Intel CPU Shortage Gains Are Real, but Its Foundry Valuation Needs More Proof

Intel is capitalizing on a CPU shortage that helped lift second-quarter revenue 25%, despite unresolved questions about the value of its foundry business.

The immediate picture looks stronger than it did a year ago. Data center sales accelerated, client revenue rose, factory performance improved, and demand exceeded Intel’s available product supply. The company expects shortages affecting substrates, memory, and other components to persist into 2027.

That combination gives Intel more pricing flexibility and better factory utilization. It also strengthens the product business that must finance years of manufacturing investment.

However, the Intel CPU shortage does not automatically validate Intel Foundry as a competitive contract manufacturer. Most foundry revenue still comes from Intel’s own product groups, while external customers contribute only a small portion.

That distinction matters because investors are judging two related but different recoveries. Intel’s x86 products are benefiting from scarce compute capacity. Its manufacturing operation must still prove that outside chip designers will commit meaningful volume.

AMD adds another complication. Its data center business is also growing rapidly, showing that tight supply can help multiple vendors without resolving Intel’s longer-term competitive position.

The central question is therefore not whether the shortage helps Intel today. It clearly does. The question is whether those gains provide enough time, cash flow, and credibility to build an external foundry business that customers trust.

The Intel CPU Shortage Is Now Visible in Revenue

Intel’s latest numbers show that constrained supply has become a measurable earnings advantage, not merely an optimistic management narrative.

Intel reported second-quarter 2026 revenue of $16.1 billion, up from $12.9 billion one year earlier. That represented its strongest annual revenue growth in more than 15 years, according to the company.

Its product segments supplied most of that expansion. Data Center and AI revenue reached $6.3 billion, a 59% annual increase. Client Computing and Physical AI revenue rose 13% to $8.9 billion.

Total Intel Products revenue increased 28% to $15.1 billion. The product organization produced $4.8 billion in operating income before Intel’s corporate allocations and intersegment eliminations.

Intel’s quarterly results also showed a 40.4% GAAP gross margin. That was 12.9 percentage points higher than the prior-year quarter.

The figures support the basic argument behind the Intel CPU shortage story. Customers need more processors than Intel can currently deliver, especially across data centers serving expanding AI workloads.

CPUs remain essential even when GPUs receive most of the attention. They run operating systems, manage data movement, coordinate accelerators, handle networking, and support workloads that never move onto a GPU.

Inference adds another source of demand. Inference is the stage when a trained AI model processes new requests and generates outputs. These systems often require substantial general-purpose compute around their accelerators.

Agentic applications can create repeated exchanges between models, databases, software tools, and external services. Each step adds orchestration work that servers must process.

Intel CEO Lip-Bu Tan connected this transition to demand for the company’s processors during its first-quarter update. He said AI was moving from foundational models toward inference and agentic computing, increasing the need for CPUs.

The company’s second-quarter filing provides firmer evidence than executive commentary alone. Intel said market demand exceeded available product supply because of its factory constraints and broader component shortages.

Higher selling prices also contributed to growth. Intel said most average selling price improvement came from customers choosing premium products. Demand-based pricing actions made a smaller contribution and partly offset higher input costs.

This qualification is important. Intel is not reporting that every percentage point of growth came from explicit price increases. Product mix, supply availability, and stronger demand all influenced the result.

The shortage nevertheless changes Intel’s economics. Scarcity allows the company to prioritize higher-value processors, sell more usable output, and spread manufacturing costs across additional revenue.

Higher factory yields strengthen the effect. Yield refers to the share of chips on a wafer that meet the required specifications. Better yields turn more manufacturing capacity into sellable products.

That creates a favorable near-term cycle. Strong demand absorbs additional output, better yields improve available supply, and premium product mix supports margins.

The original CPU shortage analysis correctly focuses on this operating improvement. Yet the same evidence also reveals why the shortage cannot settle Intel’s valuation debate.

Intel acknowledges that it cannot satisfy all current demand. Every processor it cannot ship represents deferred or lost revenue, while supply allocation can strain relationships with customers receiving fewer products.

Scarcity is valuable only while Intel manages it effectively. If the company expands the wrong capacity, misses a product transition, or sees demand normalize, today’s favorable pricing environment can fade.

For now, the result is clear. Intel enters the second half of 2026 with a much healthier product engine than it had one year earlier.

AI Infrastructure Has Made the CPU Important Again

The shortage reflects a broader change in AI infrastructure, where accelerators need growing amounts of CPU, memory, networking, and packaging support.

