# Context pack: Intel

> You are a structural analyst. The material below is from PlexusGraph — a knowledge-graph research publication. Reason with the user grounded in it: surface the structure, the feedback loops, the chokepoints and flywheels, and the non-obvious connections. When you make a claim from it, you can point to the sources.

**In one line:** Intel Is Being Asked to Save American Chip-Making — But First It Has to Save Itself

Source: https://plexusgraph.dev/companies/intel

## Brief

*Based on 468 related nodes across 27 research explorations in the semiconductors sector*

---

Intel used to be the most powerful chip company in the world. For decades, it designed and built the processors that ran almost every PC and server on the planet. Then it fell behind. A Taiwanese company called TSMC got much better at making chips than Intel did, and the rest of the industry quietly moved its business there. By the mid-2020s, Intel found itself in a strange position: too big and too strategically important to be allowed to fail, but genuinely struggling to compete.

What makes Intel's situation unusual — and worth understanding — is that its problems and its rescue plan are both being driven by forces largely outside its control. Intel is not fighting for market share in the normal business sense. It is fighting to remain viable while the United States government, its customers, and its suppliers all make bets on whether it can pull off one of the most difficult industrial comebacks in recent memory.

---

## What Intel Actually Does (And Why It's Complicated)

Most technology companies either design chips or manufacture them. Apple designs chips but pays TSMC to build them. NVIDIA designs chips but pays TSMC to build them. Intel, historically, did both — it designed its own chips and ran its own factories. This is called being an "integrated device manufacturer," or IDM.

The problem is that running chip factories at the leading edge of technology is extraordinarily expensive and difficult. TSMC spent decades doing almost nothing else, and it became very good at it. Intel, juggling both design and manufacturing, fell behind on the manufacturing side. By around 2020, the chips Intel was selling were being built on older, less efficient processes than what TSMC was offering its customers.

Intel's current strategy — launched under CEO Pat Gelsinger and continued under Lip-Bu Tan — is to not just fix its own manufacturing but to open those factories to outside customers. The idea is to become America's version of TSMC: a company that builds chips for anyone who wants them. This part of the business is called Intel Foundry.

This is much harder than it sounds.

---

## The Core Problem: A Chicken-and-Egg Trap

Imagine you are trying to get good at baking bread commercially. You need to bake thousands of loaves to learn how to do it consistently well. But customers will only place big orders once you prove you can bake consistently well. You need the orders to get good, and you need to be good to get the orders.

Intel is stuck in exactly this trap, and the research data gives it a name: the Yield-Volume Paradox.

"Yield" in chip manufacturing means the percentage of chips that come out of the factory actually working correctly. Intel's current yield on its newest process — called 18A — is somewhere between 55% and 75%. That means up to 45% of what the factory produces gets thrown away. To run a profitable foundry business, Intel needs to push that number above 70-80%. The way you improve yield is by running the factory continuously at high volume, finding defects, and eliminating them one by one. But to get high volume, you need paying customers. And customers want to see high yields before they commit.

The data identifies one factor making this worse: when Intel was cutting costs, it let a lot of experienced engineers leave. The people who knew how to diagnose and fix yield problems walked out the door. This is rated as the single highest-weight factor amplifying Intel's yield problems — the institutional knowledge is gone and has to be rebuilt.

---

## What Intel Has Going For It

**Its newest chip process is technically impressive.** Independent testing of Intel's 18A process found that it outperforms TSMC's equivalent process on speed and power efficiency — about 25% faster and significantly more power-efficient. This is not marketing language; it comes from third-party measurements. Intel genuinely has a competitive process technology, which was not true a few years ago.

**It owns a machine nobody else has.** TSMC's dominance is partly built on exclusive access to the world's most advanced chip-printing machines, made by a Dutch company called ASML. These machines use a technology called High-NA EUV lithography to etch circuits so small they are measured in atoms. ASML makes only six to eight of these machines per year. Intel received the world's first commercial unit in 2024. That is a physical, durable advantage that cannot be instantly replicated even by a competitor with unlimited money.

**The US government is backing it.** Intel has been designated a national strategic asset. It received an $8.9 billion package under the CHIPS Act, and the US government took a direct equity stake. American policymakers view Intel's success as a matter of national security — the United States does not want to depend entirely on a Taiwanese company for its most advanced chips, especially given tensions with China. This support provides financial cushion and preferential treatment that competitors in the US do not receive.

