PHLX Semiconductor Sector Index (SOX): Companies, Methodology and Research

What is the PHLX Semiconductor Sector Index (SOX)?

The PHLX Semiconductor Sector Index, known by its ticker symbol SOX, is a modified market-cap-weighted benchmark of 30 U.S.-listed semiconductor and semiconductor-equipment companies published by Nasdaq. It is the most widely cited measure of the semiconductor industry's collective equity performance and serves as the underlying index for several ETFs and options contracts.

SOX is constructed around two ICB (Industry Classification Benchmark) classifications: Semiconductors and Production Technology Equipment. This dual classification is a deliberate design choice: it ensures the index captures the entire semiconductor supply chain rather than only chip designers or only chip manufacturers. A pure-design company like NVIDIA and a pure-equipment company like ASML can both be index members because both ICB categories are in scope.

The modification in "modified market-cap weighted" refers to concentration caps applied at each quarterly rebalance. The largest component is capped at 12% of the index, the second at 10%, the third at 8%, and all remaining components at 4% each. Between rebalances, weights drift with price movements, so the caps describe target weights at rebalance dates rather than live weights at all times.

Key facts as of September 4, 2026: 30 component companies. Modified market-cap weighted with concentration caps of 12%, 10%, 8%, and 4%. Annual reconstitution effective after the third Friday in September. Quarterly rebalancing in March, June, September and December. Covers the full semiconductor value chain: design, foundry, equipment, materials, packaging, test and IP licensing.

Why SOX matters for investors

Semiconductors occupy a uniquely upstream position in every digital technology cycle. Before an AI model runs inference, before a smartphone renders a video, before an electric vehicle processes sensor data, a semiconductor must execute the instruction. That upstream position makes the semiconductor industry both a leading indicator of broader technology demand and a concentrated lens on where technology investment dollars are flowing.

SOX captures this dynamic across multiple layers simultaneously. The index includes AI accelerators (NVIDIA, AMD) whose revenue reflects the most direct monetization of AI model training and inference demand. It includes memory companies (Micron) whose pricing and volumes are often the first signal of downstream demand inflection. It includes wafer-fab equipment companies (Applied Materials, Lam Research, KLA) whose order books lead semiconductor production capacity by 12 to 18 months. And it includes analog and embedded companies (Texas Instruments, Analog Devices, NXP) whose revenue reflects industrial, automotive and consumer electronics demand cycles rather than cloud capital spending.

The practical question SOX helps an investor answer is not simply "is the semiconductor sector up or down?" The more useful question is: which part of the semiconductor stack is capturing the economics of the current technology cycle? An AI boom that benefits NVIDIA and AMD may lag in reaching memory, equipment or analog companies by several quarters. Reading SOX as a single number misses that internal rotation. Reading its subsector composition reveals it.

SOX is also relevant because it is the standard reference for semiconductor options market participants. Options on QQQ or SPY give broad market exposure; options on SOX constituents or SOX-tracking ETFs give targeted semiconductor exposure. Understanding what the index includes, how it is weighted and when it reconstitutes is foundational research for anyone building semiconductor positions or using derivatives to express a view on the sector.

The 30 SOX companies (September 2026 working set)

The table below reflects the pre-reconstitution basket as of September 4, 2026. The annual reconstitution with an effective date after the third Friday of September 2026 may add or remove members. Verify current membership against Nasdaq Global Index Watch before relying on this basket for investment decisions. The archetype column reflects the primary value-chain role, not a complete business description.

Source: September 4, 2026 pre-reconstitution basket. Annual reconstitution effective after third Friday in September 2026. Current constituents and weights are published at Nasdaq Global Index Watch.

SOX eligibility and methodology

To be eligible for SOX, a security must satisfy all of the following criteria simultaneously. Meeting some but not all does not qualify a company for inclusion. Nasdaq reviews the eligible universe annually using a reference date of the last trading day of July.

Classification requirement

The company must be classified under the ICB system as either Semiconductors (ICB subsector 9576) or Production Technology Equipment (ICB subsector 9572). This classification is applied by the index provider, not self-reported by the company. A technology company that uses semiconductors extensively, such as a cloud hyperscaler, does not qualify. A company that designs chips but also has substantial software revenue would need to meet the ICB classification threshold. The dual-ICB scope is what distinguishes SOX from a pure chip-design index and from a pure equipment index.