The first phase of generative AI spending centered overwhelmingly on GPUs. Cloud providers and model developers raced to secure accelerators for training increasingly large systems.

That framing made traditional processors look secondary. It also encouraged the belief that Nvidia’s gains necessarily came at the expense of companies associated with CPUs.

Actual data center architecture is more complicated. A GPU server still requires host processors to manage workloads, prepare data, operate storage, coordinate networking, and control access to accelerator resources.

The CPU does not replace the accelerator. It enables the rest of the system to function.

The workload mix is also changing. Model training concentrates heavy mathematical work into relatively predictable runs. Production inference must respond to users, retrieve context, call software, enforce permissions, and process varied requests.

These requirements can place more pressure on host processors. Enterprise AI deployments also combine accelerated tasks with conventional databases, analytics, security software, and business applications.

That is favorable for Intel because Xeon remains deeply embedded in enterprise and cloud infrastructure. Organizations can add accelerator capacity while continuing to purchase Intel CPUs for surrounding services.

Intel launched Xeon 6+ during the second quarter and described it as its first server product manufactured with the Intel 18A process. The company is positioning the chip for sustained performance under real operating constraints.

Intel 18A is the company’s newest leading-edge manufacturing node, a collection of transistor and production technologies used to fabricate advanced chips. Its role extends beyond product specifications.

Every successful Intel processor built on 18A provides operating evidence about the manufacturing process. It generates wafer volume, tests factory execution, and gives engineers data for improving yields.

That connection makes the CPU recovery strategically useful. Intel’s product organization can serve as the first large customer for factories that Intel eventually wants external designers to use.

The relationship works in both directions. Better manufacturing can improve Intel’s products, while successful products can absorb the factories’ large fixed costs.

Still, strong demand does not guarantee market-share recovery. AMD reported second-quarter revenue of $11.5 billion, up 50% from the prior year. Its Data Center segment more than doubled to $6.7 billion.

AMD attributed that performance to EPYC CPUs and Instinct accelerators. Its data center results show that the AI compute expansion is not exclusive to Intel.

AMD relies on external manufacturing partners, principally TSMC, rather than operating a leading-edge production network comparable to Intel’s. That model reduces the burden of funding factories but creates its own capacity dependencies.

Intel controls more of its production chain. That can become an advantage during shortages if the company assigns capacity quickly and raises factory output.

The same integration becomes a disadvantage when process delays or weak utilization force Intel to absorb the costs directly. Owning factories magnifies both successful execution and mistakes.

Current demand therefore pressures Intel and AMD in different ways. Intel must decide how much capacity to allocate among data center chips, client processors, and technology ramps. AMD must secure enough outside wafer and packaging supply to meet expanding orders.

Customers face tradeoffs too. Cloud providers need available processors, predictable roadmaps, acceptable energy use, and software compatibility. A shortage can make immediate availability more important than marginal benchmark differences.

That environment helps Intel monetize its installed base. Existing customers can add familiar Xeon systems without redesigning every layer of their infrastructure.

However, purchasing decisions made under scarcity do not always persist after supply improves. Customers may return to performance, efficiency, or total-cost comparisons once availability stops dominating procurement.

This is why the Intel CPU shortage is both valuable and temporary by nature. It gives Intel leverage now, but the company must convert that leverage into durable product preference.

The next test is whether Intel retains demand when buyers regain alternatives. Sustained Xeon adoption would strengthen the recovery more than one year of shortage-driven orders.

Intel Foundry Valuation Still Depends on Outside Customers

A successful Intel product cycle supports the factories, but it does not prove that independent chip designers will choose Intel Foundry.

Intel Foundry reported second-quarter segment revenue of $5.8 billion, up 31% year over year. At first glance, that appears to show a manufacturing business gaining commercial scale.

The composition tells a different story. Intel’s regulatory filing says foundry revenue consists substantially of transactions with its own product groups and other internal operations.

Intel eliminates those internal transactions when calculating consolidated company revenue. They represent useful factory activity, but they are not equivalent to sales won from independent chip designers.

External foundry, assembly, and test revenue totaled $293 million during the quarter. That was substantially higher than $22 million one year earlier, but remained about 5% of reported foundry segment revenue.

The manufacturing unit also recorded a $2.1 billion operating loss. That improved from a $3.2 billion loss in the comparable 2025 quarter, yet it remains a large financial burden.

Intel Products generated $4.8 billion in segment operating income during the same period. More than two-fifths of that amount was offset by the foundry segment’s operating loss before other corporate items.

