Report Interpretation

Overview

Bernstein initiates coverage of Amphenol and TE Connectivity, two diversified connectivity companies whose products move power and data across AI infrastructure, transportation and industrial systems. The central industry question is not whether connectivity content will grow, but how that content will divide between copper and optical technologies. Bernstein expects coexistence rather than rapid copper elimination, prefers Amphenol as the broader interconnect exposure and views TE as an attractively valued industrial company with data-center optionality.

Core views

Interconnects are modular plug-and-socket components that carry power or data between chips, circuit boards, cables and finished equipment without permanently soldering the parts together. They allow modules to be assembled separately, replaced, serviced or upgraded, while demanding sufficient durability for heat, vibration, moisture and electromagnetic interference. Amphenol and TE sell interconnects, connectors, cable assemblies, backplanes, power-distribution products and sensors across data centers and many other end markets. Although AI infrastructure has become central to the investment narrative, about 60% of Amphenol and roughly 85% of TE remain outside data centers. The report divides the copper-versus-optical issue into three distinct applications. Scale-out communication across domains is already predominantly optical because the distances are too long for copper. Scale-up communication between racks also moves toward optics as domains expand: Blackwell and Rubin use a one-rack NVL72 domain, while Rubin Ultra is expected to connect eight racks in an NVL576 configuration. Bernstein regards these new inter-rack optical links as neutral to additive for Amphenol and TE because they create new connectivity content without removing the copper already used inside each rack. The genuine substitution risk is whether optics eventually replaces intra-rack copper. Copper remains cheaper, lower-power, lower-latency and easier to maintain over short distances, but its usable reach declines as lane speed rises. IEEE-based figures cited in the report indicate approximately 3 meters of copper reach at 50G per lane, 2 meters at 100G and 1 meter at 200G; above 200G, reach falls below 1 meter. AI architectures are progressing from roughly 100G toward 200G and ultimately 400G-plus lanes, while racks are becoming denser and require more links. Higher speeds also increase cable thickness, signal-integrity challenges and the need for power-consuming retimers. These constraints make an eventual move toward intra-rack optical connectivity plausible, particularly if Feynman reaches 400G lanes. Bernstein nevertheless expects copper to remain the principal intra-rack medium through at least early 2028. NVIDIA and Broadcom commentary reflects a preference to use copper wherever physically possible, and the timing and specifications of Feynman remain uncertain, with greater clarity expected around GTC 2027. Near-GPU optics also faces thermal sensitivity, precise-alignment requirements, serviceability problems, laser-failure risks and packaging yields that the report says are not yet ready for mass deployment. Inference workloads should also become a larger share of future capacity additions and can tolerate lower bandwidth, preserving a meaningful role for copper. Bernstein therefore expects a hybrid copper-and-optical architecture rather than a near-term wholesale replacement. Amphenol is Bernstein's preferred way to gain exposure to the interconnect market. Communications Solutions represented about 61% of sales in 2Q26, Harsh Environment Solutions about 21% and Interconnect & Sensor Systems about 17%. IT Datacom alone is approximately 40% of the company and is growing at a mid-teens sequential rate, supported by AI demand, higher connector content per rack and hyperscale deployment. Amphenol is described as the copper-interconnect market leader in data centers and as gaining share, while its CommScope acquisition added passive-fiber products and technology such as fiber-array units. Bernstein believes this portfolio positions Amphenol to retain copper content if substitution is delayed and capture optical content if the transition begins. Amphenol's second advantage is its decentralized acquisition and operating model. It has acquired more than 50 companies in the last decade and more than 35 in the last five years. Approximately 150 local general managers control customer relationships, pricing, engineering, sourcing and manufacturing, which the report argues supports rapid decisions, cost control and integration. The CommScope Cable and Connectivity Solutions acquisition was approximately USD 10.5 billion and was characterized as reasonably priced at about 11 times EV/EBITDA. A localized manufacturing footprint across more than 40 countries and a higher-margin IT Datacom mix support management's roughly 30% incremental-margin framework. Bernstein expects Amphenol's organic growth to moderate but remain low-double-digit or better through 2030, with operating margins approaching 30%. Its forecasts call for revenue of USD 35.253 billion in 2026, USD 41.821 billion in 2027 and USD 46.777 billion in 2028, with adjusted EPS of USD 2.65, USD 3.31 and USD 3.69. Those estimates are broadly aligned with consensus in 2026 but reach 3% above consensus EPS by 2028, leaving room for further positive revisions if new connectivity content expands the addressable market. Bernstein values Amphenol at approximately 28 times NTM+1 EPS of USD 3.50, producing a USD 99 target. The report acknowledges the premium valuation but argues that low-double-digit growth and an industry-leading margin profile justify it. TE Connectivity is framed differently: it is primarily a diversified industrial company with data-center exposure rather than a data-center company with industrial exposure. Industrial Solutions and Transportation Solutions each represent approximately half of sales. Digital Data Networks is about 15% of company revenue, while energy, automation, aerospace and defense, medical, automotive, commercial transportation and sensors provide broader growth. Bernstein expects high-single-digit to low-double-digit growth through at least the first half of 2028, supported by DDN orders rising approximately 70% year to date and by momentum in energy, automation and commercial vehicles. Bernstein argues that concerns about optical substitution overstate the threat to TE. About one-third of DDN is power-related and is largely insulated from optical replacement, while rising power requirements, including 800V DC, can expand TE's addressable market. The RAM Photonics acquisition adds fiber-array-unit technology, allowing TE to participate in future optical architectures. In a 2028 bear case, Bernstein estimates that replacing intra-rack copper could create a roughly USD 0.5-1.0 billion DDN revenue headwind. The model assumes about half of current content is lost because inference remains copper-oriented, while approximately USD 0.5 billion is added back through power content. The resulting high-single-digit to low-teens earnings decline would raise TE's FY28 valuation from approximately 15 times consensus EPS of USD 14.24 to about 17 times, which the report still does not regard as expensive versus diversified industrial peers. The report also disputes a simple conclusion that Amphenol is taking TE's addressable market. Amphenol has broader exposure to scale-out and passive optics, so the companies' data-center growth rates are not directly comparable. Customers commonly use multiple suppliers, and only a limited number of companies—including TE, Amphenol and Molex—can meet data-center specifications. TE's approximately 70% DDN order growth in 3QFY26 is expected to convert into revenue during FY27. Bernstein forecasts TE revenue of USD 19.867 billion in 2026, USD 21.920 billion in 2027 and USD 23.848 billion in 2028, with adjusted EPS of USD 11.46, USD 12.98 and USD 14.79; its 2028 EPS estimate is 4% above consensus. TE's transportation franchise supplies low- and high-voltage connectors, data connectivity and sensors used in electrification and driver-assistance systems. China represents about 27% of Transportation Solutions and approximately 35% of Automotive revenue. TE has seven automotive plants, 1,100 engineers and ten sales offices there and is about 90% localized across its Chinese supply chain. China content per vehicle rose from USD 50 in 2019 to USD 87 in 2025, and about 40% of automotive design wins come from China. Bernstein notes that Chinese OEM exports are projected in the report to rise from 3 million vehicles in 2025 to 9 million in 2030, although autonomy and electrification are not presented as immediate tailwinds. For valuation, Bernstein emphasizes that TE ranks around the middle of its approximately 13-stock coverage group on growth and margins but trades at the lowest multiple. The report values TE at approximately 17 times NTM+1 EPS of USD 14.30, yielding a USD 243 target, without assuming multiple expansion. It therefore views the present valuation as downside protection against slower growth and as offering rerating potential if DDN revenue accelerates. Bernstein cautions that near-term trading may remain volatile because of the ongoing preference for Amphenol, but describes TE as more suitable for a 12-month-plus value-oriented horizon.

