High-Speed Connectors and Relays: TE Connectivity and Why Electromechanical Parts Survive in EVs

Every trend in electronics still has to cross a physical boundary. Data rates rise, process nodes shrink, vehicles move to 800V architectures, and at every seam between one board, module or subsystem and the next, the signal or current is carried by a contact, a spring or a switching element, not a transistor. This is the electromechanical layer of the ESDM stack: connectors, switches and relays.

I want to use TE Connectivity as a worked example of the discipline, explain the mechanical versus solid-state relay trade-off, and then make the argument I find most useful: why the electric vehicle, the most digitised product in the industry, is also the strongest case that electromechanical parts are not going away. The facts come from the electromechanical chapter of my ESDM book, and its evidence gaps stay marked as gaps.

Connectors: denser and faster at the same time

A connector's job sounds simple, joining two conductors so current or signal can cross a seam. Circuit boards, power sources, sub-assemblies and external interfaces cannot be soldered into one another once a system passes a single board. Each seam is a potential source of signal reflection, contact resistance and mechanical fatigue, which is why connector engineering became a discipline of its own.

It has evolved in two directions at once. The first is density. Stackable board-to-board mezzanine connectors at 0.4 mm, 0.5 mm and 0.8 mm pitch pack the contact count that dense mobile devices and AI server modules need. The second is speed. High-speed board-to-board and backplane connectors are routinely specified for PCIe Gen 5 and Gen 6 at 32 to 64 GT/s and for 112G and 224G PAM4 Ethernet backplane links. In that path a connector is not a passive afterthought. Its contact geometry, shielding and impedance profile must be engineered with the same rigour as the trace it interrupts, because a mismatched connector brings back the reflection and eye-closure problems a carefully chosen laminate was meant to avoid.

A third family serves a slower and harsher world. Ruggedised circular connectors (MIL-DTL-38999, M12 and M8) use bayonet or thread-locking because a friction fit cannot be trusted on a vibrating engine bay or an outdoor control cabinet. Together with high-voltage automotive interconnects they carry continuous currents well beyond any signal connector. So connector engineering is at least two disciplines: signal integrity at vanishing pitch, and mechanical and electrical robustness at rising voltage and current.

Both disciplines converge on one field metric: contact resistance. Automotive-grade connectors are commonly specified to hold initial contact resistance under 5 milliohms, tight enough that fretting corrosion or contact wear can be caught before it shows up as voltage drop, local heating or an intermittent fault. Reliability is measured in cycle life instead of a single pass or fail. The USCAR-2 automotive standard sets a baseline of 30 to 50 mating cycles for standard automotive connectors, which reflects how often a connector is realistically disconnected and reconnected across a vehicle's service life and is not a theoretical maximum. The book notes it does not document the contact plating alloys (gold flash, tin, silver) chosen for each tier, so I do not either.

A fine-pitch choice is a trade. Using a 0.4 mm connector instead of a coarser legacy footprint swaps connector cost and assembly precision for board area that can carry more compute, memory or power delivery. It works only if contact engineering keeps up, since a 0.4 mm contact has far less margin for misalignment during mating. That is why fine-pitch qualification leans as much on mechanical tolerance and coplanarity as on electrical performance. High-speed connectors go the other way: they hold pitch roughly constant and add shielding, differential-pair routing and impedance-matched contacts.

Relays: mechanical contact versus solid-state switching

A connector joins conductors that stay joined. A switch or relay makes and breaks that join on purpose, repeatedly. A mechanical relay energises a coil that pulls an armature across a set of contacts, so an open relay has a true galvanic break: no electrical path at all, only air and a gap. A solid-state relay (SSR) reaches a similar result with an optocoupler for input isolation and a MOSFET output stage, and no moving parts anywhere in the signal path.

The absence of moving parts is the whole case for solid state. With no armature to fatigue, no spring to weaken and no contact surface to pit or arc, an SSR gives silent, bounce-free switching and what the book's reference calls unlimited cycle life. That is why SSRs are specified into medical equipment and semiconductor test equipment, where contact bounce could register as a false signal or corrupt a test, and where particulate from arcing contacts is unacceptable in a cleanroom or beside a patient. A mechanical relay's contacts part and re-close on every cycle, carrying some risk of bounce, arcing and eventual wear. The book gives no numeric cycle-life rating for mechanical relays, so I make only the qualitative point: mechanical switching has a wear-driven life ceiling that solid-state switching lacks.

