Onsemi’s AI Data Center Growth Powers a Profit Rebound

ON Semiconductor returned to profit as quarterly revenue reached $1.60 billion, giving Google News readers a less familiar view of the AI chip boom.

The company, which now brands itself as onsemi, is benefiting from the electrical infrastructure surrounding AI processors. Its strongest opportunity is not replacing Nvidia or AMD accelerators. It is supplying components that convert and control the power those accelerators consume.

That distinction creates the central tension behind the results. AI data center sales are expanding quickly, but automotive and industrial customers still generate most of onsemi’s business. The earnings improvement therefore tests whether a fast-growing power franchise can reshape a much larger, cycle-sensitive semiconductor company.

What Google News Headlines Leave Out of the Earnings Beat

Onsemi’s second-quarter improvement was broad enough to matter, but AI data centers remain the emerging part of the story.

Onsemi reported results on August 3 for the quarter ending July 3, 2026. Revenue increased 9% from the previous year to $1.6035 billion, according to its quarterly results.

Revenue also rose 6% from the first quarter. That sequential improvement matters because it suggests the recovery was not simply an easy comparison against a weak period.

GAAP net income attributable to the company reached $226.8 million. That was up from $170.3 million one year earlier and reversed the first quarter’s $33.4 million loss.

Diluted GAAP earnings increased to $0.56 per share from $0.41 a year earlier. Non-GAAP earnings reached $0.74 per share, compared with $0.53 in the second quarter of 2025.

The distinction between GAAP and non-GAAP figures is important. Non-GAAP results remove selected restructuring, acquisition, impairment, and other charges to show management’s preferred view of continuing operations.

The company’s GAAP gross margin rose to 38.4% from 37.6% one year earlier. Its non-GAAP gross margin improved to 39.3%, up from 37.6%.

That expansion suggests onsemi earned more gross profit from each revenue dollar despite operating in several competitive semiconductor markets. It also supports management’s claim that product mix and cost discipline are improving.

Cash generation provided another meaningful signal. Cash from operations reached $459.7 million, compared with $184.3 million during the corresponding 2025 quarter.

Free cash flow, which generally measures operating cash after capital spending, reached $425.4 million. The company said that figure was roughly four times the year-earlier result.

Onsemi repurchased $332 million of shares during the quarter. Management said year-to-date shareholder returns represented approximately 105% of free cash flow.

Those figures explain why the report attracted attention beyond a simple revenue beat. Revenue advanced at a single-digit rate, while earnings, cash generation, and margins improved much faster.

CEO Hassane El-Khoury said revenue, gross margin, and earnings exceeded the midpoint of company guidance. He specifically identified strengthening demand across AI-driven applications.

More importantly, he called AI data centers onsemi’s fastest-growing business. The company now expects revenue from that market to more than double during 2026.

That forecast is a company projection, not an independently verified outcome. Onsemi has not disclosed a complete quarterly revenue breakdown for the AI data center business in its earnings release.

The missing segment detail limits what readers can conclude from the headline. Investors can confirm that the overall company improved, but not precisely how much AI contributed to quarterly profit.

This reporting gap does not invalidate the growth narrative. It means the next stage depends on clearer evidence that AI power sales are becoming financially material.

AI Data Center Power Is Becoming a Business, Not a Side Story

The critical change is that power-management chips now participate directly in AI infrastructure spending.

Onsemi entered 2026 with evidence that this business had already moved beyond experimentation. Its 2025 annual report said AI data center revenue exceeded $250 million that year.

The momentum continued during the first quarter. Onsemi said AI data center revenue more than doubled from the prior year and grew over 30% sequentially.

Its first-quarter results attributed that growth to broader adoption across the data center power tree. The company also cited work with multiple chip vendors and leading hyperscalers.

A power tree is the chain of components that converts electricity from the grid into voltages usable by processors, memory, networking equipment, and cooling systems. Every conversion stage can waste energy as heat.

That makes power efficiency commercially significant. A small percentage improvement can reduce electricity use, cooling requirements, and operating expenses across thousands of servers.

The AI infrastructure market has usually been described through graphics processors, custom accelerators, and high-bandwidth memory. Those products perform or feed the mathematical operations behind model training and inference.

Power semiconductors play a different role. They switch, regulate, and convert electricity while protecting costly computing equipment from unstable delivery.

