Bargo
Deep Dive

Credo Technology (CRDO): The Deep Dive

When Q4 FY26 revenue exceeded the entire prior year and every Blackwell rack ships with Credo inside it, the question isn't whether the thesis is real. It's how long the copper runway lasts before optical takes over.

Bargo · 2026-07-08

Credo Technology (CRDO) makes a specific chip most people have never heard of, and it sits inside every NVIDIA Blackwell rack in the world. In FY26, Credo tripled revenue, grew profits 5x, and generated $437M in a single quarter (April 2026) that alone exceeded its entire prior year. This piece explains what Credo actually does, why the numbers are this good, and what could break the story. Data as of Monday, July 7, 2026. Source: BargoAI research.

What does Credo Technology actually do?

Credo makes something called an Active Electrical Cable, or AEC. Think of it as a copper cable with a small chip built into each end. Inside a data center, when NVIDIA GPUs need to talk to each other at very high speed over short distances (a few feet, inside one rack), that data has to travel through cables. Credo's cables have the chips that clean up the electrical signal so nothing gets lost.

Two facts make this a real business:

For a hyperscaler running thousands of AI training jobs, a cable that fails less often is worth more than a cable that costs less. Downtime on one training run can cost millions. That is the moat.

What did the Q4 FY26 numbers actually say?

Credo reported fiscal Q4 (three months ending April 2026) on June 1, 2026. The one line that mattered came from CEO Bill Brennan:

Fiscal 2026 marked another defining year for Credo. Revenue exceeded $1.3 billion, more than tripling year-over-year, while non-GAAP net income increased more than 5x to $662 million.

And this one:

Notably, our revenue in the quarter exceeded our entire fiscal 2025 revenue.

Read that twice. Credo booked more revenue in one three-month quarter ($437M) than it did in the entire prior year. Very few semiconductor companies have ever done that.

Credo quarterly revenue ($M): the hockey stick

The financials in plain English

Here is what changed inside Credo over the last four quarters.

Metric Q1 FY26 Q2 FY26 Q3 FY26 Q4 FY26
Revenue $223M $268M $407M $437M
Gross margin 67.4% 67.6% 68.5% 68.2%
Operating margin 27.2% 29.4% 36.8% 35.7%
Cash on balance sheet $480M $813M $1.30B $1.44B
Total debt $16M $16M $16M $25M

Three things this table proves:

Why 68% gross margins are the real story

For context, most memory chip makers run 30% to 40% gross margins in good times. Most semiconductor companies run 50% to 60%. Credo runs 68%.

That number tells you Credo's chips are hard for customers to replace. When customers can substitute one vendor for another, margins get squeezed. When customers cannot substitute, margins hold. Credo's cables have been designed into NVIDIA's Blackwell reference architecture, which means every server maker building Blackwell racks (Dell, HPE, Supermicro, Foxconn, and so on) uses the Credo design. Swapping vendors mid-cycle is not practical.

What is the optical business, and why does it matter?

CEO Bill Brennan said this on the same call:

We believe fiscal 2027 represents an inflection point for Credo's optical business. In FY27 we expect our optical DSPs, SiPho PICs, and ZeroFlap Optics will each contribute more than $100M of revenue. In total, more than $600M of revenue.

In plain English: cables are Credo's current business. But the industry will eventually switch from copper cables to fiber-optic cables for longer distances inside data centers. Credo is building the chips that go into those optical cables too. In June 2026, Credo bought a small company called DustPhotonics that gave them the silicon photonics piece they were missing.

Management is guiding to $600M+ in optical revenue for the fiscal year ending April 2027. That is a huge number for a business that barely existed 12 months ago. It also protects Credo from the biggest bear case, which is that copper cables get replaced faster than expected.

What are the insiders doing?

Two different signals, and they matter for different reasons.

Small tax-related sales from the top three executives. On July 5, 2026, the CEO, COO, and CFO all filed Form 4s showing small share sales tied to tax withholding on vested stock (roughly 6,000 to 3,000 shares each). All three still hold substantial positions. This is routine, not a warning.

Big planned sale from the CTO. Chi Fung Cheng, the Chief Technology Officer, sold roughly $7.5M of stock across 23 separate transactions on June 30 at prices between $253 and $275. This is almost certainly a pre-scheduled selling plan (called a 10b5-1 plan) that lets insiders sell during blackout periods without violating rules. It still leaves him with a meaningful position. But the pace is worth watching. If Cheng keeps trimming aggressively over the next two quarters, that is a signal.

