Finvest
CRDO Semiconductors · AI infrastructure · Data center · Fabless chips · Thesis updated July 12, 2026

Great ramps, still few customers

01 Running thesis

Three big ramps, one big question

Credo is a high-growth data center parts company. Its best proof point is simple: in fiscal 2026, three end customers each mattered at real scale. Customer D was 33% of revenue, Customer B was 32%, and Customer E was 19%. Together, they were 84% of revenue.

That is better than a story built on one buyer. AECs, which are active cables that help move data at high speed, drove over 99% of the fiscal 2026 revenue increase. The bull case is that Credo has become a needed supplier for several AI and cloud data center builds.

The risk did not go away. A prior quarter showed Customer B, Customer D, and Customer E at 88% of revenue, while Customer C had fallen below the 10% disclosure line after being much larger before. That tells us demand can move fast from one customer to another.

The stock can work if the top three customers keep ordering and a fourth large customer appears. It can break if two large customers pause projects at the same time, or if new products like OmniConnect and Weaver do not add a second growth leg beyond AECs.

Jun 2026▲The fiscal 2026 10-K showed a more balanced full-year customer mix: Customer D at 33%, Customer B at 32%, and Customer E at 19%. Concentration is still high, but the data supports the view that Credo is serving multiple large hyperscalers.
Jun 2026→The Q4 FY2026 transcript could not be fetched, so the view relies on the fiscal 2026 10-K. The main open questions are the post-Q4 customer run-rate and the fiscal 2027 timeline for OmniConnect.
Mar 2026▼The Q3 FY2026 10-Q showed the top three customers at 88% of revenue, with Customer B at 39%, Customer D at 32%, and Customer E at 17%. Growth looked strong, but the customer swings were a warning.
Dec 2025▲The Q2 FY2026 10-Q showed a fourth major customer, Customer E, at 11% of revenue. That helped the diversification story, even as Customer C rose to 42%.
Sep 2025▲The Q1 FY2026 10-Q showed revenue spread across three major customers instead of one. That reduced the single-customer fear that had dominated the prior view.
Jul 2025→The fiscal 2025 10-K confirmed that one customer was 63% of annual revenue by end customer profile. The same filing also confirmed strong AEC-led growth and the company's first meaningful full-year profit.
Mar 2025▲A major hyperscaler ramp drove a sharp jump in product sales and net income. The upside was clear, but the quarter also made customer concentration the main risk.
Dec 2024▲A new hyperscale customer, Customer E, became the largest customer at 33% of quarterly revenue. That was an important proof point for Credo's ability to win new large accounts.
02 Business model

Fabless chips, cable revenue

Credo is fabless. That means it designs chips and connectivity products, then relies on outside manufacturers such as foundries and other supply partners to make them. This keeps Credo focused on design, customer wins, and product ramps.

The company makes money from product sales, product engineering services, and IP licensing. Product sales include integrated circuits and AECs. IP licensing lets customers use Credo's SerDes technology, which is chip technology that sends and receives very fast data streams.

Credo sells in two directions. It works with hyperscale end customers to create demand, then works with OEMs and ODMs that build the actual systems. In some wins, the end customer can require suppliers to use Credo parts.

This model can scale well when orders rise, but it has weak spots. A small number of hyperscale buyers can move revenue a lot, and Credo depends on outside supply partners to build and deliver products on time.

03 Product portfolio

What Credo sells

Growth engine

Active Electrical Cables

AECs are high-speed cables with electronics inside them. They are the main growth driver today and powered over 99% of the fiscal 2026 revenue increase.

Steady

Integrated circuits

Credo sells chips for Ethernet and data movement, including Line Card PHYs and PCIe Retimers. These products sit inside networking and compute systems.

Option

Optical DSPs

Optical DSPs help optical modules move data over fiber. Credo's lineup includes Seagull, Dove, Lark, Robin, and Bluebird.

Option

OmniConnect and Weaver

OmniConnect was introduced in November 2025. Weaver is the first product in the family, a 5nm memory fanout gearbox aimed at AI inference memory limits.

Option

SerDes chiplets

SerDes chiplets are small blocks of high-speed data technology that can be added to bigger systems. They could matter more if customers want modular chip designs.

Steady

IP solutions

Credo licenses SerDes IP to customers that want to use the technology in their own chips. This is a different revenue stream from selling hardware.

Option

microLED interconnects

Credo added microLED optical interconnect technology through the Hyperlume acquisition in fiscal 2026. The goal is future multi-terabit links through Active LED Cables.

04 Business segments

Where shipments go

North America shipments58%growing fast
Rest of World shipments42%growing fast

This mix is by shipment destination for fiscal 2026. It does not remove the bigger issue: the top three end customers made up 84% of fiscal 2026 revenue.

05 Risk factors

What could break

Two-customer pause

High impact · Medium odds

Credo's top three end customers made up 84% of fiscal 2026 revenue. One pause would hurt, but two pauses at once could reset the growth story. Data center builds are large projects, so orders can be lumpy.

We watchWatch quarterly end customer mix, especially Customer B, Customer D, and Customer E staying above the 10% line.

AEC demand cools

High impact · Medium odds

AEC products drove over 99% of the fiscal 2026 revenue increase. That makes the current story very tied to one product family. If hyperscalers slow AEC purchases, total revenue could slow fast.

We watchWatch management's comments on AEC unit shipments and whether AECs still drive most product sales growth.

New products take too long

Medium impact · Medium odds

OmniConnect and Weaver could help diversify the business. But they are new, and the internal view still has open questions on fiscal 2027 timing and revenue size. If they do not ramp, Credo stays more dependent on AECs.

We watchWatch for named OmniConnect or Weaver revenue contribution, design wins, or customer qualification updates.

Larger rivals win sockets

Medium impact · Medium odds

Credo competes with much larger semiconductor companies, including Broadcom and Marvell. These rivals have scale, broad portfolios, and deep customer ties. A lost socket can matter a lot when the customer base is concentrated.

We watchWatch for customer mix drops, weaker product gross margin, or management comments about pricing pressure.

Fabless supply chain strain

Medium impact · Low odds

Credo relies on outside manufacturing and supply partners. That is normal for a fabless chip company, but it adds delivery risk. If capacity, quality, or lead times worsen, Credo could miss customer ramps.

We watchWatch inventory, purchase commitments, and any filing language about foundry or supplier constraints.
06 Quick answers

In one breath

What does Credo Technology do?

Credo designs high-speed connectivity products for data centers. Its main growth driver today is Active Electrical Cables, which help move data between systems in AI and cloud infrastructure.

Why is customer concentration such a big issue for CRDO?

In fiscal 2026, three end customers made up 84% of revenue. That proves Credo has large wins, but it also means a spending pause from a few customers could hit revenue hard.

What is the bull case for Credo?

The bull case is that Credo's AEC products are becoming important across several hyperscale customers. If those ramps last and new products like OmniConnect add revenue, the business becomes less dependent on one product cycle.

What should investors watch next?

Watch whether Customer B, Customer D, and Customer E keep ordering in fiscal 2027. Also watch for a fourth customer above 10% of revenue and early signs that OmniConnect or Weaver can become meaningful.