Listing requirement

Securities must be listed on Nasdaq Eligible Exchanges. A July 6, 2026 methodology update broadened the listing requirement from a Nasdaq-specific exchange to the broader Nasdaq Eligible Exchange category. This change allows NYSE-listed securities such as TSMC (TSM) and Applied Materials, which are listed on Nasdaq, to qualify under the updated language. Non-U.S.-headquartered companies with a U.S. listing, such as ASML and TSMC, are eligible; the requirement is a U.S. listing, not U.S. incorporation or U.S. revenue concentration.

Market capitalization and liquidity

The minimum market capitalization threshold is $100 million. On the liquidity side, a security must have traded at least 1.5 million shares in each of the six calendar months prior to the annual reference date. This liquidity screen excludes smaller-cap or thinly traded semiconductor companies that technically meet the ICB classification but would create index replication difficulties.

Seasoning and options requirements

A newly listed company must have been trading for at least three months before the reference date to be eligible for the annual reconstitution. The three-month seasoning period prevents a recent IPO from entering the index immediately after listing, which could introduce unnecessary concentration risk from a company without an established trading history. Additionally, listed options must be available on the security. The options requirement reflects SOX's dual role as an equity benchmark and a derivatives market reference.

Index weighting methodology

SOX uses a modified market-cap weighting system rather than pure market-cap weighting. The difference is important for understanding how the index behaves when one or two companies become very large relative to the rest of the basket.

At each quarterly rebalance, Nasdaq applies the following concentration caps:

• The company with the largest weight in an uncapped calculation is capped at 12%.
• The company with the second largest weight is capped at 10%.
• The company with the third largest weight is capped at 8%.
• All remaining companies are each capped at 4%.

After the caps are applied, any excess weight above a cap is redistributed proportionally among uncapped components. This redistribution can cause some mid-tier components to carry higher weights than their raw market capitalization would imply. The practical effect is that SOX cannot become as top-heavy as an uncapped market-cap index. When NVIDIA's market cap grew dramatically through 2024 and 2025, an uncapped SOX would have given it a weight of 25% or more; the modified construction holds its target weight at 12% at rebalance dates.

Rebalance vs. reconstitution

These two events are frequently confused. Reconstitution is the annual event that changes which companies are in the index: it adds new eligible entrants and removes companies that no longer qualify or have been displaced. Rebalancing is the quarterly event that reapplies the weighting caps to the existing member list without changing composition. Rebalance reference dates are the last trading days of February, May, August, and November; effective dates are the market open on the first trading day of March, June, September, and December respectively. After each rebalance, weights again drift with daily price movements until the next rebalance date.

A note on data access

Real-time and historical index weights are proprietary data licensed through Nasdaq Global Index Watch. Publishing specific current weights without a Nasdaq data license is a terms-of-service violation. This page presents the methodology and the constituent list, not licensed weight data. Investors who need precise weight data for portfolio construction or risk management should access it directly through a Nasdaq data subscription or a licensed data vendor.

Semiconductor value-chain subsectors in SOX

SOX is not a homogeneous collection of chip companies. The 30 members span the full semiconductor supply chain, and each subsector has distinct revenue drivers, margin structures, cycle timing and competitive dynamics. Understanding which subsector a company occupies is the first step in understanding what drives its fundamentals.

Compute and AI accelerators

NVIDIA (NVDA) and AMD (AMD) design the high-performance GPUs and data-center accelerators that dominate AI training workloads. Revenue in this subsector is directly tied to capital spending by hyperscale cloud providers and enterprise AI build-outs. Gross margins in chip design tend to be high because the products are fabless (manufacturing is contracted out), but the revenue is lumpy and can be highly sensitive to data-center capex cycles.

Custom silicon and networking

Broadcom (AVGO) and Marvell (MRVL) design custom application-specific integrated circuits (ASICs) for hyperscaler networking and storage, alongside merchant silicon for switches and routers. Custom ASIC revenue is often more predictable than merchant chip revenue because it is tied to long-term partnerships with specific hyperscalers. Networking connectivity becomes more valuable as AI cluster sizes grow and the communication overhead between accelerators increases.

