Optical Selloff: Why COHR, CRDO and MRVL Got Smashed

Bond yields are going to go back down right? Right?!?

Thursday, October 8 was the ugly one. Coherent $COHR lost 9.6%. Marvell $MRVL gave back 4.6%. Credo $CRDO dropped 3.7% and Aeva $AEVA fell 4.5%. Lightwave Logic $LWLG bled all week. Not one of them had news. Lumentum $LITE fell with the group Thursday, then did something the others couldn't. It closed Friday at $1,103.36, up 5.2%, after its CEO said the company's optical parts are sold out through early 2029.

That's the tell for optical stocks this week. The trade didn't break. It got sorted. Money is crowding into the one name that can prove it's sold out and backing away from everything that still has to prove anything. The further out your revenue sits, the harder you got hit. Friday made the point again. The big caps bounced. The small caps didn't.

We've been covering this group since Nvidia's August quarter repriced the photonics trade. Here's our read on why it's getting hulk-smashed now, and who's still standing.

Rates did most of the damage

Start with the 10-year Treasury. It's above 5.2%, a 24-year high, up from about 3.9% in March. The Fed hiked in September and the October hike talk hasn't gone away. Oil isn't helping. WTI sat above $90 Friday and was climbing again Monday morning, which feeds inflation and keeps the long end heavy. We mapped the AI-stock side of this when the Fed moved.

High-multiple optical stocks are long-duration assets. A 5% risk-free rate takes a bigger bite out of a dollar earned in 2029 than a dollar earned next quarter. That's the whole ladder. Lumentum earns the dollars now. Coherent earns most of them now. Credo and Marvell earn some now and promise a lot more later. Lightwave Logic and Aeva are still mostly promise. Rank them that way and you've basically ranked the drawdowns.

Then the AI bill got questioned

Thursday also brought an OpenAI story that put a pin in the AI capex balloon. A report pegged OpenAI's annualized revenue near $50 billion at the end of September, well under the $68 billion figure that had been floating around. AI stocks tumbled. Friday morning Bloomberg said OpenAI is expected to reach $70 billion by year-end, and the group bounced.

Whiplash like that hits optics harder than it hits the hyperscalers. When the biggest AI customer's revenue wobbles, the companies selling the pipes get marked first. Transceivers, lasers, switch silicon, cables. All of it rides on capex budgets the market now wants proof of.

It's CPO week

The OCP Global Summit runs Monday through Thursday in San Jose. It's the conference where co-packaged optics gets its annual stage, and every CPO headline gets read as a tax on pluggable transceiver makers. Add the usual whispers about 1.6T module ramps and you get a group that sells first and asks later.

What the sellers skip: Lumentum is on the OCP agenda Wednesday with a talk on external laser sources for deployable CPO networks. CPO moves the laser. It doesn't remove it. Same with optical circuit switches, where Lumentum has called 2027 demand incredibly strong. The architecture fight is real for module assemblers. For the people who make the light, it's mostly a change of address.

There's also just more stock

Applied Optoelectronics $AAOI dropped a $600 million at-the-market program on the group in August. Aeva sold $115 million of shares in June. Credo and Marvell insiders have been filing sales all month. None of it changes a thesis. All of it is supply, and supply is what you notice when the bid gets thin.

Name by name

$LITE. The exception, and the benchmark. Fiscal fourth-quarter revenue hit $1.01 billion, up 109%. Management guided the next quarter to $1.25 billion and now says capacity is sold out through early 2029. Friday's close was $1,103.36. The 52-week low is $152.54. Even the leader drops 4% to 10% on red days. It keeps getting bought back because the numbers keep showing up.

$COHR. Record fiscal 2026, carried by datacenter and communications. The drag is industrial, down 16% year over year in the last reported quarter. The stock closed Thursday 23% below its June 30 level while consensus revenue for fiscal 2027 and 2028 rose 11% and 14% over the same stretch. Price down, estimates up. The multiple on next-twelve-month earnings went from 53x to about 32x. Friday's close was $312.62 against a 52-week high of $440. Bernstein started coverage at Outperform with a $350 target at the end of September. Coherent got cheaper, not worse.

$CRDO. Credo took its beating early. On September 2 it reported revenue up 115% and earnings up 131%, raised guidance, and fell 20% in a day. The market zeroed in on GAAP gross margin, which slipped to 64.5% from 68.2%, and on operating expenses that more than doubled. Customer concentration didn't help. Shares went from about $283 in mid-August to the $150s by mid-September. They've since recovered to $216.55, still 30% under the $308.67 high. Insiders are selling the bounce. Filings this month show the CTO sold about $6.1 million of stock and the COO sold 50,000 shares. Stifel kept its Buy and cut its target to $310 from $350.

$MRVL. Marvell held its Investor Day on October 6 and swung big. A long-term revenue target of $70 billion to $90 billion. A fiscal 2028 outlook of $20 billion. A $30-plus EPS forecast. The stock popped 5.8% to $287 that day. Three sessions later it was $275.28. Price targets went up across the Street, with Needham at $400 and Cantor at $350, and none of it stuck. In a 5% world, a 2030 number is exactly the kind of number that gets discounted hardest. Marvell is still up about 224% this year, so there's plenty of profit to take. The COO took some in October.

$LWLG. This is where the ladder gets steep. Lightwave Logic is pre-revenue, so the multiple is the whole stock. It printed a 52-week high of $18.71 and closed Friday at $4.97. That's a 73% drawdown, with the 200-day average up at $8.65. Market cap is about $766 million. The Street is one Buy, one Sell, consensus Hold at $10. The next report lands November 12. When rates rise there's nothing in the income statement to lean on.

$AEVA. Aeva is a lidar company that added an AI data-center optical connectivity pitch in August. That's what put it in this group, and now it's getting both discounts. The 52-week high is $31.30. Thursday's close was $13.69, down 56% from the top. Second-quarter revenue was $6.1 million against $302.9 million of liquidity, helped by that $115 million June share sale. Fugazi Research published a short report in August calling it “valuation by adjacency.” Translation: headlines with Nvidia $NVDA and Daimler instead of recurring revenue. The report also flagged about $12 million of insider sales after the August optics announcement. Goldman started it at Neutral with a $20 target. It's hard to hold a story stock when a short seller and a bulge bracket both tell you to wait.

What would un-smash it

A few things, and none of them are company news. The 10-year has to stop making new highs. Friday's chart showed its momentum stalling, which is a start. CPI prints Wednesday morning. A cool number takes the October hike off the table and the long end exhales. OCP has to pass without a CPO headline that reads like a pluggable obituary. And hyperscaler earnings arrive later this month. If the capex guides hold, the proof question gets answered for the whole supply chain at once.

For the small caps it's simpler and harder. They need a number of their own. Credo needs gross margin to stop sliding. Lightwave Logic reports November 12. In a tape that's paying for proof, a print is the only thing that moves you up the ladder.

Subscribe to StockAlpha.ai for the next read on the optical stocks tape.