CRDO - Bull · Base · Bear Investment Thesis

CRDO · June 15, 2026

Prepared June 14, 2026 · Source: SEC EDGAR — FY2025 Form 10-K (filed 2025-07-02) and XBRL company facts (CIK 0001807794). Fundamental framework only — not investment advice.

Credo Technology Group (CRDO) designs high-speed connectivity silicon for AI and hyperscale data centers — Active Electrical Cables (AECs), SerDes / retimers, optical DSPs, PCIe retimers, and Ethernet. It is fabless (TSMC-dependent) and incorporated in the Cayman Islands. The debate below turns on one fact: explosive, newly-profitable growth set against extreme customer concentration.

The financial spine (SEC XBRL)

Two facts from the FY2025 10-K frame the whole debate:

• One customer = 67% of FY2025 revenue; top 10 = ~90%.
• Revenue mix shifted to 97% product (from 85%), with high-margin IP licensing down to 3% (from 15%).

🟢 Bull — The AEC standard for the AI back-end

Bet: Credo owns an emerging connectivity category at the exact moment hyperscaler AI capex is inflecting, and the model has flipped from cash-burning to highly profitable with enormous operating leverage.

• Hyper-growth that's now profitable. Revenue went from a ~$32M trough (Apr'23) to ~$410M (Jan'26), while operating margin moved from −51% to +37% — proof of a scalable model, not just growth. Net margin reached 39%.
• Expanding gross margins into the ramp (58% → 68.5%), unusual for hardware scaling volume — implies pricing power / favorable AEC mix.
• AECs are a structural winner: more reliable and lower-power than optical for short-reach in-rack links inside dense AI clusters. Credo is the most-associated name in the category (28 AEC / 46 SerDes references in the 10-K).
• Higher-quality mix: shift from lumpy IP licensing to recurring product revenue (85% → 97% product) makes the top line more durable.

🟡 Base — Real franchise, but a hyperscaler-capex derivative

Bet: A genuine category winner whose growth rate is a leveraged bet on a few customers' build schedules — model deceleration off a parabolic base and customer diversification over time.

• The +200–270% YoY prints are off a tiny, depressed base against a single-customer ramp; growth normalizes toward "very high double-digit," not triple-digit.
• Margins likely plateau around high-60s gross / 30s operating as easy operating leverage is captured and large customers gain pricing leverage.
• The thesis works if Credo converts lead-customer validation into a broader roster. Watch the concentration line: 67% falling while revenue grows proves the franchise; staying at 67%+ means you're long one customer's roadmap.
• Fundamentally a call option on AI back-end networking capex — size it accordingly, not as a stable compounder.

🔴 Bear — One customer, one cycle, priced for perfection

Bet: The concentration and cyclicality the filings disclose, plus a fabless cost structure, make the current trajectory fragile — and the 2023 history shows how fast it can reverse.

• 67% of revenue from one customer. The 10-K is explicit that a small number of customers account for a significant portion of revenue. A single program pause, in-sourcing decision, or share shift would be catastrophic — and invisible from outside.
• It has happened before. Revenue was negative YoY for four straight quarters (Apr'23–Jan'24) with operating margin at −51%. AI demand masked the cyclicality; it didn't remove it.
• Quality-of-revenue erosion: the shift away from IP licensing (15% → 3%) drops the highest-margin, stickiest line in favor of order-cadence-dependent product sales.
• Fabless + TSMC dependency: the 10-K flags that capacity shortages can raise raw-material costs and compress gross margin — margin expansion can reverse on supply terms.
• Parabolic charts cut both ways: a stock discounting +200% growth re-rates violently on the first sequential miss — and a concentrated customer is the most likely trigger.

What would move you between cases

• The concentration number in each 10-K — 67% falling = bull/base; holding or rising = bear.
• Sequential growth + guidance in the 8-K earnings releases.
• Gross-margin direction as the lead customer's volume builds pricing leverage.
• New named customers / design wins diversifying the base.

What this is built from / caveats: financials are from SEC XBRL company facts; concentration and revenue-mix facts are verbatim from the FY2025 10-K. This analysis has no valuation, share price, backlog, or parsed forward guidance — so it does not address whether the stock is cheap or expensive today, which is central to the bear "priced for perfection" point. The "one customer is Amazon" attribution sometimes cited in the press is not named in the 10-K; the filing only discloses the 67% without identifying the customer.

This is an analytical framework generated from public SEC filings for research purposes. It is not investment advice, a recommendation, or an offer to buy or sell any security.