CRDO — Q3 FY26 Earnings Review (Bert Hochfeld)

Date: 2026-04-28 Quarter: Q3 FY26 (ended Jan 31, 2026; reported March 2, 2026) Stock: $94.21 (last close used in 3/31 analysis) | Market cap: 17.6B|EV: 16.3B FY ends: April

Summary

I initiated coverage of Credo on March 31 with a Buy at 3-4% target weight. Returning now to the Q3 FY26 print in proper earnings-review format, the conclusions only firm up. Revenue of $407M came in at +201.5% YoY and +51.9% sequentially, beating the guide midpoint by $67M, or 19.7% — the largest beat-and-raise in company history and the fifth consecutive quarter of widening beat magnitude. Non-GAAP gross margin of 68.6% was 260bp above the high end of guidance, non-GAAP operating margin of 49.6% set a record, and free cash flow of $139.7M (34.3% of revenue) marked the cash-flow inflection point I had been waiting for. The Rule of 40 score is 235. Q4 FY26 revenue is guided to 425−435M with non-GAAP gross margin of 64.0-66.0% — a guidance pattern that, on my read of the prior five quarters, is deliberately undemanding. FY27 commentary points to "nearly $2B" of revenue, more than 50% YoY. None of this is reflected in a stock that remains 56% off its $213.80 peak. Recommendation: Buy. 3-4% initial weight, add to 5-6% on Q4 margin confirmation. Thesis status moves from Intact (New Coverage) to Strengthening.

Headline KPIs (Q3 FY26)

• Revenue: 407.0M|YoY + 201.567M (+19.7%)
• Non-GAAP GM: 68.6% (guide: 64.0-66.0%) | GAAP GM: 68.5%
• Non-GAAP Op Margin: 49.6% (record, +327bp QoQ) | GAAP Op Margin: 36.8%
• Non-GAAP Net Margin: 51.3% | Non-GAAP EPS: $1.07 vs consensus $0.94
• OCF: $166.2M | FCF: $139.7M (34.3% margin) | Cash + ST inv: $1,301.5M | Debt: $0
• AEC ≈ 80% of revenue | AEC product GM ≈ 72% | IC solutions GM ≈ 62%
• Customers >10%: 3 (39%, 32%, 17% — top-3 = 88%)
• RPO: $33.9M (Q2 disclosure; Q3 not separately quantified)
• SBC: $52.2M (12.8% of revenue, down from 16.9% in Q2)
• Q4 FY26 Guide: 425−435M (mid +5.7% QoQ); NG GM 64-66%; NG OpEx $76-80M
• FY26 implied: ~1.33B( > 2002B (>50% YoY)

Five Analytical Layers

1. Revenue trajectory — accelerating into a hardware refresh

Quarterly revenue progression: $135M → $170M → $223M → $268M → $407M. The Q2 → Q3 sequential jump of 52% is not the deceleration narrative the bears were pricing. It is a re-acceleration off a 20% sequential quarter, which itself followed a 26% sequential quarter. The straight-line read is that the AEC franchise is benefitting from a 100G-to-200G transition that drives both ASP increases and unit-count expansion as hyperscalers densify infrastructure.

The five-quarter beat history — $10M, $15M, $28M, $28M, $67M — shows widening beat magnitude. This is the cleanest pattern of "deliberately undemanding estimates" I track outside of Axon. When management raised the FY26 guide four times ($800M → $960M → 1.0B → >1.3B), each raise still left room beneath the actual delivery.

The Q4 guide of $425-435M implies +5.7% sequential at the midpoint. In isolation this looks like decelerating momentum. Set against the prior pattern, the most likely outcome is a Q4 print of $460-490M, which would put FY26 closer to $1.36-1.39B. I would not bet against the sandbagging. But I would also not assume it — the brief explicitly notes "the market has high bar embedded."