The category includes traditional silicon products alongside silicon carbide and gallium nitride devices. Those materials can operate efficiently under demanding voltage, temperature, and switching conditions.

Onsemi’s portfolio covers several stages of this conversion process. That breadth gives it opportunities to sell more content into each rack without designing the central AI processor.

Its position in the Nvidia MGX ecosystem offers a concrete example. MGX is a modular architecture that lets equipment makers configure different server designs around common building blocks.

Onsemi says its components support every power-conversion stage within that MGX architecture. The relationship places its products near an important server design standard, although it does not guarantee future purchase volumes.

The company also reported platform wins with Great Wall, a Chinese cloud-infrastructure power supplier. Those wins involve EliteSiC products, silicon MOSFETs, and controllers.

A MOSFET is an electronic switch used to regulate current with high speed and precision. EliteSiC is onsemi’s brand for devices built with silicon carbide.

Onsemi separately launched GaNEXUS, a gallium nitride portfolio spanning devices from 40 volts through 650 volts. It targets data centers, robotics, and industrial infrastructure.

These technologies matter because denser AI systems require more electricity inside a confined physical space. Higher density increases the value of efficient conversion and raises the cost of wasted energy.

Industry estimates support the broader opportunity. Omdia expects annual power-chip revenue to rise from about $80 billion in 2026 to $100 billion by 2029.

The same power-chip outlook identified Infineon as the market leader and described a fragmented field of competing suppliers. Onsemi, STMicroelectronics, Texas Instruments, and Analog Devices are among the major participants.

That competitive structure cuts both ways. The market is large enough to support several vendors, but customers can compare technologies, qualify alternatives, and pressure pricing.

Onsemi’s advantage must therefore come from system-level design wins, not merely rising electricity consumption. Supplying more stages of the power tree could make its products harder to replace.

The company’s 2026 projection represents the first test. Doubling AI data center revenue would establish a larger base for 2027, when newer power architectures should move deeper into deployment.

The Real Contest Is Power Content Versus Legacy Cyclicality

Onsemi must prove that AI infrastructure growth can outrun weakness in its established automotive and industrial markets.

The company’s segment results reveal both sides of that contest. The Power Solutions Group generated $829 million during the second quarter, increasing 19% from one year earlier.

That segment also grew 13% from the first quarter. It produced more than half of total company revenue, making it central to the earnings improvement.

The Analog and Mixed-Signal Group reported $545.7 million in revenue. That figure declined 2% year over year and increased only 1% sequentially.

The Intelligent Sensing Group generated $228.8 million. Its revenue rose 7% from the previous year but fell 3% from the first quarter.

This uneven performance supports a more precise interpretation of the report. Onsemi is not experiencing identical demand strength across every product and customer category.

Power is leading the improvement. Analog performance remains restrained, while sensing shows mixed movement between annual and quarterly comparisons.

Automotive and industrial semiconductors also follow different cycles from hyperscale data center equipment. Vehicle production, factory investment, inventory adjustments, and interest rates can affect those markets independently.

AI power demand gives onsemi a new growth engine, but it does not immediately eliminate those exposures. A relatively small business must scale before it can determine consolidated results.

That is why the primary opponent is legacy cyclicality, not Nvidia. Onsemi does not need to defeat the leading accelerator company to succeed.

It needs power content per AI rack to grow quickly enough to offset slower areas. That requires both rising system shipments and a larger component footprint within each system.

The second quarter offers encouraging evidence. Overall revenue increased while gross margin, operating income, net income, and free cash flow all improved.

However, the release does not isolate the gross margin earned by AI data center products. It also does not disclose customer concentration or the duration of major platform commitments.

Those omissions matter because early design wins do not always translate into stable production revenue. Server programs can change, customer schedules can shift, and architectures can favor competing conversion methods.

The company’s reported expansion across the power tree should reduce dependence on one component. Yet broader exposure can increase execution requirements across voltage levels, materials, packaging, and control technologies.

Competition is also more complex than a list of chip vendors suggests. Customers can choose among silicon, silicon carbide, and gallium nitride depending on efficiency, cost, reliability, and production readiness.

Infineon and STMicroelectronics bring established power portfolios. Texas Instruments and Analog Devices have deep relationships across power management and signal-chain applications.

