Intel Newsroom Q2 Results Show a Recovery With an $11 Billion Catch

Intel reported $16.1 billion in second-quarter revenue, up 25% despite an $11 billion GAAP net loss that complicates its recovery story. The latest intel newsroom release presents the clearest evidence yet that AI infrastructure demand is lifting Intel’s core processor business. It also shows why one strong quarter cannot settle the debate over Intel Foundry.

The operating numbers were considerably better than the headline loss suggests. Intel generated $1.8 billion in GAAP operating income, compared with a $3.2 billion operating loss one year earlier. Adjusted net income reached $2.2 billion, or $0.42 per diluted share.

Yet Intel remains caught between two financial realities. Its product groups are benefiting from stronger demand, while its manufacturing strategy still consumes capital and produces substantial losses. That tension matters more than the earnings surprise alone.

Nvidia remains dominant in AI accelerators, while AMD competes directly with Intel in server and client processors. TSMC presents a different challenge as the contract manufacturer Intel Foundry wants to rival. Intel must therefore defend its CPU franchise while proving it can manufacture advanced chips economically for itself and outside customers.

The Intel Newsroom Numbers Point to a Real Operating Recovery

Intel’s revenue growth came from stronger products and better factory execution, not a single accounting adjustment.

The quarter ended June 27, 2026, and Intel published its results on July 23. According to the company’s quarterly results, revenue increased from $12.9 billion to $16.1 billion year over year.

That 25% increase was Intel’s strongest quarterly revenue growth in more than 15 years, according to CEO Lip-Bu Tan. It also exceeded the expectations gathered before the release. S&P Global’s earnings preview had cited a $14.4 billion revenue consensus.

The difference was not marginal. Intel delivered about $1.7 billion more revenue than that pre-earnings consensus, while adjusted earnings reached $0.42 per share.

Gross margin provided another important signal. GAAP gross margin rose from 27.5% to 40.4%, an improvement of 12.9 percentage points. Adjusted gross margin reached 41.8%, compared with 29.7% one year earlier.

This improvement means Intel retained much more revenue after direct production costs. Higher factory yields helped because each processed wafer produced more usable chips. Shorter cycle times also allowed Intel to move wafers through its factories faster.

CFO Dave Zinsner said higher yields and improved cycle times created additional product volume. This point separates the Intel Q2 2026 earnings from a demand-only story. Customers wanted more compute, but Intel also became better able to supply it.

Operating expenses moved in the other direction from revenue. Combined research, development, marketing, general, and administrative spending fell 6% on a GAAP basis to $4.5 billion. Intel consequently reported an 11.1% GAAP operating margin, up from negative 24.7%.

The income statement shows $6.5 billion in gross profit and $4.7 billion in operating expenses. That left $1.8 billion in operating income before Intel accounted for investment results, financing items, taxes, and other non-operating effects.

Cash generation also improved. Intel produced $7 billion in cash from operations during the quarter. That cash matters because its manufacturing program requires sustained spending across equipment, facilities, packaging, and process development.

Intel’s client business remained its largest reporting unit. The Client Computing and Physical AI Group generated $8.9 billion, an increase of 13% year over year. The renamed group includes Intel’s PC processors and its growing focus on local AI, robotics, and edge systems.

The more consequential change came from Data Center and AI. That business reached $6.3 billion, rising 59% from the previous year. Its operating income reached roughly $2.5 billion, compared with about $633 million in the second quarter of 2025.

Together, Intel’s product businesses generated $15.1 billion in segment revenue and $4.8 billion in operating income. Both figures show that the core product portfolio delivered more than a temporary revenue bump.

These results do not mean every part of Intel has recovered. They do establish that demand, supply, margins, and product profitability improved together. The next question is whether AI can sustain that combination.

AI Demand Has Put the CPU Back Into the Infrastructure Debate

Intel is benefiting from AI without first displacing Nvidia’s accelerators, because modern AI systems require far more than GPUs.

The early AI infrastructure race often reduced competition to one question: which company could supply the most capable accelerator? That framing favored Nvidia and left Intel defending a mature CPU business.

The market is becoming more complicated. AI clusters use accelerators for intensive model calculations, but they still need CPUs to coordinate data movement, storage, networking, application logic, and system management.