Memory

Micron Technology (MU) manufactures DRAM and NAND flash. Memory is a highly cyclical commodity business: pricing can swing 50% or more across cycles, and profitability is deeply sensitive to supply-demand balance. High-bandwidth memory (HBM), a specialized DRAM stack for AI accelerators, has become a differentiated product with better margins than commodity DRAM and represents an important shift in Micron's product mix.

Foundries

TSMC (TSM) and GlobalFoundries (GFS) manufacture chips designed by fabless customers. TSMC dominates advanced process nodes (3nm, 2nm); GlobalFoundries focuses on mature and specialty nodes for automotive, defense and analog applications. Foundry revenue is driven by utilization rates, process-node transition timing and customer concentration. TSMC's customer list is a proxy for who is winning in chip design.

Wafer-fab equipment

Applied Materials (AMAT) and Lam Research (LRCX) supply the deposition, etch and chemical mechanical planarization equipment used inside semiconductor fabs. Equipment orders lead actual chip production by 12 to 18 months because fabs must be built and equipped before wafers can be processed. Equipment company revenue is therefore a leading indicator of future semiconductor supply capacity.

Lithography

ASML (ASML) is the sole supplier of extreme ultraviolet (EUV) lithography machines, which are required to pattern the most advanced process nodes. ASML occupies a structurally monopolistic position in EUV; no credible alternative supplier exists. This creates a different risk/return profile from other equipment companies: high pricing power and long order lead times, but also concentration in a single product category and significant geopolitical export control exposure.

Process control and metrology

KLA Corporation (KLAC) and Nova Ltd. (NVMI) supply inspection and measurement equipment used to detect defects and verify dimensions during semiconductor manufacturing. As process nodes shrink, the number of inspection steps per wafer increases, making process control equipment more indispensable per advanced chip manufactured.

Analog, power and embedded

Texas Instruments (TXN), Analog Devices (ADI), Monolithic Power Systems (MPWR), onsemi (ON), NXP Semiconductors (NXPI) and Microchip Technology (MCHP) serve industrial, automotive, medical and consumer markets. Analog chips convert real-world signals (temperature, pressure, sound, light) into digital data. Revenue is less tied to AI capital spending and more tied to automotive production volumes, industrial investment cycles and consumer electronics demand. These companies often carry high margins and long product life cycles.

Wireless and RF

Qualcomm (QCOM), Skyworks Solutions (SWKS), Qorvo (QRVO) and MACOM Technology Solutions (MTSI) supply radio-frequency components for mobile devices, base stations and defense electronics. Smartphone RF content has been a mature, competitive market; MACOM and MTSI have shifted toward optical and microwave defense applications as smartphone volumes plateaued.

Photonics and optical

Coherent (COHR) supplies optical components for data-center interconnects, telecom networks and industrial lasers. As AI clusters grow in size and require higher-bandwidth connections between racks and facilities, optical interconnect becomes a critical bottleneck, which is why Coherent's revenue correlates with data-center spending even though it does not sell semiconductors in the traditional sense.

AI connectivity

Astera Labs (ALAB) and Credo Technology (CRDO) supply connectivity semiconductors and active electrical cables (AECs) that connect accelerators, memory and CPUs within AI server racks. This subsector is among the most direct AI infrastructure beneficiaries below the accelerator layer, serving the disaggregated rack architectures used in large-scale GPU clusters.

Semiconductor IP

Arm Holdings (ARM) licenses processor architecture IP used in virtually every mobile chip and an increasing share of server and automotive chips. Rambus (RMBS) licenses memory interface IP and sells semiconductor chips. IP licensing creates recurring revenue streams with high operating leverage and does not require semiconductor manufacturing, which gives these companies different margin profiles from chip designers or equipment makers.

Test equipment

Teradyne (TER) supplies automated test equipment used to verify chip functionality after manufacturing. Test equipment demand tracks advanced chip complexity: more transistors per die require more test time and more sophisticated test systems.

Semiconductor materials

Entegris (ENTG) supplies specialty chemicals, filtration systems and handling equipment used in semiconductor fabrication. Materials companies sit at the chemical and physical layer below equipment, supplying the consumables and purification systems that fabs require at every process step. Revenue is tied to wafer starts at operating fabs rather than to new fab construction.