**A major customer just signed on.** Apple — the most demanding chip customer in the world — has agreed to have Intel manufacture chips for it using the 18A process. Apple does not do this as a favor. If Apple is willing to trust Intel with its most important components, it sends a powerful signal to every other company sitting on the fence about whether to try Intel's foundry. This matters more than the direct revenue.

**A very large anchor customer may be coming.** In April 2026, an alliance was announced between Intel, a company called Terafab, and customers connected to Elon Musk's businesses — Tesla, SpaceX, and xAI. If this results in binding purchase commitments worth roughly $25 billion, it would provide exactly the volume Intel needs to climb its yield curve. The research rates this as the single highest-leverage event in Intel's near-term situation. The caveat: as of April 2026, it is an announcement, not a signed contract.

---

## What Could Still Go Wrong

**The density problem.** Intel's 18A process is faster than TSMC's comparable process, but it fits fewer transistors into the same area — about 31% fewer. For AI training chips, which companies like NVIDIA and the big cloud providers need in enormous quantities, fitting more transistors into less space is more important than raw speed. This means Intel's process advantage is real but points in the wrong direction for the market segment that would generate the most foundry volume. Intel is technically competitive but structurally misaligned with where the money is.

**TSMC is not standing still.** TSMC is building six factories in the United States, committing $165 billion. One of its Arizona factories has actually achieved higher yields than its equivalent factories in Taiwan — proving that high-quality chip manufacturing can happen in America. TSMC brings decades of process expertise and a customer base that already knows how to design chips for its systems. Intel's "national foundry champion" story gets harder to tell as TSMC builds more US soil.

**The JV is complicated.** Intel and TSMC are negotiating a joint venture where TSMC would take a stake in Intel Foundry and contribute its manufacturing process knowledge. This sounds like a solution to Intel's expertise problem. But the US government has to approve it, and there are national security concerns about a Taiwanese company taking ownership of American strategic manufacturing capacity. The same political environment that supports Intel as a national champion also constrains what kinds of deals Intel can make.

**The software side of chips is already lost.** Intel tried to build a competitor to NVIDIA's AI accelerator chips, called Gaudi. It failed, largely because the software ecosystem that developers use to program NVIDIA chips — called CUDA — has nineteen years of development behind it. Developers build their AI systems using CUDA tools, and switching away from NVIDIA requires rewriting enormous amounts of software. Intel had no answer to this. Gaudi has effectively been abandoned as a market competitor. This means Intel's foundry must attract NVIDIA as a customer, but NVIDIA uses TSMC and has no obvious reason to switch.

---

## The Non-Obvious Finding

The most structurally interesting thing about Intel, which is not immediately obvious from its products or financial results, is that its strategic position is almost entirely externally imposed rather than internally chosen.

Intel did not decide to become America's foundry champion because it was the best business opportunity available. It was designated into this role by geopolitical circumstances — specifically, the US government's fear of dependency on Taiwan for chips, and China's accelerating efforts to build its own chip industry. Intel's rescue plan depends on tariffs staying in place, CHIPS Act funding remaining committed, and the political calculus around Taiwan continuing to favor domestic US semiconductor investment. If any of those external supports shift, Intel's business case for the foundry pivot changes significantly.

This is different from a company that is strong because of what it has built. Intel is being asked to be strong because of where it sits on a map.

---

## The 18-Month Window That Decides Everything

All of Intel's critical decisions are converging on the same roughly 18-month window: Can 18A yields cross the commercial threshold? Will the Ohio factory get enough customer commitments to justify full buildout? Will the TSMC joint venture get finalized before Intel runs out of runway? Will the Terafab announcement convert from press release to purchase orders?

None of these are independent. They reinforce or undermine each other. If Terafab commits, yields improve. If yields improve, Apple and other customers commit. If customers commit, Ohio gets funded. If Ohio gets funded and the JV finalizes, Intel has a credible path to the next process node — 14A — and a genuine competitive position in advanced chip manufacturing.

If any of those links breaks, the cascade goes the other way.

---

## Bottom Line

Intel is a company that is genuinely competitive on technical grounds — its 18A chip process works and has been independently validated — but structurally fragile on commercial grounds. Its path to survival runs through a narrow window where volume commitments, yield improvements, a government-constrained joint venture with its main competitor, and a massive anchor customer deal all have to come together in roughly the same 18-month period.