Agentic AI increases that supporting workload. An AI agent is software that can plan and perform a sequence of actions, such as searching databases, writing code, or calling business applications. Those activities create substantial general-purpose computing demand around the model itself.

Intel says this shift is increasing demand for Xeon server processors, custom ASICs, advanced packaging, and foundry capacity. An ASIC is a chip designed for a particular workload, rather than the broad flexibility offered by a CPU.

The company is not positioning Xeon 6+ as a direct replacement for every accelerator. Instead, it wants Intel processors to become the general-purpose foundation surrounding specialized AI compute.

Intel’s own infrastructure demonstrations reflect that approach. One system combined Intel Xeon processors, SambaNova’s dataflow processors, and Nvidia Blackwell GPUs. This mixed architecture recognizes that customers often assemble AI infrastructure from several types of silicon.

The strategy creates an unusual competitive relationship with Nvidia. Intel competes for data center spending, but its processors can also operate inside systems that use Nvidia accelerators. Intel does not need every customer to reject Nvidia before it can sell more CPUs.

AMD presents a more direct processor challenge. Its EPYC server products target the same cloud, enterprise, and AI-support workloads as Xeon. AMD also sells Instinct accelerators, giving customers another path toward combining CPUs and AI chips within one supplier’s portfolio.

Intel’s 59% Data Center and AI growth suggests it captured part of the current infrastructure expansion. Reuters reported that the company’s server-chip demand was strong enough to support an increased investment plan.

However, one quarter does not reveal how much demand came from durable deployments and how much reflected constrained customers advancing orders. Component shortages can cause buyers to secure supply earlier than usual.

Intel also faces the risk that large cloud providers design more of their own silicon. Amazon, Google, Microsoft, and other infrastructure operators increasingly use custom processors and accelerators for selected workloads. Custom chips can reduce dependence on general-purpose suppliers when volumes justify the development cost.

Intel sees that trend as both a threat and an opportunity. Internally designed cloud silicon can reduce demand for standard Xeon products. Yet Intel can pursue the associated ASIC, packaging, and manufacturing work.

That dual position helps explain Tan’s emphasis on CPUs, ASICs, packaging, and wafer manufacturing in the same statement. The company is attempting to earn revenue from several layers of the AI system, even when Intel does not own its central accelerator.

The strategy also extends beyond data centers. Intel said more than 130 customers were adopting or testing its Core Ultra Series 3 and Core Series 3 processors for robotics and edge AI.

Edge AI runs models close to cameras, machines, vehicles, or other data sources. It can reduce latency and limit how much information must travel to a remote data center.

Intel introduced OpenVINO Physical AI to support models involving vision, language, reasoning, and motion control. The framework is intended to help developers deploy these workloads across Intel hardware.

This edge activity remains smaller than Intel’s established PC and server businesses. Still, it gives the company another route into AI spending without competing only for the largest training clusters.

The essential test is profitability. Revenue driven by urgent demand looks less valuable if expensive product ramps and component inflation absorb the benefit. Intel’s improved gross margin says the balance moved in the right direction during Q2, but the foundry results show how much work remains.

Intel Foundry Is Improving but Still Losing Billions

Intel Foundry narrowed its loss and increased output, yet it has not shown that external customers can support the business at scale.

Intel Foundry reported $5.8 billion in segment revenue, up 31% year over year. Its operating loss narrowed from $3.2 billion to $2.1 billion.

That improvement is substantial. A loss exceeding $2 billion in one quarter is still the largest obstacle to treating Intel’s recovery as complete.

Most reported foundry revenue does not represent independent customers choosing Intel over TSMC. Intel’s product divisions pay the foundry segment for manufacturing services, and those internal transactions disappear during consolidation.

Intel reported $5.5 billion in intersegment eliminations for the quarter. That figure was almost as large as the foundry unit’s $5.8 billion in reported revenue.

The accounting does not make the segment irrelevant. It allows investors to see the cost and performance of Intel’s manufacturing organization separately from its product groups. However, the headline revenue cannot serve as evidence of comparable outside demand.

External customer commitments matter because Intel’s factories require extraordinary scale. When factories operate below capacity, fixed costs are spread across fewer usable chips. More customer volume can improve utilization and reduce the cost assigned to each unit.

Intel 18A is central to that effort. The process uses RibbonFET transistors and backside power delivery, two changes designed to improve performance and energy efficiency. It also represents Intel’s attempt to restore credibility after previous manufacturing delays.