Analysis framework

Bernstein first defines the interconnect market and separates scale-out, inter-rack scale-up and intra-rack scale-up links. It then compares copper and optics by distance, bandwidth, power, latency, serviceability and engineering constraints; builds a directional rack-content model; evaluates technology-transition scenarios for each company; analyzes segment exposure, orders, margins and acquisitions; compares forecasts with consensus; and applies forward P/E multiples to NTM+1 earnings.

Methodology notes

• Competition & strategyValue chain analysis

  Connectivity architecture and product-position analysis

  The report maps where copper cables, optical assemblies, connectors, power products and fiber-array units sit within AI racks and between racks, then assesses which parts of Amphenol's and TE's portfolios capture each content opportunity.

• Valuation methodsP/E and PEG Valuation

  NTM+1 P/E valuation

  Bernstein applies approximately 28 times NTM+1 EPS of USD 3.50 to Amphenol and 17 times NTM+1 EPS of USD 14.30 to TE, deriving targets of USD 99 and USD 243.

• Event-Driven and Behavioral FinanceExpectation Gap and Expectation Management

  Investor-concern versus operating-evidence comparison

  The report tests market concerns about rapid optical substitution and TE losing share against roadmap timing, customer multisourcing, DDN orders, company forecasts and valuation.

• Other

  Directional rack-level connectivity-content model

  Connector and cable content is estimated per chip and multiplied by chips per rack, while optical content is estimated per compute tray. The model includes a 20% redundancy allowance and is explicitly intended to show order of magnitude rather than a precise future bill of materials.

