Coherent
It makes the plug-in modules that turn electrical signals into light so AI chips can pass data, and grows the lasers inside them itself.
The scores
Growth3/4 ↑
Revenue grew 22.5%, but it narrows to one segment, one doubling customer, and a rate roughly two thirds of its own market's.
Business quality2/4 ↓
Not 3 because the advantage was tested in 2025 and lost ground: its own report calls the product standard and a rival passed it.
Financial position3/4 ↑
Not 4 because operations funded 7.2% of the capital programme; debt is small, liquidity large and no maturity falls inside the horizon.
Execution record3/4 →
Eight straight earnings beats, margin up in 8 of 9 quarters, capacity a quarter early - but this chief executive has guided for two years.
Per share capture2/4 ↓
A convertible landed and $2B of new shares were issued at a protected price; revenue per share grew 4.4 points slower than revenue.
Potential2/4 →
The step to $3B a quarter needs four product lines with almost no revenue yet, and somebody else's capital budget.
Risk profile2/4 ↓
Not 3 because two risks resolve in one step inside the horizon: the export-control inquiry, and a customer at a fifth of sales.
Valuation2/4 →
The price pays the June 2027 plan in full and the good case past it; not 1, because two rivals have grown faster than that.
Evidence quality3/4 →
Everything central is filed and current, but the competitive finding rests on one industry source and the forward half on management claims.
If it works 0.5× · if it fails you keep about 36%.
The Verdict
We rate Coherent unattractive. It makes the optical modules AI data centres use to move data between chips, and grows its own lasers for them. The business is mixed: last year's profit was a record but produced almost no cash, and its market grew faster than it did while a Chinese rival passed it. The price is full, not cheap: it counts management's plan as done, with no credit for beating it. As a bet it is fragile: even if its new product lines work, the shares are worth only about half today's price, and if they fail, about a third is left. The price leaves no cushion either way.
The long caseIf AI clusters move light onto the chip package and switch it without electronics, Coherent has new parts to sell, provided buyers sign rather than build their own.
The evidence under this - 8 claims
the periods are not the same, and Coherent accelerated through its own year: Q4 fiscal 2026 datacenter revenue grew 66% year over year, well above the full-year 41%. Against that, Innolight's most recent quarter grew 207%. The gap narrows on the latest quarter and does not close.
LightCounting states the caveat itself: "the list accounts for sales of optical transceivers only, excluding all other products, which explains lower rankings of Coherent and Lumentum." Coherent's Datacenter & Communications segment also sells lasers, optical circuit switches, telecom transport and subsystems, so its rank on transceivers alone is not its rank in optics. It is nonetheless the rank in the product line that is the growth story.
the capital spending is deliberate rather than a leak: it funds the indium phosphide capacity that management says is the binding constraint on revenue, against an order book it says covers fiscal 2027 completely, with a stated payback of roughly 18 months. Capital spending on that basis is an investment decision, not a deterioration.
the same statement is the bull case and the bear case. If output really doubles, the below-market growth of fiscal 2026 was a supply problem that is about to end. If it does not, the $2,581.0M of inventory and the $1,103M of new plant were built for volume that will not ship. The 10-K's own risk factors name both outcomes, including 'insufficient customer demand to utilize our increased capacity'.
25 times operating profit is a chosen reference, not a law. On the company's own non-GAAP basis the multiple is 41.8 times, and if the exit rate is held and margins reach target the same enterprise value would sit near 21 times non-GAAP operating profit - demanding, but not absurd for a business its market thinks is compounding.
the first-quarter guide of $2.2B to $2.4B implies about 9.3% sequential growth in each of the following three quarters to reach it - faster than any quarter of fiscal 2026 except the 13.3% just delivered. The gross margin target needs another 260 basis points on a non-GAAP basis from the 39.5% to 41.5% guided for Q1, and management first gave the 42% figure at an investor day without a date on it.
including all three of those items instead gives $897.9M, within $4M of the same answer, so the choice of definition barely moves it. What does move it is the non-GAAP presentation, which removes both the amortisation and the stock, and that is the basis the company guides and is judged on.
The company
Between AI chips, at today's speeds, copper cable stops carrying the signal reliably, so data travels as light. Coherent makes the module that plugs into each switch and does the conversion, and, unusually, it makes more and more of the laser inside rather than buying it.
The catch is in its own annual report: that module is one of its very few industry-standard products, so a buyer can swap suppliers. A quarter of the company sells industrial lasers and materials.
