against Nvidia. Nvidia creates much of the demand through accelerator platforms, while power suppliers compete around those processors.

Onsemi's direct pressure comes from established analog and power-semiconductor companies. Infineon has substantial positions in silicon carbide, gallium nitride, power modules, and data center power.

STMicroelectronics also supplies silicon, silicon carbide, and gallium nitride components. Texas Instruments and Analog Devices bring extensive analog, control, and power-management portfolios.

These companies can pursue the same architectural transition. They can sell individual components, integrated modules, controllers, or reference designs covering several stages.

Onsemi's intended differentiation combines product breadth with internal manufacturing. It argues that silicon, silicon carbide, gallium nitride, and vertical gallium nitride provide options across the full system.

The company is also expanding its analog capabilities through Treo, a 65-nanometer mixed-signal platform. Mixed-signal chips process both continuous analog signals and digital information.

Onsemi says Treo-based products are already sampling. Sampling means selected customers receive early devices for testing before larger commercial production begins.

Its vertical gallium nitride effort adds another route. Unlike conventional lateral GaN devices, vertical GaN moves current through the material vertically.

Onsemi says this structure supports higher current density and voltage handling. It is sampling 700-volt and 1,200-volt devices, with volume production targeted for late 2026.

Those claims remain subject to customer validation. Manufacturing a device at scale does not guarantee that hyperscalers will qualify it for long-lived infrastructure deployments.

The company must also prove that a broad portfolio creates more value than specialized products from several suppliers. Large customers often qualify multiple vendors to protect supply and strengthen purchasing leverage.

That purchasing behavior limits any supplier's ability to own the entire power tree. It also turns reference designs and engineering support into important parts of the competition.

Onsemi nevertheless has a credible opening. Its Power Solutions Group generated $829 million during the quarter, up 19% year over year.

The group represented more than half of total company revenue. Its scale gives onsemi manufacturing experience, customer relationships, and an existing channel for new data center products.

Advanced Solutions Group revenue declined 2% year over year, while Intelligent Sensing Group revenue rose 1%. That contrast reinforces the central role of power products in the quarter.

The competitive question is therefore specific. Can onsemi convert a broad component catalog into higher content per rack before rivals secure the same designs?

An answer will emerge through design wins, production ramps, and sustained margin expansion. Headline growth alone cannot reveal which supplier controls the most valuable sockets.

What the Double-Growth Claim Does Not Show

Onsemi's outlook is notable, but limited disclosure makes its AI data center business difficult to measure independently.

The company reports revenue across automotive, industrial, and other markets. AI data center sales sit inside the other category rather than a separately disclosed segment.

Second-quarter other-market revenue totaled $400 million, up 34% sequentially. That category contains more than AI data center products, so it cannot serve as a direct revenue figure.

Automotive revenue reached $781 million, while industrial revenue totaled $423 million. Together, those established markets still accounted for most of onsemi's quarterly sales.

This mix creates the article's central reversal. AI is becoming onsemi's fastest-growing story, but automotive and industrial demand still determine much of its near-term financial performance.

The company also used relative growth rather than an absolute AI revenue figure. Growth above 100% sounds substantial, but investors cannot calculate its total-company contribution without a starting value.

That does not make the claim misleading. It means readers should distinguish a verified growth rate from an undisclosed revenue base.

The same caution applies to per-rack opportunity estimates. Onsemi's projected increase from approximately $15,000 to $115,000 depends on future architecture, rack density, and supplier share.

The larger estimate concerns a 2030 configuration using 800-volt distribution and racks drawing between 600 kilowatts and one megawatt. Most deployed racks have not reached that configuration.

Data center architecture can also change before those systems become common. Operators are evaluating competing voltage levels, cooling designs, battery systems, and conversion topologies.

A topology is the arrangement of components used to convert and distribute electricity. Different topologies can alter the quantity and type of semiconductors required.

Customers may integrate more functionality into modules. They may also eliminate conversion stages, source parts from several vendors, or negotiate lower component costs at scale.

Onsemi itself describes power-tree compression as an opportunity. Fewer conversion stages can improve efficiency, but compression can also remove sockets available to component suppliers.

The company must capture more value in each remaining stage to offset that possibility. Wide-bandgap devices and integrated controls form part of that strategy.

Execution risk extends beyond product design. Onsemi announced plans to divest two manufacturing facilities during July as part of its Fab Right strategy.

Fab Right aims to improve factory utilization and reduce structural costs. It also requires onsemi to balance internal capacity with external manufacturing during product transitions.

The proposed Synaptics acquisition adds another variable. Onsemi agreed to acquire the connected-compute specialist in an all-stock transaction announced during June.

A deal analysis described the combination as a move beyond power and sensing into compute, connectivity, and control.

That strategy addresses physical AI systems, including robotics and intelligent edge devices. It does not directly validate the data center revenue forecast.