AI exposure: how to classify SOX companies

A common mistake is treating "AI exposure" as synonymous with "sells AI accelerators." A company can be a significant beneficiary of the AI infrastructure build-out without designing a single GPU. Classifying SOX members by their position in the AI value chain produces a more useful map than classifying them by whether they appear in AI headlines.

The relevant AI exposure categories for SOX companies are:

• Direct AI compute: Companies whose revenue directly tracks AI training and inference workloads. NVIDIA and AMD are the clearest examples. Revenue growth here is tightly correlated with hyperscaler AI capex but can also be volatile if capex cycles slow or if a new architecture reduces compute requirements per workload.
• AI memory: Micron's high-bandwidth memory business. HBM is a specialized product required by GPU clusters; commodity DRAM and NAND have a weaker AI correlation. The distinction between Micron's HBM revenue and its commodity memory revenue matters more than the aggregate revenue figure for AI exposure analysis.
• AI networking and interconnect: Broadcom and Marvell for custom silicon and merchant network switching; Astera Labs and Credo for connectivity within racks. As AI clusters scale from hundreds to tens of thousands of accelerators, the bandwidth requirements between chips grow faster than the compute itself, making this layer increasingly critical.
• AI manufacturing infrastructure: ASML, Applied Materials, Lam Research and KLA benefit from AI-driven demand for advanced process nodes, particularly 3nm and 2nm logic nodes used for accelerator production. This is an indirect benefit with a longer lag; fab equipment orders today translate into production capacity 12 to 24 months later.
• AI process complexity: Nova and KLA for metrology, Entegris for advanced materials. These companies benefit because each new process node at which AI chips are manufactured requires more inspection steps and tighter chemical purity tolerances.
• AI power delivery: Monolithic Power Systems and onsemi for power management. A large-scale GPU cluster can draw tens of megawatts of power; managing that power efficiently at the chip, rack and facility level creates demand for sophisticated power management semiconductors.

Companies like Texas Instruments, NXP, Microchip, Qualcomm, Skyworks and Qorvo have limited direct AI revenue correlation. Their revenue is driven by industrial production, automotive shipments and smartphone unit sales. These companies serve as a distinct economic signal within SOX, reflecting the non-AI semiconductor demand cycle and providing an embedded contrast to the AI-driven components.

The practical implication is that SOX as an index simultaneously tracks two partly independent demand cycles: an AI-infrastructure cycle and an automotive-industrial cycle. These cycles can move in different directions at the same time, which is why SOX can underperform during periods when AI spending is strong but automotive or industrial demand is weak, or vice versa.

Semiconductor cycle map

Understanding where each SOX company sits in the production sequence matters because signals propagate through the supply chain with meaningful lags. The stylized sequence is:

1. End demand: consumer electronics purchases, enterprise IT budgets, automotive production schedules, data-center capex plans.
2. Chip design decisions: fabless designers (NVIDIA, AMD, Qualcomm, Broadcom, Marvell, ARM licensees) adjust their wafer start forecasts and submit purchase orders to foundries.
3. Foundry utilization: TSMC and GlobalFoundries adjust production plans, which determines near-term memory and logic supply.
4. Fab equipment orders: if foundries decide to expand capacity, they order equipment from Applied Materials, Lam Research, KLA and ASML. These orders typically lead actual production capacity additions by 12 to 18 months.
5. Semiconductor materials: Entegris and other materials suppliers fulfill purchase orders tied to current wafer start rates at operating fabs.
6. Test: Teradyne's test equipment revenue tracks chip production volumes with a short lag.
7. System deployment: finished chips are integrated into servers, phones, vehicles and industrial equipment and deployed to end customers.

Several overlays complicate this linear sequence. Inventory cycles: end customers and distributors routinely overbuild or underbuild chip inventories relative to actual consumption, creating demand signals that lead or lag real end-use demand by two to four quarters. The 2022-2023 semiconductor correction was largely an inventory digestion event, not a collapse in end demand, and equipment company order rates fell earlier than foundry revenues because equipment spending is the most deferrable item in the supply chain. Process-node transitions: every time a major foundry transitions to a new process node, equipment intensity per wafer increases because new deposition, etch and inspection steps are added. ASML's EUV demand is a function of how quickly the industry migrates to advanced nodes, not just of how many wafers are started total. Memory cycles: DRAM and NAND pricing follows a distinct supply-driven cycle that is partly independent of logic demand; Micron's profitability can move in the opposite direction from NVIDIA's during periods when logic is in short supply but memory is oversupplied. Geopolitical constraints: export controls on advanced semiconductor equipment and chips to specific geographies create distortions in where fab capacity is built and which equipment companies can grow revenue in which markets.