The government wants Intel to succeed. Some important customers are making bets that it will. But the core yield-volume paradox — needing orders to improve, and needing improvement to get orders — has not been solved yet. The Terafab alliance is the most promising single development in the data, but as of April 2026, it remains a commitment to a commitment.

Intel is not a failing company. It is a company attempting a very hard thing, with real assets and real support, on a tighter timeline than anyone would choose.

## Deep analysis

*468 related nodes, 2746 connections across 27 explorations in the semiconductors sector.*

# Intel — Company Brief
**Semiconductors | April 2026**
*Based on 468 nodes, 2,746 connections across 27 research explorations*

---

## Structural Position

Intel occupies the most structurally contested position in the semiconductor industry: it is simultaneously a government-designated national strategic asset, a financially distressed manufacturer, and the swing variable in the US-China technology bifurcation. No other company in the graph bears this combination of dependencies.

The connection topology is diagnostic. Intel's two most connected nodes are **Intel Foundry National Champion Bet** (44 connections) and **Manufacturing Geopolitical Bifurcation Lock-In** (36 connections) — both structural roles imposed by external forces rather than derived from Intel's own competitive strengths. The third most connected node, **TSMC Geopolitical Chokepoint** (33 connections), is a *competitor's* strategic position that Intel is being asked to counterbalance. This pattern reveals Intel's core structural situation: it is not primarily a product company competing in markets, but a geopolitical instrument competing for viability.

The **Intel Foundry 2026-2027 Make-or-Break Window** synthesizes four simultaneous binary decisions: 18A yield crossing commercial threshold, 14A Ohio customer commitment, TSMC-Intel JV formalization, and Terafab volume commitment. All four are converging on the same 18-month window, producing an unusually compressed risk event for a company of Intel's scale.

---

## Key Strengths

**1. Process Technology Credibility — Partially Validated**
The **Intel 18A Process Node** (w=8.5) carries genuine technical merit. The **18A vs N2 Performance-Density Tradeoff** node documents TechInsights measurements: 18A scores 2.53 on performance vs TSMC N2's 2.27 — a 25% performance and 36% power-reduction advantage over Intel's prior node. **Intel Panther Lake 18A Public Validation** and the **AWS AI Fabric Chip Tape-Out** both carry validate edges to 18A at weights of w=9.3 and w=9 respectively, providing external third-party confirmation.

**2. First-Mover Position on High-NA EUV**
Intel holds the world's first commercial **ASML High-NA EUV Angstrom Gate** (Twinscan EXE:5200B), received in 2024. With ASML producing only 6-8 units per year scaling to 20 by 2027-28, Intel's early allocation position in the **ASML High-NA EUV Allocation Race** (w=8.5) represents a durable physical advantage for sub-2nm production. This is a lead that cannot be instantly replicated even with capital.

**3. Terafab Alliance — Structural Volume Solution**
The **Intel-Terafab-Musk Alliance** (w=9, April 7 2026) potentially resolves the company's deepest structural problem. The edge weight on `[potentially_breaks]` Intel Foundry Yield-Volume Paradox is w=9.8 — the highest resolution edge in the Intel subgraph. A committed $25B anchor customer combining Tesla AI5, SpaceX compute, and xAI training workloads on 18A/14A addresses the yield-ramp problem by providing guaranteed volume independent of external customer confidence.

**4. Apple Customer Win — Legitimacy Signal**
The **Apple-Intel 18A Foundry Deal** (w=8.5) carries outsized signaling value beyond its direct revenue contribution. Apple is the world's most demanding and technically sophisticated chip customer. The **IDM Trust Paradox** node explicitly notes that Apple's willingness to fab with Intel partially resolves the trust barrier that has constrained all other fabless customer conversations.

**5. Geopolitical Insurance Premium**
The **Trump Chip Tariff Domestic Differential** provides structural pricing protection for Intel's US manufacturing. The Terafab alliance `[benefits_from]` this tariff structure at w=7.5. Intel's designation as a national foundry champion, backed by a restructured $8.9B CHIPS Act package plus 9.9% equity stake, provides downside support that TSMC Arizona does not fully replicate.

