AEC, Retimers & Rack-Scale Connectivity — Strategy

Where AEC, Retimers & Rack-Scale Connectivity is heading over the next 12 months, grounded in product-axis evidence and verbatim demand from the last 90 days. The judgment column is the engine's read — operators verify and refine.

Product trajectories

Credo HiWire AEC (400G/800G Active Electrical Cable) — active electrical cable (AEC) with retimer DSP → steady

Opportunity: Hyperscaler AI fabrics need ≥800G in-rack copper links that survive long reaches and high BER where passive DACs fail; HiWire's retimer DSP commands a 2-3x ASP premium over passive DAC (verified_fact bd2fad3c) and is the line that drove Credo to ~$437M FY25 revenue +125% YoY (e7e3dfbb) with >65% GM (8343daa9).

HiWire is the de-facto incumbent 800G AEC and the cleanest pure-play AI-fabric copper product in the topic; the unresolved risks are hyperscaler concentration (single-customer pause in CY24, 9c0f6cb5) and the structurally lower hyperscaler GM (70dae263) as Semtech CopperEdge attacks the same sockets.

Semtech CopperEdge AEC — active electrical cable (AEC) with retimer DSP → steady

Opportunity: Hyperscalers want a second source against Credo HiWire on 800G in-rack AEC at the same $400-600 ASP band (bd2fad3c); Semtech reallocated proceeds from the PerSe divestiture specifically into CopperEdge R&D (ceb734c1) to chase that socket.

CopperEdge is the only credible #2 AEC retimer SKU in the evidence, but FY25 datacenter segment is still <$100M (41d781f5) — strategically committed (capital redirected from PerSe) but pre-scale; whether it actually displaces HiWire at a hyperscaler is the open question.

Enfabrica ACF (Accelerated Compute Fabric) 800Gbps chip — high-radix network fabric chip (AEC-adjacent retimer/switch hybrid) → steady

Opportunity: Hyperscalers exploring whether a fabric chip with built-in retiming/switching can collapse the AEC + ToR-switch stack for AI clusters; Enfabrica is sampling 800Gbps silicon to those customers (5067528d).

Adjacent threat / option to the discrete AEC retimer category — ACF is sampling but has zero publicly disclosed production design wins (0fc6cfe4), so it is a 2026+ wildcard rather than a current revenue contender against HiWire/CopperEdge.

Credo SerDes IP licensing — 112G/224G SerDes IP block licensed to ASIC/SoC customers → steady

Opportunity: The same 112G SerDes that powers HiWire AEC retimer chips is licensable IP to customers building their own ASICs to the CEI-112G-Linear spec (5cc686e0); a high-margin recurring stream that monetizes the underlying retimer technology without manufacturing AECs (b453170a).

Quiet but high-quality monetization of the AEC retimer technology stack — only ~$30M in FY25 (b453170a), so it's a moat indicator (Credo's SerDes is good enough that others license it) rather than a near-term swing factor on the topic.

Credo Optical DSP (Dove / Seagull family) — 800G DR/FR optical DSP retimer · weak signal

Opportunity: Same hyperscaler customer base that buys HiWire AEC also needs 800G DR/FR optical retimers; Credo is leveraging its AEC channel to plug into pluggable optics (0cf241d0).

Strategically the natural AEC adjacency for Credo but still <15% of FY25 revenue (0cf241d0); read as optionality on Credo's SerDes IP rather than as a topic-defining product today.

See the Products and Hiring modules for the full landscape and who's investing in which direction.