The evidence under this - 4 claims
How it makes money
It is paid per module shipped, mostly by a handful of cloud companies, and nothing filed ties them to it. One unnamed customer was 20% of last year's revenue, double the year before; a second was 12%.
The evidence under this - 2 claims
the same statement is the bull case and the bear case. If output really doubles, the below-market growth of fiscal 2026 was a supply problem that is about to end. If it does not, the $2,581.0M of inventory and the $1,103M of new plant were built for volume that will not ship. The 10-K's own risk factors name both outcomes, including 'insufficient customer demand to utilize our increased capacity'.
Where the money comes from, stream by stream
| Stream | Share of sales (fiscal 2026) | How it earns it |
|---|---|---|
| Datacenter & Communications | 74.1% | Optical transceivers, lasers and optical switches sold to cloud and AI data-centre builders, plus telecom and data-centre-interconnect optics, paid per unit shipped. |
| Industrial | 25.9% | Industrial lasers, engineered materials and components for manufacturing, semiconductor equipment and electronics makers; shrinking after two divestitures. |
Who runs it
Jim Anderson has run Coherent since June 2024, after six years running Lattice Semiconductor; Sherri Luther, his finance chief there, followed him. Under them sales have grown fast and two side businesses were sold. Insiders own under 1% of the company and have been selling, not buying. The yearly bonus pays on sales and an adjusted profit measure, the long-term shares on beating other electronics stocks, and a new special award only if the share price keeps rising. Nothing in their pay rewards turning profit into cash.
The evidence under this - 5 claims
Nothing in the bonus or the units pays on cash flow, inventory or return on capital, so the fiscal 2026 year that converted 10.1% of profit into cash (C15) costs the plan nothing.
The awards are outside the regular plan and add to it; they reward the share price alone, not revenue, profit or cash, and the 8-K calls them necessary for retention after a 300% return.
Why it wins, and who could beat it
Why customers come back
The scarce part is the laser wafer. Coherent grows its own, on six-inch lines that management says yield about 4 times the chips of a three-inch wafer at half the cost. A rival that buys its lasers pays someone else's margin and ships only what it is allocated.
The evidence is margin: 37.50 cents of gross profit per dollar of sales last year, up from 35.17. But margin cannot tell a lasting cost edge from a shortage that will end, and the module around the laser is standard.
The evidence under this - 4 claims
the same statement is the bull case and the bear case. If output really doubles, the below-market growth of fiscal 2026 was a supply problem that is about to end. If it does not, the $2,581.0M of inventory and the $1,103M of new plant were built for volume that will not ship. The 10-K's own risk factors name both outcomes, including 'insufficient customer demand to utilize our increased capacity'.
The threat
Two Chinese makers are winning the cloud orders: Innolight's 2025 revenue rose 61% to $5.3B, and Eoptolink's rose 189% to $3.5B, passing Coherent by LightCounting's count. NVIDIA is building the optics into the switch itself and says its first is in full production. And one buyer needs no supplier at all: Google designed its own optical switch, one of Coherent's new product lines.
The evidence under this - 4 claims
LightCounting states the caveat itself: "the list accounts for sales of optical transceivers only, excluding all other products, which explains lower rankings of Coherent and Lumentum." Coherent's Datacenter & Communications segment also sells lasers, optical circuit switches, telecom transport and subsystems, so its rank on transceivers alone is not its rank in optics. It is nonetheless the rank in the product line that is the growth story.
co-packaged optics does not remove Coherent from the socket, it changes what Coherent sells into it: Coherent supplies the external laser modules, connectors and fibre attach, and its Texas plant is ramping the ultra-high-power laser for co-packaged optics "including those covered by our NVIDIA partnership". Management states dollar content is comparable to a transceiver. That is a management claim about a product line with almost no revenue yet.
LightCounting's May 2026 report says optical circuit switches "are gaining adoption beyond Google", so other buyers are buying the switch rather than building it, and Coherent says its own optical circuit switch revenue grew sequentially.