2. Margin structure — software economics in hardware silicon

68.6% non-GAAP gross margin in a fabless connectivity semiconductor business is exceptional. The Q3 expansion of 90bp QoQ on a guided step-down to 64-66% is the second consecutive quarter where management's GM forecast has been beaten by 200-260bp. The CFO's framing — "long-term expectation in the 63% to 65% range" — is now openly inconsistent with the demonstrated execution. This is the classic sandbagging signature: when management's stated long-term GM target is consistently beaten by 300-500bp, the company is not running below capability — the guide is a floor.

Operating margin of 49.6% non-GAAP is the kind of figure I see at mature application software companies, not at a hyper-growth hardware vendor. The 327bp QoQ expansion came despite a 35% sequential lift in non-GAAP OpEx for R&D — meaning the operating leverage on this revenue base is structurally enormous. R&D is being funded by the AEC margin profile while the next product pillars are built underneath.

There is one caution. AEC product GM of 72% combined with IC solutions GM of 62% means the blended margin is highly sensitive to product mix. As ZeroFlap Optics, ALCs, and OmniConnect ramp from FY27, blended GM will compress before maturing back toward AEC levels. Management's 63-65% long-term target is conservative but the directional signal — initial GM compression on new products — is real.

3. Free cash flow and capital structure — the inflection

FCF of $139.7M at 34.3% of revenue is the cash flow inflection I flagged as the key watch item in my March 31 initiation. Q3 OCF of $166.2M was up $104.6M sequentially. With no debt, $1.3B of cash and short-term investments (boosted by ATM proceeds), and inventory of $208M positioned ahead of demand, the balance sheet is now a fortress. Capital intensity is light — fabless model, $26.5M Q3 capex against $407M revenue.

The 12.8% SBC as a percent of revenue is well below the 16.9% in Q2. SBC in absolute dollars is still rising ($52.2M), but the dilution math gets better as revenue scales faster than SBC. Diluted share count of 192M (GAAP) / 194.9M (non-GAAP) is up only modestly QoQ, with the Q4 guide pointing to ~197M.

4. Customer concentration — the unresolved structural risk

Top-3 customers represent 88% of revenue. Top-1 is 39% (down from 42% in Q2 and 61% in Q4 FY25). Top-2 surged to 32%. The fourth hyperscaler that briefly crossed 10% in Q2 dropped back below. CEO commentary: "The customer concentration is stubborn… driven by the fact that we've got three world-class hyperscalers that are all ramping at different rates."

This is the prior-learning that surfaced for me before this analysis: near-zero RPO ratio as structural demand-visibility risk. CRDO's $33.9M RPO against $407M quarterly revenue is 0.08X — far below the 0.25X threshold I have set as the demand-certainty floor. This is a short-cycle PO business. Five quarters of widening beats says current demand is exceptional. The contractual visibility says nothing about Q+2 or Q+3.

The implication for sizing is not that I avoid the position — it is that I cap the weight more tightly than I would for a similarly-margined contracted-backlog business. 5-6% is my hard ceiling here, even on confirmation. A hyperscaler capex pause of the type we saw briefly during the DeepSeek shock would land directly on the income statement with no contracted backlog absorbing it.

5. Platform / TAM — five pillars, three multi-billion-dollar additions

The Q3 transcript confirmed the platform thesis I built in the stock analysis. The CoMira acquisition closed in the quarter (link-layer, ECC, security IP). PCI-SIG compliance was achieved on the Toucan PCIe retimer. A 224G multiprotocol AI scale-up retimer was announced with Upscale AI as first customer. ZeroFlap Optics has been pulled forward to Q1 FY27 production with TensorWave first, and four additional hyperscaler/neocloud customers in qualification.

Combined TAM commentary: ">$10B in the coming years," up from ~$3B 18 months ago. AEC alone is ~5B(20 − 251.33B FY26 revenue, that is ~13% blended TAM penetration — which is the kind of runway that matters.

The CPO (co-packaged optics) competitive concern that has dominated the narrative is, on the CEO's framing, "not an either-or" — copper remains the cost/power-efficient solution for inter-rack connectivity, and CRDO carries best-in-class on both the AEC (copper) and IC (optical DSP) sides. NVIDIA's stated preference for copper-where-possible is a structural endorsement.