• Other

  Intra-rack optical substitution stress test

  For TE, Bernstein models a 2028 bear case with USD 0.5-1.0 billion of DDN revenue pressure, partial preservation of inference-related copper and roughly USD 0.5 billion of additional power content, then recalculates the earnings and valuation impact.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Amphenol Corporation (US.APH)

  Primary covered company and Bernstein's preferred exposure to the interconnect market

  Strengths

  Market leadership in data-center copper interconnects, approximately 40% IT Datacom exposure, passive-fiber and FAU capabilities following CommScope, diversified end markets, strong acquisition execution and a decentralized operating model.

  Weaknesses

  The shares trade at a premium multiple, and Bernstein expects organic growth and margin expansion to moderate from current levels.

  Comparison

  Amphenol has broader scale-out and passive-optical exposure than TE and is better positioned if intra-rack optical adoption accelerates.

  Risks

  Faster-than-expected CPO adoption, competitive pressure and failure to sustain acquisition integration or premium growth could weaken the thesis.

• TE Connectivity Ltd (US.TEL)

  Primary covered company offering diversified industrial exposure with data-center optionality

  Strengths

  Broad industrial and transportation exposure, approximately 70% year-to-date DDN order growth, insulated power content, RAM Photonics FAU technology, strong Chinese automotive localization and the lowest valuation in Bernstein's comparison group.

  Weaknesses

  Data-center growth is expected to trail Amphenol, the optical portfolio is less complete, and automotive autonomy and electrification benefits may take longer to develop.

  Comparison

  TE is less concentrated in data centers than Amphenol and trades at approximately 17x NTM+1 earnings versus Amphenol at approximately 28x.

  Risks

  Faster CPO adoption, competition from Amphenol and Molex, and a slower automotive-thesis ramp are explicitly identified risks.

Key data

• Amphenol rating and targetOutperform; USD 99Coverage initiation; current price USD 80.72 as of 21 September 2026
• TE Connectivity rating and targetOutperform; USD 243Coverage initiation; current price USD 210.01 as of 21 September 2026
• Copper reach by lane speedApproximately 3m at 50G, 2m at 100G and 1m at 200GAbove 200G per lane, the cited reach falls below 1 meter
• Expected intra-rack transition timing2028 at the earliestBernstein sees limited to no APH or TEL risk through at least early 2028
• Amphenol IT Datacom exposureApproximately 40% of company revenueDescribed as growing at a mid-teens sequential rate
• Amphenol acquisition recordMore than 50 acquisitions in ten years; more than 35 in five yearsSupported by approximately 150 decentralized general managers
• Amphenol 2028 forecastRevenue USD 46.777B; adjusted EPS USD 3.69Revenue is 1% and EPS 3% above consensus
• Amphenol valuationApproximately 28x NTM+1 EPS of USD 3.50Produces the USD 99 target price
• TE data-center exposureApproximately 15% of revenueAbout one-third of DDN is power-related
• TE DDN order growthApproximately 70% year to dateExpected to support stronger FY27 revenue
• TE optical-substitution bear caseApproximately USD 0.5-1.0B DDN revenue headwindPartly offset by approximately USD 0.5B of additional power opportunity
• TE 2028 forecastRevenue USD 23.848B; adjusted EPS USD 14.79Revenue is 3% and EPS 4% above consensus
• TE valuationApproximately 17x NTM+1 EPS of USD 14.30Produces the USD 243 target price
• TE China automotive exposureApproximately 35% of Automotive revenueChina is about 27% of Transportation Solutions revenue

Impact & implications

Bernstein's analysis implies that growing AI bandwidth and rack density should expand total connectivity content even as the technology mix evolves. Inter-rack optics is viewed as additive in the near term, while delayed intra-rack substitution protects existing copper content. Amphenol is positioned to participate under copper, optical or hybrid configurations, whereas TE's diversification, power products and industrial growth reduce the earnings significance of data-center copper risk. The report therefore supports Outperform ratings for both companies, with a stronger strategic preference for Amphenol and a valuation-based case for TE.

Risks

• Faster-than-expected adoption of co-packaged optics could accelerate the loss of intra-rack copper content.
• Competitive pressure from peers, including Amphenol and Molex, could affect TE's data-center growth or market share.
• The automotive benefits from autonomy, electrification and related content growth could take longer than expected to materialize.

What to watch

• Watch GTC 2027 for clearer Feynman specifications, launch timing and evidence on whether lane speeds will reach 400G.
• Track whether intra-rack optical packaging resolves its thermal, serviceability, reliability and yield constraints.
• Monitor whether TE's approximately 70% DDN order growth converts into stronger revenue during FY27.
• Follow the mix between training and inference workloads because inference can preserve demand for lower-bandwidth copper links.
• Track Amphenol's integration of CommScope and the readiness of its passive-fiber and CPO-related portfolio for 2028 or whenever customers require it.