What the tape said: the June 26 drop and recovery

The stock had one very ugly day in the last 60. On June 26, 2026, CRDO dropped 11.2% to $238 on 38.6 million shares of volume, which is roughly four to five times the normal daily amount. That is a distribution day, meaning big holders were dumping.

Then something changed. Over the following week and a half, the stock recovered about 70% of that drop.

The tell that institutional buyers had turned buyers again came on July 2. That day the stock closed red (down 6.6%), but the order flow underneath was positive: net buying of $11M in the SIP tape, even as the price fell. That is what happens when institutions are quietly accumulating on red days while retail sells. Two sessions later on July 6, the stock rallied 9.8% on strong volume, confirming the pattern.

The current price of $246 is above the June 26 low but below the pre-drop $275 high. The tape says: distribution phase is over, accumulation is happening, but the stock hasn't fully reclaimed old highs.

What the options market is pricing

Two levels matter here:

The stock trades at $246, close to the middle of the $230 to $270 range. Options are currently in a "long gamma" regime, which means dealer hedging dampens moves and the stock tends to chop rather than trend.

One trade worth flagging: on the January 2027 chain (six months out), there are 1,010 open contracts on the $230 strike call. That is roughly $2.5M of premium sitting on a single bet that the stock is well above $230 by January 2027. Someone with real money believes the recovery continues.

What the Street thinks

Nineteen analysts cover Credo. Eighteen rate it Buy or Strong Buy. One (Rosenblatt) rates it Neutral.

Metric Value
Consensus rating Strong Buy (1.32 out of 5, where 1 = Strong Buy)
Analysts 19
Mean price target $269.81
Upside to mean +9.5%
Highest price target $350 (BofA raised)
Lowest price target $184

The bull case has been articulated most explicitly by BofA (target $340) and Evercore (target $325). Both argue the AEC ramp is stronger than the Street was modeling, and Evercore is separately building a case that the ratio of CPUs to GPUs in AI servers goes from 1-to-1 today to 2-to-1 or even 10-to-1 by 2030, driving huge growth in connectivity chips regardless of what happens in AI training.

Can this be a multibagger from here?

At today's $246, Credo has a market cap of roughly $50B. Rough math for what it would take to compound from here:

The main risks

The one date that matters: September 2, 2026

Credo reports Q1 FY27 (three months ending July 2026) after market close on September 2, 2026. Two questions get answered that afternoon:

If both answers are yes, the tape likely reclaims $275 and pushes toward the BofA $340 target. If either answer disappoints, the current multiple compresses toward 20x forward and the stock trades back to the $200 range.

Bottom line

Credo is one of the highest-quality names in AI infrastructure. Every Blackwell rack ships with its chips inside. Revenue tripled year over year. Margins expanded. Cash on the balance sheet grew from $480M to $1.44B in four quarters. No debt.

The bear case is real but sits in the future: at some point, optical replaces copper for these connections, and Credo's original business erodes. Management is trying to get ahead of that with their own optical chips. FY27 tells you whether the transition is working.

The June 26 distribution day was ugly on volume. The recovery over the following two weeks was the institutional response. Someone believes in the story enough to sit on $2.5M of six-month call options.

This is not investment advice. All live financials, options positioning, and signals are on bargo.ai.

Sources

Company disclosures

Sell-side research (attribution, notes are subscriber-gated)

Corporate action attribution

Live data feeds used

Reviewed by the Bargo editorial desk. Financials and quotes are from Credo's Q4 FY2026 earnings release on June 1, 2026 and management's forward guidance. Market signals from live Bargo feeds as of July 8, 2026. This is research and educational content, not investment advice.

Related research

Semiconductors
ACM Research: China's Semiconductor Decoupling Play Hits a Reset

The only US-listed pure play on China's domestic semi equipment boom just washed out 23% on a market-wide distribution day. The options market is betting it snaps back.
HBM
HBM week: SK Hynix declares victory as Samsung enters the game

Two signals landed on the memory market inside the same week. SK Hynix ran a marketing barrage claiming HBM dominance. Samsung Electronics began HBM4 mass production, entering the market as a real third supplier. Both stories are true. The pricing implication for NVIDIA, AMD, and Micron is not what the headlines suggest.
Semiconductors
The semi bloodbath: what Diffusion Rule uncertainty just flushed

Every major AI-semi stock fell 3 to 15 percent in a single session. The trigger was not earnings, not China data, not a hyperscaler downgrade. It was a Biden-era export rule the Trump administration wants to scrap, and the market's inability to price what replaces it.