The company said Intel 18A-P entered risk production during the quarter. Risk production is an early manufacturing stage when a process begins producing customer designs before full high-volume maturity.

Intel also said a subset of Panther Lake processors had entered high-volume production on Intel 18A. Panther Lake is the code name for a generation of Core Ultra client processors.

These milestones carry more weight than a presentation slide because they connect the process to actual products. Still, production status alone does not disclose manufacturing yield, customer volumes, unit economics, or long-term reliability.

Intel says better yields helped Q2 supply. Yield measures the share of chips on a wafer that function within the required specifications. Small improvements can significantly change costs when each wafer is expensive.

The company did not provide a detailed public yield percentage in its earnings release. Readers should therefore avoid treating management’s directional comments as proof that 18A economics match TSMC’s leading processes.

Intel Foundry also reported $2.1 billion in operating losses while Intel Products produced $4.8 billion in operating income. In effect, the manufacturing business consumed a large portion of the product groups’ operating profit.

That relationship is the main opponent in Intel’s earnings story: improving product economics versus the continuing cost of rebuilding manufacturing.

TSMC operates a different model. It manufactures chips for many customers but does not sell competing PC or server processors under its own brand. That neutrality has helped it build deep relationships with chip designers.

Intel must persuade outside customers that its manufacturing roadmap is dependable and that their designs will receive appropriate protection and support. It must do so while its own product teams remain major factory customers.

Advanced packaging offers another route into those relationships. Packaging connects processors, memory, accelerators, and other components inside a complete device. AI systems increasingly depend on these connections because no single piece of silicon can economically handle every task.

Customers might adopt Intel packaging services before committing their most important processors to an Intel wafer process. That gives the foundry organization a potential entry point, though packaging wins cannot substitute indefinitely for leading-edge wafer volume.

Intel’s manufacturing improvements are therefore meaningful but incomplete. The quarter reduced the probability that foundry execution is simply stalled. It did not prove that Intel has created a self-supporting contract-manufacturing competitor.

The $11 Billion Loss Is an Accounting Shock, Not an Operating Collapse

Intel’s GAAP loss came primarily from a mark-to-market charge tied to government escrowed shares, but dismissing it entirely would also be misleading.

Intel reported an $11 billion net loss attributable to the company, equal to a loss of $2.16 per diluted share. At first glance, that appears incompatible with stronger revenue, wider margins, and positive operating income.

The bridge sits below operating income. Intel recorded about $12.6 billion in “interest and other, net,” compared with a $95 million expense one year earlier.

The major adjustment involved escrowed Intel shares connected to an agreement with the United States government. These shares are held for release as Intel performs under the Secure Enclave program and receives associated funds.

Intel must remeasure the related derivative liability as the value of its shares changes. The Q2 charge represented the change in fair value for shares released during the quarter and shares still held in escrow.

This mark-to-market treatment creates a counterintuitive result. A rising Intel share value can increase the measured liability and produce an accounting loss, even when the underlying business improves.

Intel excluded this effect from its non-GAAP results. It reported adjusted net income of $2.2 billion and adjusted earnings of $0.42 per share.

The adjusted figure offers a clearer view of current operations because the escrowed-share remeasurement does not represent ordinary processor production or customer demand. It also cannot replace GAAP reporting.

Intel’s own non-GAAP explanation warns that adjusted measures should not be considered superior to results calculated under generally accepted accounting principles.

The distinction matters for two reasons. First, readers evaluating Intel’s operating recovery should not interpret the $11 billion loss as evidence that its factories and product groups lost that amount during Q2.

Second, the escrowed shares represent an economic arrangement involving potential dilution and value transfer. Calling the charge non-operating does not make the underlying obligation imaginary.

Intel’s earlier quarterly filing described 149 million escrowed shares that had not been released at the end of Q1. It also explained that some shares depended on further government disbursements.

Investors should consequently track both operating and GAAP results. Operating income shows whether Intel’s products and factories are improving. GAAP results capture financing, equity, and contractual effects that still influence shareholder value.

Cash flow adds a third perspective. Intel generated $7 billion in operating cash during Q2, which supports the claim that the quarter’s business performance was stronger than the net-loss headline.

Yet Intel’s balance sheet also reflects a capital-intensive strategy. Cash and cash equivalents stood at $12.9 billion at quarter-end, while short-term investments were $16.9 billion. Long-term debt reached $48.5 billion.