The rivals: NVDA, AVGO, GOOGL, GOOGL, and 3 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| Innolight Technology (Zhongji Innolight, 300308.SZ)Priced against it: Listed abroad | Datacenter & Communications | On sale | Listed abroad | |
| Eoptolink Technology (300502.SZ)Priced against it: Listed abroad | Datacenter & Communications | On sale | Listed abroad | |
| NVDA | NVIDIA Corporation, with Quantum-X InfiniBand Photonics and Spectrum-X Ethernet Photonics co-packaged optics switchesPriced against it: Small part of it | Datacenter & Communications | On sale | Small part of it |
| AVGO | Broadcom Inc., with the Tomahawk 6 Davisson co-packaged optics Ethernet switchPriced against it: Small part of it | Datacenter & Communications | On sale | Small part of it |
| GOOGL | Google (Alphabet Inc.), with its internally developed MEMS optical circuit switchPriced against it: Different product | Datacenter & Communications | On sale | Different product |
| Google (Alphabet Inc.), with its internally developed MEMS optical circuit switch | New datacentre lines | On sale | ||
| Wuhan Raycus Fiber Laser Technologies, Maxphotonics and TrumpfPriced against it: Listed abroad | Industrial | On sale | Listed abroad | |
| NVDA · AVGO | NVIDIA Corporation and Broadcom Inc., with their own co-packaged optics switchesPriced against it: Small part of it | New datacentre lines | On sale | Small part of it |
How big it could get
How big this could get
LightCounting counted $23.8B of transceiver sales in 2025 and estimates the market grew about 58%. The buyers are still accelerating: Alphabet alone spent $80.6B on property and equipment in the six months to June 2026, 103% more than a year earlier. Nobody publishes how many modules ship or at what price, so the market cannot be sized in units.
The evidence under this - 2 claims
Who has signed
NVIDIA signed a multi-billion-dollar purchase commitment in March, with a capacity agreement running to 2030; no volume or price is disclosed and neither agreement is filed. Management says fiscal 2027 is basically booked out, with orders into 2028, though no backlog is published.
The evidence under this - 2 claims
Coherent publishes no backlog figure in the 10-K or the results release, files none of these agreements as exhibits, and discloses no volume or term for any of them. Its own risk factors describe orders that "may be reduced, delayed, or cancelled" and warn of "excess or obsolete inventory" if forecast orders do not materialise, which is the company describing what its order book is worth as a promise.
The results
The numbers
Sales grew 23% in the year to Jun 2026, about the same pace as the 23% in the year before.
It keeps 11% of sales as profit, up from −5.0% in the year to Jun 2023.
It keeps 37¢ of each sales dollar after making what it sells, and spends 15¢ of it on selling and admin, its biggest running cost.
Shareholders' money earned 9.7% a year.
It used $1.02B of cash in the latest year, so the $1.99B it holds lasts about 1.9 years at that pace.
It owes $1.24B more than the cash it holds, about 15.5 years of the cash its operations bring in.
The share count rose 68% in 4.0 years, and pay in shares costs 2.7% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in June.
| Jun 2022 | Jun 2023 | Jun 2024 | Jun 2025 | Jun 2026 | |
|---|---|---|---|---|---|
| Size and growth | |||||
| Sales | $3.32B | $5.16B | $4.71B | $5.81B | $7.12B |
| Sales growth | +6.8% | +56% | −8.8% | +23% | +23% |
| Profit | |||||
| Gross margin | 38% | 31% | 31% | 35% | 37% |
| Operating margin | 12% | −0.7% | 2.0% | - | - |
| Net margin | 7.1% | −5.0% | −3.3% | 0.8% | 11% |
| What growth costs, per sales dollar | |||||
| Research and development | 11% | 9.7% | 10% | 10% | 10% |
| Selling, general and admin | 14% | 20% | 18% | 16% | 15% |
| Pay in shares | 2.2% | 2.9% | 2.7% | 2.8% | 2.7% |
| Capital spending | 9.5% | 8.5% | 7.4% | 7.6% | 15% |
| Cash | |||||
| Cash from operations | $413M | $634M | $546M | $634M | $79.5M |
| Free cash flow | $99.0M | $198M | $199M | $193M | −$1.02B |
| Balance sheet | |||||
| Cash and short-term investments | $2.58B | $821M | $926M | $909M | $1.99B |
| Debt | $2.30B | $4.31B | $4.10B | $3.69B | $3.22B |
| Net cash | $282M | −$3.49B | −$3.17B | −$2.78B | −$1.24B |
| Returns and shares | |||||
| Return on capital | 6.5% | - | - | - | - |
| Shares (diluted) | 117M | 138M | 152M | 155M | 195M |
What growth costs
Growing has made Coherent more profitable on paper: making each product and running the company cost less per dollar of sales, while research spending grows in step with sales. What grows faster than sales is spending on factories, which doubled as a share of sales last year, so growth widens the margin but has so far eaten the cash.