An investor tracking SOX should map each constituent to its position in this cycle and maintain awareness of the relevant leading indicators for each subsector rather than relying on the index's aggregate price movement as the only data point.

Reconstitution calendar and basket freshness

SOX reconstitutes annually. The process begins with an eligible universe review using the last trading day of July as the reference date. Nasdaq applies the eligibility criteria described above, ranks the qualifying securities by market capitalization, and determines the new composition. New components enter the index and removed components exit at market open on the first trading day after the third Friday of September.

For September 2026, the reconstitution effective date falls after September 19, 2026. As of September 4, 2026, the 30-company basket listed on this page represents the pre-reconstitution composition. After the effective date, the basket may differ: companies that no longer meet eligibility criteria will be removed, and newly eligible companies with sufficient market cap may be added. The total number of components does not have to remain exactly 30 if the eligible universe produces a different count, though SOX has historically targeted approximately 30 components.

Swoopr will update the constituent table on this page following each annual reconstitution. For historical baskets, the constituents/history sub-page (linked below in the Research and Data section) will preserve each year's composition for reference. The methodology ensures the index always reflects the currently eligible semiconductor universe rather than a fixed legacy roster, which is why several companies that were SOX members in 2015 are no longer in the index today.

SOX vs. related benchmarks

SOX vs. Nasdaq-100 (NDX)

The Nasdaq-100 contains the 100 largest non-financial Nasdaq-listed companies, many of which are technology companies including several SOX members. The critical difference is breadth and focus. NDX includes large-cap names from consumer discretionary, healthcare and other sectors that have nothing to do with semiconductors; SOX is sector-specific by design. When an investor wants to understand how the semiconductor industry is performing relative to broader technology or the overall market, comparing SOX to NDX is more analytically useful than comparing two broad indexes. Conversely, NDX provides diversification that SOX deliberately removes.

SOX vs. S&P 500 Information Technology Index

The S&P 500 IT sector index includes large-cap IT companies from the S&P 500 universe, including major software companies (Microsoft, Salesforce), payment processors and IT services firms that do not manufacture or design semiconductors at all. The semiconductor weighting within the S&P 500 IT index is substantial but diluted by these non-semiconductor IT names. SOX is a purer semiconductor proxy. An investor who believes semiconductors specifically will outperform broader IT benefits from SOX's tighter sector focus.

SOX vs. SOXQ (Invesco PHLX Semiconductor ETF)

SOXQ is an ETF that seeks to track SOX. It is the most accessible vehicle for investors who want index-level semiconductor exposure without trading individual stocks or futures. However, SOXQ's live holdings are not identical to the canonical SOX index weights for several reasons: ETF portfolio management involves cash drag, dividend reinvestment timing, corporate action processing and periodic rebalancing that may not coincide exactly with the index rebalance. For research purposes, the definitive constituent and weight data is always the index itself as published by Nasdaq, not the ETF's disclosed holdings. Do not substitute an ETF's portfolio disclosure for the index specification when studying methodology.

SOX research and data pages

• SOX Index Methodology: full detail on ICB eligibility criteria, weighting caps, reconstitution calendar, rebalancing schedule and July 2026 methodology update
• SOX Current Constituents: current 30-company basket with primary ICB classification and value-chain archetype
• SOX Constituent History: year-by-year composition changes and reconstitution additions and removals
• SOX Reconstitution Calendar: annual reconstitution reference dates, effective dates and historical additions and removals
• SOX Subsector Breakdown: component groupings by value-chain role with subsector-level analysis
• SOX AI Exposure Map: classification of each SOX component by AI value-chain position
• SOX Annual Returns: calendar-year and trailing-period index returns
• SOX-Linked ETFs and Products: SOXQ, options contracts and other vehicles that reference SOX

Frequently asked questions

References

• Nasdaq: PHLX Semiconductor Sector Index Overview
• Nasdaq: SOX Index Methodology (PDF)
• SEC EDGAR: Company Filings Search