Each technology holds the ground where its trait matters. Tactile switches remain the default human-interface element because they give feedback a solid-state option cannot match at similar cost. Mechanical relays stay strong wherever true galvanic isolation, tolerance of high inrush current, or simple field-serviceable switching is worth more than bounce-free operation, general industrial power switching among them. SSRs take the segments where bounce, acoustic noise or arc particulate is the risk to be removed.

The economics are simple, even though the book attaches no unit costs. A mechanical relay is a coil, a spring and contacts, a mature process with decades of volume behind it, which tends to keep its bill of materials low. An SSR's premium is the price of removing mechanical wear-out from a system where it is unacceptable and not just inconvenient. That is a subsystem-level decision, which is why both technologies keep shipping in volume. And neither escapes the connector layer: a relay still terminates into a board or harness through a connector, so contact resistance and mating-cycle discipline apply to switching just as they do to signal and power interfaces.

TE Connectivity: one platform across three regimes

The connector industry is led by a few specialist champions instead of a field of interchangeable suppliers: TE Connectivity (Switzerland and the USA), Amphenol (USA), Molex (USA), Hirose Electric (Japan), Samtec (USA) and Rosenberger (Germany). Their positions rest on decades of contact design and materials science, since a reliability record built with an automotive or data-centre customer is slow and expensive for a rival to displace on price.

TE is the largest by the available evidence, but the book's own figures differ on how large. TE's executive summary says over 20% of the global connector market. A more granular figure sourced to Bishop & Associates for 2024 gives about 14.8%, derived from roughly $12.88 billion of TE connector revenue against a global market of about $87 billion, which sits near the middle of the $80 billion to $90 billion range. Both agree TE holds the largest single share. They disagree on magnitude, and I report the range of roughly 15% to 20%.

What a revenue figure understates is the span of physical regimes one platform covers: automotive RF and Ethernet signalling, high-voltage EV power delivery, and data-centre backplanes. Each has its own dominant constraint, and TE's product families are specified for it.

FAKRA is the workhorse for signals that matter to vehicle function but are not bandwidth-hungry today. MATE-AX is rated to 20 GHz and 28 Gbps because ADAS cameras and radar produce high-resolution streams, a requirement that grew out of multi-camera, radar and LiDAR fusion. AMP+ IPT lives in another regime entirely: its job is not signal integrity but making and holding the physical connection between the battery pack and the traction inverter, at temperatures up to 140 C that reflect a high-current power path. STRADA Whisper sits at the opposite extreme from FAKRA, at backplane speeds of the class the opening section tied to PCIe Gen 5/6 and 112G/224G Ethernet.

Automotive-grade connectors are also qualified against vibration standards, USCAR-2 Class 3 and Class 4 engine-vibration compliance alongside ISO 16750 environmental qualification. A bandwidth rating only matters if the contact survives the vibration it will meet without intermittent disconnection. Signal or power performance qualified together with mechanical durability is the connector discipline, applied by one company across automotive, EV and data-centre lines.

In India, TE Connectivity India in Bengaluru and Amphenol Interconnect India in Pune are the two multinational anchors of the interconnect ecosystem, with Kaynes Technology and Centum Electronics specialising in cable harness and custom interconnect integration around them. The book has no figure for India's total domestic connector production value and marks it evidence required.

The EV test case: why the contacts stay

If any application should squeeze electromechanical parts out in favour of pure electronics, it is the electric vehicle. Semiconductor content per car has risen from roughly $500 in a legacy combustion vehicle to somewhere in the $1,500 to $2,000-plus range for an electric, ADAS-equipped vehicle, a three to four times increase, driven by domain controllers running 254 to 500 TOPS, lockstep safety microcontrollers qualified to ISO 26262 ASIL-D, and battery-management ICs that monitor individual cell voltage to +/-1.5 mV accuracy. (Other chapters of the same book cite a somewhat wider range, so read this as an order of magnitude.) Every one of those describes more computation and control than the previous generation. And yet the layer that ties that silicon to the battery and the motor has not become less electromechanical. The stakes on it have risen.