**Durability Assessment:** High-NA EUV allocation is durable (physical constraint, multi-year lead). Geopolitical insurance is contingent on policy continuity. Terafab and Apple deals are recent and unproven at volume — structurally promising but not yet validated.

---

## Structural Vulnerabilities

**1. The Yield-Volume Paradox — Immediate, Partially Internal**
The **Intel Foundry Yield-Volume Paradox** (w=8.5) is the central constraint. Current 18A yields of 55-75% must reach 70-80%+ for commercial profitability, but the primary mechanism for reaching that threshold — volume through-put to pipe-clean fabs — requires customer commitments that themselves require yield proof. The **Intel Foundry Institutional Knowledge Liquidation** node carries a `[amplifies]` edge to this paradox at w=10 — the highest weight edge in the dataset targeting Intel's vulnerabilities — indicating that engineer departures during Intel's cost-cutting phase have compounded the yield-ramp difficulty.

**2. Density Deficit in the AI Chip Market — Structural, Immediate**
The **18A vs N2 Performance-Density Tradeoff** documents that TSMC N2 is 31% denser than Intel 18A (313 MTr/mm² vs 238 MTr/mm²). The **AI Chip Density Imperative** `[amplifies]` this tradeoff at w=9. For AI training chips — where NVIDIA and hyperscaler custom silicon concentrate demand — die area efficiency directly translates to cost per FLOP. Intel's performance advantage is real but structurally misaligned with the market segment that would generate foundry volume at scale.

**3. Operating Loss Trap — Immediate, Partially Internal**
The **Intel Foundry Yield-Volume Paradox** `[causes]` the **Intel Foundry Operating Loss Trap** at w=9, and separately `[amplifies]` it at w=9. The **Intel Foundry $15B Backlog vs $307M Revenue Gap** node measures the scale: a pipeline of announced deals that has not converted to reportable revenue, creating ongoing cash burn at the foundry division without near-term relief.

**4. TSMC Arizona Competitive Pressure — Long-Term, External**
The **TSMC Arizona Yield Inversion** (w=8.5) documents that TSMC's Arizona Fab 21 achieved 92% yield on 4nm/5nm — 4% *higher* than comparable Taiwan fabs — falsifying the narrative that US fabrication cannot match Asian quality. Combined with $165B committed across six planned fabs, TSMC is building a US presence that directly competes with Intel's national champion positioning without Intel's trust barriers or yield problems.

**5. PDK Design Ecosystem Lock-In — Long-Term, External**
The **US Chip Manufacturing "Too Late" Threshold** node identifies PDK lock-in compounding as the primary "too late" mechanism: every quarter TSMC wins AI chip designs deepens customer PDK investment that Intel must subsequently overcome. The **PDK Design Ecosystem Lock-In** `[amplifies]` this threshold at w=8. This is external to Intel's control and grows with elapsed time.

**6. TSMC-Intel JV Structural Ambiguity — Near-Term, Partially Internal**
The **TSMC-Intel Foundry Joint Venture** (w=9) is characterized as a "preliminary agreement" with TSMC taking up to 20% stake and contributing process recipes. However, it is `[constrained_by]` the **Intel Foundry Spinoff Government Veto** at w=7.5 — indicating regulatory and national security review complications that could delay or alter terms. The JV simultaneously `[undermines]` the Intel Foundry National Champion Bet at w=7, revealing a tension between the strategic narrative and the operational reality.

---

## Competitive Dynamics

**Intel vs. TSMC**
The relationship is simultaneously competitive and cooperative — a structurally unusual configuration. TSMC's **Accumulated Process Recipe Moat** (20 connections to Intel) represents decades of proprietary process development that `[amplifies]` Intel's Yield-Volume Paradox at w=8.5 and `[constrains]` the TSMC-Intel JV's High-NA EUV access at w=7.5. Yet the JV `[transfers]` this same moat to Intel at w=9 — making TSMC Intel's primary competitor and its primary rescue mechanism simultaneously.

The **TSMC Arizona GigaFab Strategy** `[undermines]` Intel Foundry National Champion Bet at w=7.5 and `[undermines]` the Yield-Volume Paradox at w=7. TSMC is building a US-domestic narrative that directly competes with Intel's political positioning, backed by demonstrably higher yields. The **TSMC-Intel US Sovereign Duopoly Thesis** offers one resolution: a scenario where both operate complementary roles rather than compete directly — Intel as defense/leading-edge, TSMC as high-volume advanced.