The evidence under this - 2 claims
- Selling, general and administrative carries acquired-intangible amortisation (C48), which runs off on its own, so part of the fall is the II-VI purchase ageing rather than operating leverage.
- Fiscal 2024 was a trough year, revenue below fiscal 2023's $5,160.1M, so ratios measured from it flatter the leverage.
- The research figure is the filer's ResearchAndDevelopmentExpense tag and matches the income-statement line of $723.0M (C43); there is no narrower or wider definition here.
the capital spending is deliberate rather than a leak: it funds the indium phosphide capacity that management says is the binding constraint on revenue, against an order book it says covers fiscal 2027 completely, with a stated payback of roughly 18 months. Capital spending on that basis is an investment decision, not a deterioration.
The last report (12 Aug)
Quarter ended June 2026, filed figures.
In the fiscal year ended June 2026 revenue grew 22.5% to $7.12B. The data-centre side rose 40.5%; the industrial side shrank 10.3% and fell in every quarter.
The profit is weaker than the headline. $198.1M of pre-tax earnings came from two business sales that will not repeat, and a 7.2% tax rate flattered the rest. Above all, $786.9M of earnings produced $79.5M of operating cash: $1.18B went into inventory, paid for mainly by taking 156 days to pay suppliers instead of 82.
What it did to the thesisIt confirmed the demand for data-centre optics and also the core of the Unattractive view. Much of the profit came from one-off sales and a low tax charge, and almost none of it became cash, because inventory was funded by paying suppliers far more slowly.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $1.74 against $1.62 expected.
The evidence under this - 7 claims
the capital spending is deliberate rather than a leak: it funds the indium phosphide capacity that management says is the binding constraint on revenue, against an order book it says covers fiscal 2027 completely, with a stated payback of roughly 18 months. Capital spending on that basis is an investment decision, not a deterioration.
the gross figure overstates the strain in two ways. Accounts payable released $748.5M back, so the working-capital block took $1,102.1M net rather than $1,183.0M. And on days, the cash conversion cycle widened only from 118 days to 124: inventory days rose 73 and payable days rose 74, so suppliers funded nearly all of the build. Separately, other operating net assets took a further $408.3M that the cycle does not capture.
What management did with the money
NVIDIA bought $2B of new shares at $256.80 in March, and that money, with new borrowing, paid for $1.10B of new plant; the business's own cash covered 7.2% of it. The share count rose 25.8%, but most of that was preferred stock converting: fully converted, it rose 3.7%.
The evidence under this - 3 claims
the company ended the year better capitalised than it started: cash and short-term investments of $1,987M against $909M, total debt down from $3,687M to $3,222M, and management puts net leverage at 0.7 times from 2.0 times. Whatever funded it, the balance sheet improved.
The price, and where we disagree
In short. Full. The price already pays for management's plan for mid-2027 working in full. If it lands, the shares are only fairly priced; if it slips, there is no cushion.
The shares returned 202.9% over the past year, more than twice the typical stock in its group. Then they fell 8.0% the day after the record results and closed 32.1% below their high: the market's own reaction to the plan was negative.
Insiders sold 39 times in the year and bought nothing, and the first quarter of the new plan, promised at $2.2B to $2.4B of sales, reports in November.
The evidence under this - 4 claims
A guide is the company's own forecast and this one is unusually load-bearing, because the June 2027 target of more than $3B a quarter is built off it: reaching that from a $2.3B midpoint needs about 9.3% sequential growth in each of the three following quarters, faster than any quarter of fiscal 2026 except the 13.3% just delivered. The gross margin range of 39.5% to 41.5% is also on the adjusted basis, which removes the acquired-intangible amortisation and share-based pay the accounts charge, so it is not comparable to the 37.50% the filings report for fiscal 2026.
How we read the multiple
When the thesis was written the business, counting its debt, cost about 67.6 times the $901.5M of operating profit it earned last fiscal year, after every real cost including stock paid to staff.
To fall to 25 times, operating profit must:
- Grow about 2.7 times.
- That is roughly what management's plan delivers: more than $3B of sales a quarter by mid-2027.
- It also needs 42 cents kept from each sales dollar after the cost of making the product.
- So a buyer today pays for that plan working in full.
Against its own history. Far dearer than its own normal on profit and on sales, while its cash return has turned negative: buyers are paying for growth that has not yet turned into cash.