Vicor pursues specialized high-density power architectures. Hyperscalers can also influence component requirements through custom server and rack designs.

No single earnings report settles that competition. Onsemi’s best defense is converting named platform participation into recurring, diversified production orders.

The company’s Power Solutions Group growth shows that the broader power portfolio already has momentum. The unresolved question is how much comes specifically from AI systems.

Google News coverage can compress that distinction into a simple AI-growth headline. The financial statements show a more demanding reality.

AI must become a durable contributor while onsemi continues managing automotive, industrial, analog, and sensing cycles. The reversal becomes structural only when that balance repeatedly improves.

Better Margins Make the AI Story More Credible

Revenue growth matters, but operating leverage is what turns an AI design win into a better business.

Onsemi’s second-quarter GAAP operating income reached $258.6 million. That compared with $193.4 million one year earlier and a $53.4 million operating loss during the first quarter.

First-quarter restructuring and impairment charges contributed to that sequential swing. Readers should not treat the entire improvement as an organic change in demand.

The year-over-year comparison still shows progress. Revenue grew 9%, while GAAP operating income increased by roughly one-third.

GAAP net income attributable to onsemi rose by approximately one-third as well. Diluted GAAP earnings per share increased faster because income grew while the diluted share count declined.

CFO Thad Trent described that pattern as operating leverage. The term means profit changes faster than revenue because fixed costs and product economics amplify incremental sales.

He said year-over-year earnings per share increased four times faster than revenue. The company attributed that result to gross-margin expansion and disciplined cost management.

The cash-flow figures reinforce the point. Operating cash rose 150%, while capital spending fell to $34.3 million from $78.2 million a year earlier.

Lower capital spending helped free cash flow reach $425.4 million. That improvement cannot be credited to AI demand alone, but it strengthens the company’s financial position.

Inventory remains a useful counterweight to the positive figures. Onsemi carried approximately $2.05 billion of inventory at quarter-end, slightly above the level recorded at the end of 2025.

High inventory can support customer service during a recovery. It can also create pricing or write-down risk if demand fails to match production assumptions.

The company must therefore balance growth with manufacturing discipline. Power semiconductors serve long-lived platforms, but customers often maintain multiple qualified sources.

Onsemi has been reshaping its manufacturing network through a strategy it calls “fab right.” The approach aims to match internal production with the technologies where ownership provides an advantage.

The company announced plans in July to divest two manufacturing facilities. Such actions can improve asset efficiency, but transitions also introduce supply, qualification, and execution risks.

Its proposed Synaptics acquisition adds another financial variable. Onsemi agreed in June to acquire Synaptics through an all-stock transaction with an enterprise value of approximately $7 billion.

The Synaptics agreement is intended to combine power and sensing with connectivity, compute, control, and software. The companies expect the transaction to expand their reach into physical AI systems.

Physical AI refers to systems that sense conditions, make decisions, and act in the physical world. Examples include robots, industrial machines, vehicles, and smart devices.

That strategy extends beyond the current data center earnings story. It could diversify onsemi’s AI exposure, but integration will require management attention and regulatory approval.

The transaction also creates tension around capital allocation. Onsemi is repurchasing shares, managing manufacturing changes, and preparing for a significant acquisition at the same time.

Strong free cash flow makes that agenda easier to support. It does not remove the possibility of integration costs, customer disruption, or slower-than-expected synergies.

For AI data center growth to improve the company’s quality, margins must remain healthy as volumes expand. Revenue gained through aggressive pricing would offer less durable value.

The second-quarter margin trend supports management’s case. Future disclosures must show that the improvement survives changes in product mix and customer negotiations.

The Numbers Still Do Not Prove an AI-Led Transformation

Management’s forecast is credible enough to track, but current disclosure cannot establish that AI is driving most of onsemi’s recovery.

The most important skeptical point concerns scale. Onsemi exceeded $250 million in AI data center revenue during 2025, while total annual revenue was much larger.

Even if that business more than doubles in 2026, automotive and industrial demand will continue shaping consolidated performance. AI is becoming material without becoming dominant.

The second issue is attribution. Onsemi reports AI data center activity within broader financial structures rather than as a standalone segment.

Readers cannot independently calculate its quarterly revenue, margin, customer count, or backlog from the earnings release. They must rely on directional statements from management.