Five-Factor Snapshot

Valuation Re-Mark

Updating the cohort comparison from my March 31 stock analysis with no material movement since:

The framework I apply — segment AI compute exposure into training, commodity inference, and frontier inference, then layer Jevons demand expansion — is highly favorable to CRDO because connectivity infrastructure benefits from total compute volume regardless of the GPU/ASIC mix. CRDO's AEC TAM expands whether NVIDIA holds 80% share of GPUs or falls to 60% with custom silicon picking up slack. More compute, more interconnects, more cables. The segmentation is the wrong axis to worry about.

The cycle-narrative discount that suppresses Micron despite 74% GM is a relevant cross-read. CRDO is not historically cyclical — it is a 7-year-old company with no prior cycle to revert to. The discount here is not "this will revert," it is the customer concentration and short-cycle PO structure. Different mechanism, but the result is the same: structurally cheap entry on equivalent or superior fundamentals.

Prior Beliefs / Updated Beliefs

Risks (in order of materiality)

1. Hyperscaler capex pause / customer concentration shock. Top-3 = 88% of revenue, RPO 0.08X quarterly. A single hyperscaler rephasing orders during a budget reset could remove 30%+ of revenue in a quarter. This is the dominant near-term risk and the reason I cap at 5-6%, not 7-8%.
2. GM mix compression as new pillars ramp. ZeroFlap Optics, ALCs, and OmniConnect will start at lower margins. Management's 63-65% long-term GM target is the most likely directional signal. The current 68.6% is unlikely to be the steady state.
3. CPO and competitive substitution. Co-packaged optics narrative is mostly noise today, but at higher rack densities it is a real architectural alternative. CRDO's hybrid copper+optical positioning is defensible, not bulletproof.
4. Hardware cyclicality and exit timing. Even great hardware companies get re-rated lower at the end of capex cycles. The asymmetric upside is now; the asymmetric downside is at the cycle peak, which is unknown.
5. SBC absolute dollar growth. $52.2M Q3 SBC will keep pressuring share count. The 12.8% ratio is improving; the absolute dollars are not.

Conditions Under Which I Am Wrong

• Q4 FY26 prints below $425M (the low end of guide). That would break the five-quarter beat pattern and signal real demand softening.
• Non-GAAP GM falls below 65% in Q4 with no AEC mix-shift explanation. That would invalidate the structural margin thesis.
• A top-3 hyperscaler is publicly disclosed as pausing or reducing 2026 AI infrastructure capex.
• Customer concentration worsens to >90% top-3 with no fifth hyperscaler ramping.

Recommendation

Buy. 3-4% initial weight, scaling to 5-6% on Q4 FY26 confirmation (Non-GAAP GM ≥66% and revenue ≥$430M, ideally with a beat). Hard ceiling 6% given customer concentration and absent contractual backlog.

The stock is down 56% from $213.80 to $94.21 while the fundamentals have improved on every measurable axis. EV/FY27 Revenue at 8.1X for a 50%+ grower with 49.6% operating margins, 34.3% FCF margins, and a $1.3B net cash balance sheet is the kind of disconnect that resolves toward fundamentals over a 12-24 month window. Panics tend to be indiscriminate. This one has overshot.

FY27 math at 2.0Brevenueanda454.30-5.00 in non-GAAP EPS. At a 25-30X multiple — generous discount to the 30-40X warranted by the growth and margin profile — that is $108-150 per share, or 15-60% upside from current. Add in the option value of FY28 contributions from ALCs and OmniConnect — neither of which is in any analyst model today — and the upside skew is substantial.

Decision rule applied: cohort EV/S ≥ 20% discount with management credibility "reliable" and no broken thesis condition met = Buy at target weight, capped by structural risk profile.

Analysis based on Q3 FY26 actual results (reported March 2, 2026), prior Bert stock analysis (March 31, 2026), and Atlas Q3 FY26 earnings review baseline.