**Intel vs. NVIDIA**
Intel is not NVIDIA's direct competitor but is structurally subordinate to it. The **NVIDIA GPU Monopoly Economics** (24 connections to Intel) and **Nvidia CUDA Ecosystem Lock-in** (15 connections) define the market Intel's foundry must serve but cannot currently win. The **CUDA 19-Year Software Moat** `[undermines]` Intel Gaudi AI Chip Market Collapse at w=9 and `[amplifies]` the Fabless Cliff at w=7.5. Intel's **Gaudi3** accelerator has been explicitly proven out as a failure in the **Intel Gaudi3 Software Ecosystem Collapse** node, used as evidence in the **CUDA Fortress vs Inference Open Market Topology** synthesis. Intel's foundry strategy must attract NVIDIA as a customer — an entity that currently relies entirely on TSMC and has no evident incentive to diversify.

**Intel vs. China (SMIC/Huawei)**
The **China Shenzhen EUV Prototype** (w=8.5, December 2025) is identified as "the single biggest threat to the entire Western semiconductor export control strategy." A functional domestic EUV prototype built by Huawei and SiCarrier Technologies `[undermines]` the ASML High-NA EUV Allocation Race at w=8 — the same race in which Intel holds first position. If China closes the EUV gap, the bifurcation thesis that underlies Intel's foundry premium erodes. The **China Semiconductor Self-Sufficiency Drive** (15 connections to Intel) and the **China 15th FYP Digital Economy Pivot** both compound this pressure by shifting China's strategy from import substitution to deployment-scale data flywheel advantages, reducing the technology gap's strategic relevance.

---

## Regulatory Exposure

Intel's regulatory exposure is primarily *favorable* rather than constraining — a structural differentiation from most companies in the graph.

**Supportive Regulatory Environment:**
The **Trump Chip Tariff Domestic Differential** creates pricing advantages for US-manufactured chips. The CHIPS Act $8.9B package plus equity stake represents active government subsidy. The **US Defense Foundry Dependency** node, partially resolved by the Terafab alliance at w=7, reflects continued DoD investment in Intel as a domestic source for defense-grade semiconductors. These subsidies reduce Intel's effective cost of capital for fab construction.

**Constraining Regulatory Factors:**
The **Intel Foundry Spinoff Government Veto** constrains the TSMC-Intel JV at w=7.5 — any structural transaction that could transfer fab control to a foreign entity faces national security review. This limits Intel's strategic flexibility precisely at the moment when the JV would be most valuable. The **US BIS Export Control Ratchet** `[controls]` the ASML EUV Monopoly at w=9 — the same export control regime that blocks China's access to EUV benefits Intel's relative position but also creates reciprocal constraints on Intel's ability to sell products or services to Chinese customers.

**Comparison to Peers:**
TSMC faces similar national security review exposure (the **TSMC Geopolitical Chokepoint** implies regulatory leverage in both directions), but TSMC's regulatory relationship with the US is more passive — it receives subsidies without facing the spinoff/control constraints Intel does because TSMC does not carry Intel's IDM status. NVIDIA faces export control exposure on H100/B200 sales to China but has no manufacturing regulatory dependencies.

---

## Strategic Leverage Points

**1. Terafab Volume Commitment → 14A Ohio Decision**
The highest-weight resolution edge in the dataset: Terafab `[potentially_breaks]` Yield-Volume Paradox at w=9.8, and `[enables]` Ohio 14A Binary Decision at w=10. If Terafab volume commitments are formalized before H2 2026, they simultaneously resolve the yield ramp problem (volume), validate 14A customer demand (Ohio decision), and strengthen Intel's negotiating position in the TSMC JV. This is the single action that addresses the most constraints simultaneously. The risk: Terafab is an April 2026 announcement; conversion to binding purchase agreements has not been confirmed in the data.

**2. TSMC JV Finalization — Recipe Transfer**
The JV `[resolves]` Yield-Volume Paradox at w=9 and `[transfers]` TSMC Accumulated Process Recipe Moat at w=9. If finalized before the **2026 H2 Customer Commitment Cliff**, it provides Intel fabs with TSMC process expertise that could accelerate 18A yield to commercial threshold. The JV `[influences]` the Make-or-Break Window at w=8.5. The political constraint (Government Veto, w=7.5) makes this the highest-leverage but also highest-friction path.