Start with the pack. A high-voltage EV pack assembles 96 to 192 lithium-ion cells in series to reach 400V to 800V, hundreds of electrochemical cells that some physical scheme must join into one current path and break cleanly and safely on service, crash or shutdown. The battery-management layer is itself partly electromechanical. Passive and active cell balancing uses 10-ohm shunt resistors to bleed or redirect charge between cells, a physical resistive intervention and not a purely computational one. The isolated communication link between the high-voltage cell-monitoring hardware and the vehicle's 12V CAN logic, the isoSPI daisy chain linking cell monitors such as the ADI ADBMS6818 to a master BMS controller, depends on a galvanic isolation barrier built to withstand 5,000V RMS or more. No software in the BMS controller substitutes for an actual barrier between a several-hundred-volt stack and low-voltage logic a technician might touch.

The power path out of the pack says the same from another angle. Once the BMS decides to deliver power to the traction inverter, that power leaves through a rated high-voltage connector, and AMP+ IPT at 1,000V DC and 300A is TE's answer to that. No silicon carbide efficiency in the inverter changes the fact that several hundred amps of DC must cross a physical mating interface to reach it, and that interface's contact resistance, thermal rating and vibration durability are electromechanical questions. The same holds for charging. EV high-voltage interconnects rated to 800V DC and 300A-plus serve packs and fast-charging inlets alike. The marketed fast-charge figure, 10% to 80% state of charge in roughly 15 to 18 minutes on 800V platforms such as the Porsche Taycan and Hyundai E-GMP, is achievable only if the charging connector can carry that current density without overheating at the contact.

Safety regulation makes the stakes explicit

ISO 26262 classifies the battery management system alongside steering, braking and the traction inverter as ASIL-C or ASIL-D, the highest criticality tier, and ASIL-D hardware must reach a fail-operational or fail-safe state within 50 milliseconds of a primary hardware failure. A BMS that has to isolate a faulted pack in that window must be able to command a physical, high-current disconnect quickly and reliably. The isolation barrier and the high-voltage connector infrastructure are the physical base under that requirement. The book documents the timing rule and the isolation spec in general terms but does not name the high-voltage contactor parts, makers or cycle-life ratings that do the disconnecting in a production BMS. That is a genuine evidence gap, and I leave it open instead of supplying a number.

The wider pattern holds beyond cars. As the digital content of a system grows, the interconnect and switching layer beneath it does not shrink in proportion. It grows in criticality, because every added sensor, controller and control loop needs a physical path to the power and signal it depends on, and each path is a connector, a relay or an isolation barrier.

What that means for policy and procurement

A localisation strategy that secures wafer fabrication, chip design or AI compute while treating connectors, relays and harnesses as commodity is solving part of the problem. A vehicle with a domestically designed inverter and a domestically made BMS chip still cannot ship without a qualified supply of high-voltage connectors, isolation barriers and relays to bind them into a powertrain. That is why India's interconnect ecosystem, Amphenol Interconnect India and TE Connectivity India alongside Kaynes and Centum, belongs in the same conversation as fabs and packaging, not as a downstream detail to be settled later. I would put it more strongly than the book's phrasing does: electromechanical capability is a precondition for a mature electronics ecosystem, not a lagging indicator of one.

What to take away

• Connectors are engineered in two directions: down to 0.4 mm pitch for density, and up to PCIe Gen 5/6 and 112G/224G PAM4 for speed. Contact resistance under 5 milliohms and 30 to 50 mating cycles under USCAR-2 are the reliability floor.
• Mechanical and solid-state relays divide the market by failure mode. SSRs remove contact wear, and mechanical relays keep the true galvanic break. Neither displaces the other.
• TE Connectivity holds roughly 15% to 20% of a global connector market of about $80 billion to $90 billion. The two source figures (over 20% and about 14.8%) disagree, and the range is the honest answer.
• In the EV, more silicon has meant more dependence on 96 to 192 series cells, a 5,000V RMS isolation barrier and a 1,000V, 300A connector, not less. ASIL-D timing of 50 milliseconds makes the physical disconnect path safety-critical.
• Two gaps remain in the source: contact plating alloys and the specific HV contactor components. India's domestic connector production value is also unstated.

Every doubling of a vehicle's compute budget still has to close through a handful of contacts rated in amps and volts. That boundary has not moved.