The evidence under this - 3 claims
25 times operating profit is a chosen reference, not a law. On the company's own non-GAAP basis the multiple is 41.8 times, and if the exit rate is held and margins reach target the same enterprise value would sit near 21 times non-GAAP operating profit - demanding, but not absurd for a business its market thinks is compounding.
including all three of those items instead gives $897.9M, within $4M of the same answer, so the choice of definition barely moves it. What does move it is the non-GAAP presentation, which removes both the amortisation and the stock, and that is the basis the company guides and is judged on.
the first-quarter guide of $2.2B to $2.4B implies about 9.3% sequential growth in each of the following three quarters to reach it - faster than any quarter of fiscal 2026 except the 13.3% just delivered. The gross margin target needs another 260 basis points on a non-GAAP basis from the 39.5% to 41.5% guided for Q1, and management first gave the 42% figure at an investor day without a date on it.
Priced against its rivals
None of the listed rivals above sells the same thing to the same buyers as a large part of what it does. NVIDIA and Broadcom do too, but as a small part of much bigger companies, so their price says little about this one. A price comparison needs at least two, so the price is judged against its own history instead.
The bet: what it pays if it works
This bet is on one part of Coherent: four new datacentre product lines that have barely started selling. The rest of the company is counted separately, at a price the thesis would accept for what it earned last year.
Even if the new lines take half of management's planned jump in sales, the shares look expensive. Nothing for them is signed yet, and one big buyer builds its own switch. If they earn little, the rest of Coherent is worth well under today's price.
| New datacentre lines revenue in two to three years | Even fifths of that step needed | At the stated multiple |
|---|---|---|
| $760.00M | about 1 at 100% kept | 0.4× |
| $1.90B (the big win) | about 2 at 100% kept | 0.5× |
| $2.40B (market forecast) | about 3 at 100% kept | 0.6× |
| $3.04B | about 4 at 100% kept | 0.6× |
Independent forecasts put outside-bought optical circuit switches near $2.5 billion in 2029 and co-packaged engines near $10 billion in 2030; with management's multi-rail figure, the stated shares give about $2.4 billion a year. On that forecast, $2.40B pays 0.6× from today's value: an Expensive bet on the market case. New datacentre lines would be worth about $12.00B, 0.3× what the price pays for it.
To pay 2× (a Fair bet) it would need about $19.00B of New datacentre lines revenue (about 25 even fifths of that step); to pay 5× (a Cheap bet), about $54.27B of New datacentre lines revenue (about 71 even fifths of that step), at the same multiple.
Each row is New datacentre lines alone, plus the rest of the company (Transceivers, the rest of Datacenter & Communications, and Industrial) counted at about $22.54B.
The steps it has to take
- Done Optical circuit switch on sale as a Coherent product line
- Not started First co-packaged optics revenue, guided for the December 2026 quarter due 2027-02
- Not started A filed customer commitment naming one of the new lines due 2027-08
- Not started A quarter above $3B of revenue by the end of fiscal 2027 due 2027-08
- Not started Revenue disclosed for the new lines, so their share of the step can be read due 2027-08
- Not started Multi-rail and datacentre thermal revenue begin, guided for calendar 2027 due 2027-12
The detail behind the label
The evidence under this - 6 claims
the first-quarter guide of $2.2B to $2.4B implies about 9.3% sequential growth in each of the following three quarters to reach it - faster than any quarter of fiscal 2026 except the 13.3% just delivered. The gross margin target needs another 260 basis points on a non-GAAP basis from the 39.5% to 41.5% guided for Q1, and management first gave the 42% figure at an investor day without a date on it.
Coherent publishes no backlog figure in the 10-K or the results release, files none of these agreements as exhibits, and discloses no volume or term for any of them. Its own risk factors describe orders that "may be reduced, delayed, or cancelled" and warn of "excess or obsolete inventory" if forecast orders do not materialise, which is the company describing what its order book is worth as a promise.
nothing outside the company supports it, and the arithmetic runs the other way: a pluggable transceiver is a finished module Coherent assembles and sells whole, while in a co-packaged switch the optical engine sits inside somebody else's package and Coherent supplies the laser and the passive parts. First co-packaged revenue is guided for the December 2026 quarter, so no realised figure exists to check it against.