**3. 18A Yield Improvement → PDK Lock-In Reversal**
The **Fab Yield Learning Curve Economics** node is the mechanism by which volume converts to yield improvement. Every 7% monthly yield improvement reported narrows the window during which TSMC's PDK lock-in advantage compounds. Intel's ability to accelerate this curve — through both Terafab volume and TSMC recipe transfer — is the primary counter to the "too late" thesis.

**4. Density Gap Mitigation via Packaging**
The **CoWoS Advanced Packaging Chokepoint** and **Advanced Packaging** nodes suggest a path around the 18A density deficit: advanced packaging stacks multiple lower-density dies to achieve effective density comparable to monolithic high-density chips. Intel's control of its own advanced packaging infrastructure (Intel Foundry provides both logic and packaging) offers a potential architectural workaround for AI chip customers who cannot absorb TSMC's CoWoS wait times. This is not explicitly mapped as a leverage point in the data but is implied by the packaging chokepoint structure.

**5. Defense Market Segmentation**
The **US Defense Foundry Dependency** node offers a segment where Intel's national champion status is not a narrative but a procurement requirement. Defense customers are structurally blocked from TSMC (foreign entity) and cannot use SMIC (adversary). Intel's 18A performance advantage (speed over density) aligns better with defense application profiles than commercial AI training. Concentrating initial foundry capacity on defense workloads would generate revenue, avoid the TSMC competitive comparison, and maintain DoD subsidy flows without requiring Intel to win commercial AI chip market share in the near term.

---

## Open Questions

**1. Terafab Binding Commitment Timeline**
The April 7, 2026 announcement is present in the data but the contractual structure — whether volume commitments are binding, phased, or conditional — is not specified. The w=9.8 resolution edge is contingent on this distinction. An MOU without purchase commitment does not resolve the Yield-Volume Paradox.

**2. TSMC JV Government Veto Conditions**
The **Intel Foundry Spinoff Government Veto** constrains the JV at w=7.5 but the specific trigger conditions are not documented. Whether TSMC's 20% stake crosses national security review thresholds, and on what timeline a review would conclude, is absent from the graph.

**3. 18A Current Yield Trajectory**
The data states yields are "currently 55-75% (improving ~7%/month)" as of the Make-or-Break Window node, but this is a range with significant uncertainty. The commercial threshold of 70-80%+ could already be reached within the lower bound or remain 3-6 months away at the upper bound. The actual yield curve slope determines whether the H2 2026 cliff is survivable without Terafab volume.

**4. Panther Lake Volume Ramp**
The **Panther Lake 18A Self-Validation Loop** validates 18A and enables the Apple deal, but Panther Lake is Intel's own product — self-validation for an external foundry strategy has limits. Whether Panther Lake production volumes are sufficient to meaningfully advance the yield learning curve for external customer specifications is not addressed.

**5. Gaudi Recovery Path**
The **Intel Gaudi3 Software Ecosystem Collapse** is treated as a closed outcome in the CUDA Fortress analysis. Whether Intel has a credible path to recover GPU market share (through Gaudi4 or successor), or has implicitly conceded the accelerator market to focus entirely on foundry, is not resolved in the data. This matters because a successful accelerator product would provide internal foundry volume independent of external customers.

**6. 14A Technical Readiness**
The **Intel 14A High-NA EUV Node** depends on the Ohio facility and ASML High-NA EUV availability, but 14A PDK maturity is described only as "PDK 0.5 currently at customer evaluation" — an early development stage. Whether 14A will achieve competitive technical specifications by the time Ohio capacity decisions are required is not documented.

**7. Institutional Knowledge Recovery**
The **Intel Foundry Institutional Knowledge Liquidation** carries the highest weight edge targeting Intel's vulnerabilities (w=10 to Yield-Volume Paradox). Whether Lip-Bu Tan's engineering culture reset (w=8.5 targeting the Make-or-Break Window) has reversed the talent departure trend, and on what timeline recovered expertise translates to yield improvement, is absent from the data.

---

*Brief compiled from graph corpus. All node weights and edge weights as reported. Figures current to April 2026 data snapshot.*