This is the key conflict behind Intel foundry valuation. Investors can reasonably value a stronger product franchise more highly, especially during a demand expansion.

Assigning additional value to an external foundry opportunity requires another set of assumptions. Those assumptions concern customer wins, production volume, utilization, yields, capital intensity, and eventual margins.

Internal production can demonstrate that a process works. It cannot fully answer whether outside customers consider Intel an attractive manufacturing partner.

External customers face high switching costs. Moving an advanced design to another foundry requires engineering work, compatible design tools, validated intellectual property, production planning, and lengthy testing.

A customer must also trust the foundry’s future roadmap. Chip programs take years, so a designer needs confidence that its manufacturing partner will support later products.

Intel’s history as both a chip designer and manufacturer adds another concern. Potential customers may compete with Intel’s product divisions or worry about operational independence.

The company has reorganized its reporting and operating structure to create clearer boundaries. It has also recruited design-tool companies and packaging partners to broaden its manufacturing platform.

Those changes are necessary, but customer commitments offer stronger proof than organizational design. Foundry economics improve when external clients place repeat orders large enough to fill expensive facilities.

Intel’s own filing draws this boundary plainly. It says substantially all foundry activity currently supports internal manufacturing, while the company aims to develop a more significant external business.

This means internal 18A volume should be treated as process validation and utilization, not as final proof of commercial foundry demand.

Intel has reported external engagements around advanced packaging and future process technologies. Some collaborations involve meaningful technical work, but an engagement is not automatically a high-volume production contract.

The distinction matters most for Intel 14A, the planned successor to 18A and 18A-P. Intel designed 14A from its beginning with external customers in mind.

Intel committed during the second quarter to complete 14A development. It expects internal products to use the process and is advancing manufacturing projects connected to the node.

The company has also said expansion speed will depend on committed demand from Intel’s product roadmap and external design wins. That capital discipline is sensible because advanced factories require years of investment.

Yet it creates a demanding sequence. Intel must develop competitive technology, persuade customers to design for it, build appropriate capacity, achieve acceptable yields, and deliver on schedule.

A CPU shortage helps finance that sequence. It does not remove any of those execution steps.

Intel’s foundry priorities emphasize customer trust, ecosystem support, advanced packaging, and engagement with lead 14A customers. Those are relevant foundations.

Investors still need evidence that such engagement becomes committed production. Without meaningful external volume, Intel Foundry remains primarily an internal manufacturing organization with an external option attached.

That option can have substantial strategic value. The United States wants more leading-edge domestic manufacturing, and customers increasingly consider geographic concentration when planning supply chains.

Strategic importance, however, does not guarantee attractive shareholder returns. A factory can matter to national policy while producing inadequate commercial margins.

Intel foundry valuation therefore needs two separate calculations. One concerns the manufacturing network’s value to Intel’s processors. The other concerns its ability to earn returns from third-party customers.

The first case is improving as Intel ships more products. The second remains under construction.

Better Yields Narrow the Loss, but They Do Not Settle the Case

Intel’s manufacturing progress is economically meaningful, although improving yields and internal volume remain incomplete measures of foundry success.

The foundry operating loss narrowed by roughly $1.1 billion from the prior-year quarter. Revenue grew, manufacturing charges declined, and higher yields improved output.

This is not cosmetic progress. Semiconductor factories carry high fixed costs, so better utilization and more usable dies can materially change unit economics.

Improved cycle times also matter. Cycle time measures how long wafers spend moving through the manufacturing process. Faster movement can release capacity and shorten the period between production spending and customer delivery.

Intel said volume exceeded its expectations because factory yields and cycle times improved. In a supply-constrained market, every added unit has a ready buyer.

That makes the present environment especially favorable for manufacturing optimization. Additional output converts quickly into revenue instead of building unwanted inventory.

Intel 18A is central to the argument. At the start of 2026, Intel began high-volume production of its first products using the process.

The company entered risk production for 18A-P in June. Risk production is an early manufacturing stage used to validate a process and prepare designs before full-volume output.

Intel also said it used ASML’s High NA EUV equipment for high-volume manufacturing of some Core Ultra Series 3 processors. High numerical aperture extreme-ultraviolet lithography enables finer chip features through more precise patterning.

These milestones indicate technical movement, but the investment question concerns repeatability and economics. A process must deliver suitable yields across products and over time, not merely reach a milestone.

Manufacturing costs can remain elevated during a new-node ramp. Equipment expenses, process development, low early yields, and underused capacity all weigh on margins before production matures.

