Is Credo Under Threat From Optical?

The Short Answer

Yes, but Credo knows it and is pivoting. The threat is real at the physics level — copper has a hard distance and bandwidth ceiling that optical does not. But the timeline gives Credo a multi-year runway, and the company is aggressively building an optical product line to ride the transition rather than be killed by it.

How Credo's Business Works Today

Credo essentially invented the AEC (Active Electrical Cable) market — copper cables with embedded DSPs that extend copper's usable range to 5-7 meters while consuming 25-50% less power than optical alternatives. Their distinctive purple cables are visible in hyperscaler rack photos from AWS, Microsoft, xAI, and Meta.

Current dominance:

• 73-88% market share in AECs (estimates vary by quarter)
• Q3 FY26 revenue: $407M (tripled YoY, up 52% sequentially)
• Q4 FY26 guidance: $425-435M
• FY26 total expected ~$1.2B, up 173% YoY
• AEC market on pace to hit $4B by 2028

Where AECs sit in the stack: Inside the rack, connecting GPUs to switches. Nvidia's GB200 NVL72 uses 5,184 massive copper cables to lash 72 Blackwell GPUs together in an all-to-all shared memory config. This is scale-up — copper's stronghold.

The Physics Problem: Why Optical Eventually Wins

The fundamental issue is per-lane speed vs. distance:

At 200G per lane (the 1.6T generation, ramping now), passive DAC can only reach ~1 meter. AECs extend this to ~3m with aggressive DSP processing, but that's the physical limit. The raw bit error rate at 200G/lane hits 2-4 x 10^-3 before FEC — orders of magnitude worse than previous generations.

The implication: Every speed generation shrinks copper's usable radius. At 3.2T (400G/lane), AECs may not be viable at all for anything beyond sub-meter connections.

The Two Domains: Scale-Up vs. Scale-Out

This is the critical distinction that determines Credo's exposure:

Scale-up (within the rack/pod): GPUs talking to each other via NVLink, NVSwitch. This is copper's stronghold. Nvidia uses copper here because it's lower power (~16W vs. optical modules), lower latency, and cheaper — saving ~20kW per cabinet. Credo's AECs dominate this domain.

Scale-out (rack-to-rack, across the data center): Switches talking to switches, connecting racks into clusters of hundreds or thousands of GPUs. This is already optical. An 18,000-GPU GB300 cluster requires 18,432 800G transceivers and 27,648 1.6T transceivers for scale-out.

The threat narrative is: As GPU counts per rack increase and rack architectures evolve, even scale-up connections may need to go optical. If Nvidia's next-gen (Rubin/Vera) designs move to optical for intra-rack connections, Credo's core AEC business shrinks.

The Three Optical Threats, Ranked

1. Pluggable Optical Transceivers (Near-term, moderate threat)

• 800G pluggables already dominate scale-out
• 1.6T pluggables ramping 2026-2027
• These don't directly compete with AECs today (different domain) but expand the "optical zone" closer to the GPU
• Key players: Coherent, Lumentum, II-VI, InnoLight

2. Linear-Drive Optics / LPO (Medium-term, higher threat)

• Removes the DSP from the optical transceiver, reducing power and cost
• Makes optics more competitive with AECs on power consumption (the main AEC advantage)
• If LPO closes the power gap, the distance advantage of optics becomes decisive
• Timeline: ramping in 800G generation, mainstream at 1.6T

3. Co-Packaged Optics / CPO (Long-term, existential threat)

• Embeds optical transceivers directly onto the switch/GPU package — no separate cable or transceiver needed
• Eliminates electrical-to-optical conversion overhead entirely
• Market projected to exceed $20B by 2036 (37% CAGR)
• Timeline: commercial scale 2027-2029, mainstream at 3.2T/6.4T generation
• If CPO works at scale, it eliminates the need for both AECs and pluggable transceivers

Credo's Defense: The Three-Product Pivot

Credo management clearly sees the threat and is executing a transition. They've made two acquisitions and have three new product lines:

1. ZeroFlap (ZF) Optics — Pluggable transceivers

• 400G/800G optical transceivers with Credo's reliability tech (the "ZeroFlap" brand = zero link flaps/drops)
• Production ramp in fiscal Q1 2027 (July 2026), across 4+ customers
• This puts Credo directly into the pluggable optics market competing with Coherent, Lumentum, etc.
• Demonstrated at OFC 2026