The company plans to increase investment as it responds to expected AI demand. That decision creates a classic timing risk. Intel must order equipment and expand capacity before it knows exactly how long the demand cycle will last.

Capital spending can produce attractive returns when factories fill with profitable orders. It can damage cash flow when customer demand changes, yields disappoint, or equipment sits underused.

Intel’s $11 billion loss should therefore neither erase the operational recovery nor disappear from the analysis. The useful interpretation sits between those extremes.

Revenue, margins, operating income, and operating cash flow show a business that improved sharply. Foundry losses, debt, government-linked equity obligations, and future capital requirements show that the recovery remains expensive.

Three Signals Will Decide Whether Intel’s Recovery Lasts

The next phase depends on Q3 margins, external foundry demand, and whether data center growth remains profitable.

The first signal is Intel’s third-quarter delivery against its own outlook. The company expects revenue between $15.8 billion and $16.8 billion, with a midpoint of $16.3 billion.

Intel also expects a 41% GAAP gross margin and a 42% adjusted gross margin. GAAP earnings are forecast at $0.31 per diluted share, while adjusted earnings are expected to reach $0.38.

Meeting that outlook would show that Q2 was not simply a burst of delayed shipments or unusually favorable factory output. Results near the top of the range would strengthen the case for sustained AI-related demand.

Margins deserve more attention than revenue alone. Intel will be ramping newer products and increasing investments while dealing with component constraints. If revenue holds but gross margin falls sharply, the company may be buying growth through an unfavorable product mix.

The second signal is externally funded foundry activity. Investors need more than higher segment revenue because internal manufacturing charges dominate the current total.

A significant outside customer committing a leading processor to Intel 18A-P or a later process would validate several claims at once. It would indicate confidence in design tools, yields, capacity, schedules, and Intel’s ability to protect customer information.

The identity of a customer matters less than the scope of the commitment. A small test chip or packaging engagement would not carry the same weight as a high-volume computing product.

Foundry operating losses must also continue narrowing. Better yields can lower unit costs, while more volume can spread the fixed cost of factories across a larger production base.

If losses remain near $2 billion each quarter despite rising internal revenue, Intel’s manufacturing economics will remain under pressure. If losses decline while external revenue grows, the foundry recovery will become much more credible.

The third signal is the composition of Data Center and AI growth. Q2’s 59% increase was the strongest evidence that Intel participates in AI infrastructure spending beyond the accelerator market.

Intel must show that the growth continues without sacrificing operating margin. Product revenue that converts into profit can help finance foundry investments and strengthen the company’s balance sheet.

The comparison with AMD will be especially useful. Continued share competition in server CPUs will reveal whether Intel’s growth came mainly from market expansion or from a more durable improvement in its competitive position.

Nvidia’s platform strategy also remains relevant. Nvidia increasingly integrates accelerators, CPUs, networking, and software into complete systems. Intel’s mixed-hardware approach must remain attractive when customers can buy a more unified platform elsewhere.

Large cloud providers create another test. Their custom chips can become Intel Foundry customers, but they can also replace standard Intel processors. The net effect depends on which side of that relationship grows faster.

For enterprise buyers, the practical issue is supplier choice. Stronger Intel execution could expand available server capacity and reduce dependence on a narrow set of AI hardware providers. Unreliable delivery would send buyers back toward established alternatives.

Developers should watch whether Intel’s software support keeps pace with its hardware. OpenVINO, common AI frameworks, and mixed-accelerator compatibility determine whether available chips can be used without excessive engineering work.

Knowledge workers will feel the results indirectly through AI-enabled PCs, local inference, workplace agents, and cloud service capacity. More efficient competition can improve availability, but supplier spending eventually needs productive customer workloads.

The intel newsroom release establishes a genuine turning point in Intel financial results. The company grew quickly, restored operating profit, expanded margins, and turned factory improvements into greater supply.

It also reported an enormous GAAP loss, continued losing billions inside Intel Foundry, and committed to more investment before the demand cycle is fully proven.

Watch the next earnings release for three answers: Does gross margin hold, do outside foundry customers commit meaningful volume, and does data center growth remain profitable?

Those signals will determine whether Q2 marked the beginning of a durable Intel recovery or simply its strongest quarter inside an unfinished turnaround.