Onsemi must therefore manage a cyclical recovery, factory restructuring, new product ramps, and a major acquisition. Strong second-quarter execution reduces concern but does not remove those demands.

The Earnings Beat Changes Who Faces Pressure

Onsemi's results force power-chip competitors and data center buyers to treat electrical architecture as a strategic layer of AI infrastructure.

The immediate pressure falls on rival suppliers seeking positions in next-generation racks. A customer design win can remain in production for years after qualification.

That long cycle rewards early engineering engagement. Vendors want their switches, controllers, drivers, and modules included before a hyperscaler finalizes its architecture.

Onsemi's raised outlook signals that those decisions are already creating revenue. Its first-quarter disclosure cited multiple chip vendors and leading hyperscalers.

The company did not identify those customers. Readers should not infer a specific commercial relationship from general architecture diagrams or industry partnerships.

Still, the result suggests spending is broadening beyond accelerators and networking. Power delivery is joining cooling, memory, and interconnects as a visible AI bottleneck.

That shift affects enterprise buyers too. A server's processor performance does not determine deployment capacity when a facility lacks sufficient utility power or cooling.

Procurement teams increasingly need to evaluate the entire system. That includes conversion efficiency, redundancy, serviceability, qualification history, and the supplier's manufacturing capacity.

Developers feel the consequences indirectly. Constrained infrastructure can affect cloud availability, deployment schedules, inference capacity, and the cost of operating compute-heavy workloads.

Knowledge workers following AI products should also notice the connection. Model availability depends on physical infrastructure that expands more slowly than software demand.

Google news results often flatten this chain into a contest among accelerator brands. Onsemi's quarter reveals a wider set of suppliers participating in AI spending.

The change does not diminish Nvidia or AMD. Their platforms establish processor road maps that shape voltage, rack, networking, and cooling requirements.

Instead, it expands the competitive field. Every new accelerator generation creates engineering work for companies responsible for delivering electricity efficiently.

Onsemi's third-quarter forecast raises the pressure further. Revenue guidance starts above the second-quarter total, while the gross-margin midpoint also rises.

Before the result, analysts expected second-quarter non-GAAP earnings near $0.72 per share, according to a pre-earnings assessment. Onsemi ultimately reported $0.74.

The difference is modest by itself. Its significance comes from pairing the beat with improving margins, stronger cash generation, and a higher AI data center outlook.

Competitors now need to demonstrate similar operating leverage. Revenue growth becomes less persuasive when it requires disproportionate spending or comes from low-margin commodity products.

Onsemi is promising the opposite pattern. Its long-term model targets revenue growth, gross-margin expansion, and higher free-cash-flow margins together.

Those targets remain management goals rather than achieved results. The second quarter supplies supporting evidence, not final proof.

Three Signals Will Test Onsemi's AI Power Bet

The next test is whether onsemi can turn a fast-growing category into visible, repeatable, and defensible companywide growth.

The first signal is third-quarter performance against guidance. Onsemi expects revenue between $1.65 billion and $1.75 billion, with a non-GAAP gross margin between 40% and 42%.

A result near the upper ends would strengthen the recovery narrative. It would show that second-quarter momentum continued while profitability improved.

A result near or below the lower bounds would weaken the connection between AI demand and broader operating performance. It would also refocus attention on automotive and industrial cyclicality.

The second signal is greater AI data center disclosure. Investors need an absolute revenue figure, a clearer customer mix, or a more specific description of product contributions.

Another percentage-growth statement would confirm direction but not scale. Separate revenue data would show whether AI is becoming material to the overall company.

Readers should also watch whether onsemi continues reporting adoption across several power-tree stages. Broader attachment would support its claim that portfolio coverage raises content per rack.

The third signal is customer qualification for vertical GaN and 800-volt systems. Onsemi targets volume production of its vertical GaN devices in late 2026.

Production timing matters because samples do not create a durable competitive position. Customers must validate reliability, efficiency, packaging, and manufacturing consistency.

Named design wins would strengthen the argument that onsemi can convert technical breadth into revenue. Delays or vague updates would leave more room for Infineon, STMicroelectronics, and other suppliers.

The company's scheduled September analyst day offers another useful checkpoint. Management can explain its AI revenue base, product road map, and assumptions behind the 2030 rack opportunity.

Readers should separate three layers when evaluating those disclosures. Reported quarterly revenue is historical, the 2026 growth outlook is forward-looking, and the 2030 opportunity is a company estimate.

That distinction matters when google news headlines compress all three into a single growth story. The confirmed result is a strong quarter with improving revenue, margins, earnings, and cash flow.

The forward-looking case is more ambitious. Onsemi expects AI data center revenue to more than double as power requirements spread across denser computing systems.

The unresolved issue is competitive capture. Rising electricity demand benefits the category, but it does not guarantee one vendor a dominant share.

Follow the next earnings release, absolute AI revenue disclosure, and late-2026 GaN qualification milestones. Together, those signals will show whether this was an early infrastructure shift or a temporary headline surge.