Intel’s detailed second-quarter filing reported a 36% foundry operating loss rate. That was much better than 72% one year earlier, but still far from a self-sustaining business.

The filing also shows why headline foundry revenue deserves careful interpretation. Of $5.8 billion in segment revenue, approximately $5.5 billion was removed through intersegment eliminations.

That internal activity has value because Intel needs manufacturing for its own processors. However, consolidated investors cannot count the same internal transfer as new outside revenue.

A bullish interpretation says internal processors provide the scale needed to mature 18A and prepare Intel Foundry for external customers. Strong CPU demand reduces idle capacity during that process.

A skeptical interpretation says Intel’s product divisions are temporarily masking the absence of large third-party manufacturing commitments. Under that view, scarcity improves utilization without establishing an independent foundry franchise.

Both readings can be true for a time. Internal demand can genuinely improve manufacturing while external commercial proof remains limited.

The risk emerges when investors treat progress in one category as confirmation of the other. Higher Xeon revenue shows product demand. It does not identify which external customer will fill a future 14A factory.

Likewise, a narrower foundry loss demonstrates improving operations. It does not establish the normalized margin of a mature external business.

Intel must also manage capital intensity. The company is increasing investment in equipment, clean-room space, and substrates to support expected growth.

Spending can raise future output, but returns depend on demand remaining available when capacity arrives. Semiconductor expansion decisions occur well before final products generate revenue.

The Intel CPU shortage complicates forecasting because it can pull orders forward. Customers worried about availability may order earlier, hold more inventory, or accept products they would otherwise replace.

Shortages can also encourage competitors and customers to seek alternatives. AMD can gain server share, cloud providers can develop custom silicon, and large buyers can diversify suppliers.

Intel’s 2025 annual filing acknowledged another significant uncertainty. At that point, it had not secured a major external foundry customer for its leading-edge nodes.

The company subsequently committed to completing 14A and reported progress with potential customers. Still, it has not publicly supplied enough contract detail to calculate future external wafer volume reliably.

That gap does not prove failure. Foundry agreements often involve confidentiality, long product cycles, and staged commitments.

It does mean the market must distinguish observable results from future expectations. Observable results include current external revenue, node milestones, yields, and segment losses.

Expectations include unnamed customer decisions, future high-volume ramps, and eventual margins. A credible valuation should assign different confidence levels to each category.

Intel’s recovery deserves recognition because the product and manufacturing trends have improved together. It also deserves scrutiny because a favorable shortage can hide whether those improvements endure under normal supply.

What the Intel CPU Shortage Does Not Prove

Scarcity strengthens Intel’s bargaining position, but it does not prove lasting share gains, normalized margins, or external foundry demand.

The first uncertainty concerns duration. Intel expects component constraints involving substrates, memory, and other critical inputs to continue into 2027.

That forecast supports near-term pricing and product mix. It also highlights the company’s inability to capture every available order.

Supply constraints are not purely beneficial. Intel can lose business when it cannot deliver, even if the units it ships produce better economics.

The second uncertainty concerns customer behavior after supply normalizes. Buyers currently prioritizing availability may resume aggressive price and performance comparisons when capacity becomes easier to secure.

AMD’s growth demonstrates that customers have another x86 option. Its second-quarter Data Center revenue exceeded Intel’s DCAI revenue, although the segments include different product mixes and should not be treated as exact equivalents.

AMD also reported accelerating EPYC demand. That suggests Intel is participating in an expanding market, not necessarily reclaiming every point of competitive ground.

Custom processors create another pressure. Large cloud providers increasingly design silicon for specific workloads, potentially reducing their reliance on standard merchant chips.

General-purpose CPUs remain necessary, but customers can change how many they deploy and which tasks they perform. Architecture choices made during the AI buildout will influence demand long after the shortage ends.

The third uncertainty is product execution. Intel needs competitive processors when customers have more freedom to choose.

A supply shortage can support sales of older or lower-performing inventory. It cannot permanently compensate for weak performance per watt, delayed roadmaps, or inferior total ownership costs.

The fourth uncertainty concerns foundry customer concentration. A single large design win would increase confidence, but it could also make Intel dependent on one buyer.

A stronger signal would combine multiple external customers, repeat designs, growing wafer volume, and improving margins. That pattern would show a business rather than an isolated contract.

The fifth uncertainty concerns 14A timing. Intel said it committed to completing the process and is progressing toward milestones used by potential customers.