The third issue is timing. Design wins usually precede volume production, sometimes by several quarters.

Participation in an ecosystem does not disclose how many systems will ship. A supplier can hold an approved position without capturing every configuration or customer deployment.

The fourth issue is architecture risk. Data center operators are exploring different methods for distributing power across increasingly dense racks.

A transition toward higher-voltage distribution can create opportunities for advanced power semiconductors. It can also shift value between components, materials, and suppliers.

Onsemi offers silicon, silicon carbide, and gallium nitride products, which provides flexibility. Competitors are investing across many of the same technologies.

The fifth issue is customer spending. Hyperscalers are committing substantial resources to AI infrastructure, but those budgets remain sensitive to utilization and financial returns.

A slowdown in accelerator deployments would eventually reach power suppliers. Onsemi would face that effect alongside its existing exposure to vehicle and industrial demand.

The sixth issue is geographic and policy risk. The company cited a platform win involving a Chinese infrastructure supplier.

Technology trade restrictions can change product availability, customer plans, and supply relationships. The earnings release did not quantify the revenue attached to that win.

Finally, the quarter benefited from cost actions and lower capital spending. Those improvements are real, but they differ from revenue growth driven by expanding end demand.

A fair interpretation separates three claims. Onsemi’s total financial performance improved, its AI data center business is growing quickly, and management expects that growth to accelerate.

The available evidence does not prove that AI caused every margin or cash-flow improvement. It also does not establish that the company has permanently escaped semiconductor cyclicality.

This distinction protects readers from both extremes. The report should not be dismissed because AI revenue lacks full segment disclosure.

It should also not be treated as proof that onsemi has become a pure AI company. The business remains diversified, and that diversification creates both resilience and complexity.

The strongest conclusion is narrower. AI power has become large enough to influence onsemi’s growth narrative, and current results give management time to scale it.

Three Signals That Will Test Onsemi’s AI Power Bet

The next quarter must connect management’s AI forecast with measurable company-level results.

The first signal is third-quarter execution. Onsemi expects revenue between $1.65 billion and $1.75 billion.

It projects a GAAP gross margin between 39.9% and 41.9%. The corresponding non-GAAP range is 40% to 42%.

GAAP diluted earnings guidance ranges from $0.79 to $0.91 per share. Non-GAAP guidance ranges from $0.81 to $0.93.

Results near the upper portions of those ranges would strengthen the operating-leverage argument. A revenue increase without continued margin progress would weaken it.

The second signal is more specific AI disclosure. Management should provide quarterly data center revenue, growth, design-to-production conversions, or clearer customer diversification.

Any of those metrics would help separate actual shipments from broad market enthusiasm. Continued reliance on percentage growth alone would preserve uncertainty about the business’s scale.

The third signal is evidence from new power architectures and platform deployments. Investors should watch for production milestones involving MGX systems, Great Wall, GaNEXUS, and higher-voltage solutions.

Repeated wins across different customers would support onsemi’s “across the power tree” strategy. Dependence on one architecture or a few deployments would make growth less durable.

The September 2026 analyst day offers a near-term opportunity for greater detail. Management can define the addressable market, product roadmap, customer mix, and expected margin profile.

That event also gives onsemi a chance to clarify how the Synaptics transaction fits its priorities. The proposed acquisition addresses physical AI, while current earnings momentum centers on data center power.

A coherent strategy should explain how those opportunities share technology, customers, or sales channels. Otherwise, investors may view them as separate narratives competing for resources.

For enterprise buyers and infrastructure teams, these results carry a practical message. AI computing capacity depends on more than access to accelerators.

Power delivery, conversion losses, thermal limits, component availability, and rack design can influence deployment schedules. Those constraints affect the real cost of expanding AI services.

For developers, the connection is less direct but still important. Infrastructure efficiency affects the capacity available for model training and inference.

For technology decision-makers following Google News, the headline is therefore only the starting point. Onsemi has delivered a meaningful profit recovery while its AI data center business accelerates.

The harder test comes next. Watch whether rising guidance becomes reported revenue, whether margins remain above 40%, and whether platform wins become diversified production shipments.

Those three signals will show whether AI power is changing onsemi’s financial identity or simply lifting one favorable phase of another semiconductor cycle.