Thesis. The best economics in the connectivity complex — a 68.5% gross margin on 201% revenue growth, which is what selling intellectual property looks like rather than hardware. The catch is that its copper cables partly compete with the optical links its own category peers sell.
Reviewed 11 September 2026, after fiscal Q3 2026 results. Next review after the next earnings on 2 December 2026.
What it does
Two products, both solving the same problem: moving data very fast over short distances.
SerDes — serializer/deserializer — is chip design IP. It squeezes many slow parallel data lanes into one very fast serial stream, which every high-speed link needs. Credo licenses these designs to others.
AECs — active electrical cables — are copper cables with signal-boosting chips built into each end. Inside a rack, over short runs, they replace optical links at lower cost and lower power.
Here is the physics that makes this a business: as each PCIe generation gets faster, the signal degrades over the length of copper trace it has to travel. At some point you need either active help or light. Credo sells the active help.
Bull case
- Fiscal Q3 2026 revenue up 201.5% at a 68.5% GAAP gross margin — by a wide margin the best economics in this category
- That margin proves the model. A 68.5% gross margin means Credo licenses designs rather than assembling hardware. Structurally a better business than transceiver manufacturing
- AECs solve a binding physical constraint, not a preference. Each faster PCIe generation makes signal integrity over trace length harder, which creates the need
- It sits directly in the rack architecture decision at hyperscalers, which is where the AI build-out actually gets designed
- The market has noticed — a one-year return already above 200% before a further 31.7% year to date
Bear case
- It is in strategic conflict with its own category. Where Credo wins a rack slot with copper, an optical transceiver loses one. It sits alongside the photonics names because it trades with them, not because their interests align
- The product line is narrow and absorbable. Broadcom or Marvell could integrate the retimer and SerDes function into their own silicon and remove the need for a discrete Credo part entirely
- It is small relative to the giants it competes with for design sockets
- Customer concentration among a handful of hyperscalers, with percentages not publicly disclosed
- The valuation prices continued triple-digit growth, which leaves no room for a single lost design socket
Major customers
Specific percentages are not publicly disclosed. Concentration among a handful of hyperscalers is the dominant structural risk.
- Hyperscalers — the AEC business sells directly into hyperscaler rack designs. Amazon and Microsoft have been reported among the largest accounts, though reported rather than disclosed
- Switch and system vendors — SerDes IP licensed into their networking silicon
- Optical module makers — Credo's DSP and SerDes designs appear inside third-party transceivers, which is the one place its interests do align with the optics names
Read-through — and this is the most useful divergence signal in the file.
Credo's copper AECs replace optical links for short in-rack runs. So if Credo is strong while COHR and LITE are weak, that is not relative strength — it is evidence that rack architecture is shifting away from optics. Tracked together, these three pages measure a real engineering decision as it happens.
Growth here is driven by rack architecture decisions, not transceiver volumes, so watch hyperscaler rack design announcements rather than optics demand.
What would change the view
- AEC design wins against losses at each hyperscaler — socket by socket is how this business is actually won
- Gross margin durability above 65%, which is the evidence the IP model holds
- Any integrated competing solution from Broadcom or Marvell
- The divergence signal — Credo strong while optics weak means architecture is moving
- Customer concentration disclosure, if it ever arrives
Update log
- 11 Sep 2026 — Reviewed. Verdict held at Constructive.
- 7 Aug 2026 — Full review after fiscal Q3 2026 results. The 68.5% gross margin at 201% growth became the centre of the thesis; the strategic conflict with the optical names and the integration risk from Broadcom and Marvell flagged as the limits.
Research and education only — not investment advice.
Made with Bullet