Those customers must decide whether to invest engineering resources before Intel can display mature high-volume results. Intel must earn commitments before expanding too far, yet customers want confidence in capacity before committing.

This coordination problem is common in contract manufacturing. It is harder for a new leading-edge supplier competing against an established company such as TSMC.

TSMC offers customers a long record of neutral manufacturing, large-scale execution, and broad design support. Intel must match enough of that experience while also presenting a reason to switch.

Domestic production, advanced packaging, supply-chain diversity, and closer collaboration can form that reason. Each advantage still needs to outweigh switching costs and execution risk.

A final uncertainty concerns valuation itself. A rising share price can reflect better product earnings, reduced financial distress, strategic manufacturing value, or optimism about external foundry success.

Those drivers should not receive the same valuation multiple. Current CPU profits are observable, while future foundry cash flows depend on milestones that have not all occurred.

Investors should also separate accounting effects from operating progress. Intel’s second-quarter GAAP net loss included a large non-operating charge associated with a derivative liability.

Non-GAAP profitability and segment operating performance better illustrate the improving core business. However, non-GAAP measures do not eliminate the cash demands of factory investment.

The balanced conclusion is not that Intel Foundry lacks value. Its manufacturing assets, process development, packaging capabilities, and strategic location clearly have value.

The issue is how much value can be justified today. Internal manufacturing utility is established, while a profitable third-party franchise remains less certain.

That is why Intel foundry valuation needs more proof even as the product recovery improves. The shortage provides time and financial support, but it cannot serve as a substitute for external demand.

Three Signals Will Determine Whether Intel’s Gains Last

The next phase depends on sustained product economics, disclosed external foundry traction, and measurable progress toward profitable manufacturing.

The first signal is Intel’s third-quarter product performance. The company guided for revenue between $15.8 billion and $16.8 billion, with a 41% GAAP gross margin.

Investors should compare actual results with that outlook and examine the sources of any improvement. Revenue supported by higher unit volume carries a different implication from revenue driven mainly by scarcity pricing.

Data Center and AI performance deserves particular attention. Continued Xeon growth would support Intel’s claim that inference and agentic workloads are expanding CPU requirements.

Client results also matter because capacity decisions can shift supply between product groups. Better data center economics are less reassuring if shortages unnecessarily damage Intel’s client position.

The second signal is a significant external commitment for Intel 14A or 18A-P. The customer’s identity matters less than the commercial substance of the agreement.

A useful disclosure would clarify whether the commitment involves a production design, anticipated volume, manufacturing timing, and repeat opportunities. Evaluation work alone offers weaker evidence.

Intel does not need to reveal confidential contract terms. It does need to provide enough information for investors to distinguish testing from an economically meaningful design win.

The third signal is the foundry loss trajectory. The second-quarter operating loss improved to $2.1 billion from $3.2 billion one year earlier.

Further improvement would strengthen the case that yields, utilization, and cost controls are moving in the right direction. Progress should eventually include a larger share of external revenue.

A narrower loss driven only by internal transfers or temporary accounting benefits would carry less weight. Investors should examine both segment expenses and intersegment eliminations.

These signals must be read together. Strong product earnings without outside foundry customers would support Intel as an integrated chipmaker, but not every assumption behind a broad foundry premium.

An external design win without better manufacturing economics would validate customer interest, but not profitability. Lower losses without competitive products or outside demand might reflect cost reduction rather than durable growth.

The strongest outcome combines all three. Intel would sustain CPU demand, convert potential foundry customers into production clients, and narrow manufacturing losses through real operating improvements.

The weakest outcome also has three parts. Supply constraints ease, customers shift orders elsewhere, and external foundry decisions remain limited while capital spending continues.

Today’s evidence sits between those outcomes. Intel’s product recovery is real, manufacturing execution has improved, and the AI buildout is creating more CPU demand than many observers expected.

The verification gap sits inside the external foundry story. Intel still needs customer and financial evidence that stands independently from its own processor volume.

For enterprise buyers, the immediate lesson is practical. CPU availability, supplier diversity, power requirements, and roadmap stability should remain part of infrastructure planning.

For developers, the shift toward inference means application architecture deserves attention beyond accelerator choice. Data retrieval, orchestration, security, and tool execution can determine system-wide compute needs.

For investors, the discipline is simpler. Treat Intel CPU shortage gains as current operating evidence, and treat a profitable external foundry as an outcome still requiring proof.

Watch the next earnings report, the next credible 14A customer commitment, and the foundry loss trend. Together, those signals will show whether Intel is building a durable recovery or enjoying a valuable period of scarcity.