2. Active LED Cables (ALCs) — MicroLED-based optical cables

• Acquired Hyperlume (Sept 2025, ~$40-50M) for MicroLED optical interconnect technology
• Uses MicroLEDs instead of lasers — thinner cable, matches copper's power and reliability, but extends reach to 30 meters (vs. 3m for AECs)
• Targets the "row scale" middle ground between intra-rack copper and long-haul optical
• Sampling fiscal 2027, production fiscal 2028
• This is the most creative part of the strategy — it's effectively "an AEC but optical," preserving Credo's cable form factor and customer relationships

3. OmniConnect Gearbox — Protocol translation silicon

• Acquired CoMira Solutions for link-layer IP
• Gearbox chips that translate between different interconnect speeds and protocols
• Ramp in fiscal 2028
• Makes Credo relevant regardless of whether the cable is copper or optical

Management's TAM claim: Each of these three new product lines represents a [1B+addressablemarket](https://markets.financialcontent.com/stocks/article/finterra−2026−3−16−the−ai−pipeline−a−deep−dive−into−credo−technology−group−crdo−and−the−future−of−16t−connectivity),expandingCredo′stotalTAMto 1B+addressablemarket](https://markets.financialcontent.com/stocks/article/finterra−2026−3−16−the−ai−pipeline−a−deep−dive−into−credo−technology−group−crdo−and−the−future−of−16t−connectivity),expandingCredo′stotalTAMto 10B by end of decade.

The Bear Case

• AEC revenue = the vast majority of today's $1.2B. The optical products are pre-revenue or early-revenue. If AEC demand declines before optical ramps, there's a gap.
• FY27 growth slowing to ~40% (from 173% in FY26) — this is partly the AEC saturation signal.
• Inventory days rose from 83 to 140 — potential demand softening or over-building.
• Gross margins dipping to 64-66% during the product transition.
• Competition in optical is fierce. Credo is entering a market with entrenched players (Coherent, Lumentum, Broadcom, Marvell) who have decades of optical experience. Can a copper company out-execute optical specialists?
• CPO could skip the pluggable generation entirely. If hyperscalers jump from pluggable optics to CPO faster than expected, Credo's ZF Optics product has a compressed window.
• Nvidia's choices are king. If Nvidia's next architecture (Rubin, 2H 2026) shifts intra-rack connections to optical, Credo's core AEC business takes a structural hit. If Nvidia keeps copper for scale-up (likely for at least one more generation), Credo has more runway.

The Bull Case

• Copper isn't dying tomorrow. At 800G (today's mainstream), AECs still have strong advantages in power, cost, and reliability. The 1.6T ramp extends the AEC opportunity through 2027-2028.
• Nvidia is still pro-copper for scale-up. Jensen Huang publicly reaffirmed copper's role at GTC, and the Blackwell/GB200 architecture is heavily copper-dependent. The next generation (Rubin) is also expected to use copper for NVLink.
• The ALC concept is genuinely novel. MicroLED-based cables at 30m reach, with copper-like power and reliability, don't exist from anyone else. If this works, Credo has a differentiated optical product — not just a me-too transceiver.
• Credo's DSP silicon expertise transfers. Whether the medium is copper or photons, the signal processing IP (SerDes, FEC, equalization) is the value. Credo isn't a cable company — it's a connectivity silicon company that happens to sell cables.
• The total market is expanding faster than copper's share is shrinking. Even if AECs go from 100% of intra-rack to 50%, the number of racks is growing 5-10x. Absolute AEC volumes could still grow even as optical takes share.

Timeline Summary

Bottom Line

Optical is a real, physics-based threat to Credo's core AEC business — but it's a slow-moving one with a 2-3 year runway. The company is executing a deliberate pivot with acquisitions (Hyperlume, CoMira) and new products (ZF Optics, ALCs, OmniConnect) that could make it a copper-AND-optical connectivity company rather than a copper-only casualty.

The risk is execution: can a company that built its franchise on copper compete in optical against specialists who've been doing it for decades? The ALC/MicroLED bet is the most interesting angle — if it works, it's a differentiated product no one else has. If it doesn't, Credo is left competing head-to-head against Coherent and Marvell in pluggable transceivers, which is a much harder fight.