LightCounting's May 2026 report says optical circuit switches "are gaining adoption beyond Google", so other buyers are buying the switch rather than building it, and Coherent says its own optical circuit switch revenue grew sequentially.
including all three of those items instead gives $897.9M, within $4M of the same answer, so the choice of definition barely moves it. What does move it is the non-GAAP presentation, which removes both the amortisation and the stock, and that is the basis the company guides and is judged on.
25 times operating profit is a chosen reference, not a law. On the company's own non-GAAP basis the multiple is 41.8 times, and if the exit rate is held and margins reach target the same enterprise value would sit near 21 times non-GAAP operating profit - demanding, but not absurd for a business its market thinks is compounding.
Where we disagree with the market
A core winner of the AI optics boom: 24 of 31 analysts say buy, and NVIDIA put $2B into the shares.
Even if the plan arrives it is only fairly paid for, and meanwhile Coherent grows slower than its market and is funded by its customer.
Two facts sit outside the company's own release: an independent count says a rival passed Coherent, and a record year produced almost no cash.
The case against us is serious. Management says lasers, not demand, limit shipments, and that its laser output doubles by November. Fiscal 2027 is said to be booked out, and NVIDIA committed $2B to the shares.
What analysts say, and the evidence
The evidence under this - 6 claims
the periods are not the same, and Coherent accelerated through its own year: Q4 fiscal 2026 datacenter revenue grew 66% year over year, well above the full-year 41%. Against that, Innolight's most recent quarter grew 207%. The gap narrows on the latest quarter and does not close.
the capital spending is deliberate rather than a leak: it funds the indium phosphide capacity that management says is the binding constraint on revenue, against an order book it says covers fiscal 2027 completely, with a stated payback of roughly 18 months. Capital spending on that basis is an investment decision, not a deterioration.
the same statement is the bull case and the bear case. If output really doubles, the below-market growth of fiscal 2026 was a supply problem that is about to end. If it does not, the $2,581.0M of inventory and the $1,103M of new plant were built for volume that will not ship. The 10-K's own risk factors name both outcomes, including 'insufficient customer demand to utilize our increased capacity'.
Coherent publishes no backlog figure in the 10-K or the results release, files none of these agreements as exhibits, and discloses no volume or term for any of them. Its own risk factors describe orders that "may be reduced, delayed, or cancelled" and warn of "excess or obsolete inventory" if forecast orders do not materialise, which is the company describing what its order book is worth as a promise.
What could go wrong, and how we would know
The risks
- Profit that is not cash. Inventory reached 212 days of costs, funded by paying suppliers slower, and that can reverse.
- One customer, unnamed. The largest customer went from 10% to 20% of revenue and accounts for about 64% of the year's revenue increase.
- The module could disappear. NVIDIA, its customer and shareholder, is putting the optics inside the switch. Management says Coherent earns about as much per port selling the laser.
- An export inquiry. The US export regulator is asking about possible violations, and Coherent says it cannot estimate the loss.
The evidence under this - 6 claims
the gross figure overstates the strain in two ways. Accounts payable released $748.5M back, so the working-capital block took $1,102.1M net rather than $1,183.0M. And on days, the cash conversion cycle widened only from 118 days to 124: inventory days rose 73 and payable days rose 74, so suppliers funded nearly all of the build. Separately, other operating net assets took a further $408.3M that the cycle does not capture.
co-packaged optics does not remove Coherent from the socket, it changes what Coherent sells into it: Coherent supplies the external laser modules, connectors and fibre attach, and its Texas plant is ramping the ultra-high-power laser for co-packaged optics "including those covered by our NVIDIA partnership". Management states dollar content is comparable to a transceiver. That is a management claim about a product line with almost no revenue yet.
nothing outside the company supports it, and the arithmetic runs the other way: a pluggable transceiver is a finished module Coherent assembles and sells whole, while in a co-packaged switch the optical engine sits inside somebody else's package and Coherent supplies the laser and the passive parts. First co-packaged revenue is guided for the December 2026 quarter, so no realised figure exists to check it against.
What we still do not know
- Who the 20% customer is. The filing names neither customer above a tenth of revenue.
- Whether the margin is a lasting cost edge or extra profit from a shortage that ends when lasers stop being scarce.
The evidence under this - 2 claims
What would change our view
- Unattractive→AttractiveThe margin plan is delivered, at a multiple that no longer assumes it
It needs all of:
- ○Full-year adjusted gross margin reaches 42%, and operating cash exceeds 60% of earnings, so the profit arrives as cash.checked at the next results, due by 4 Nov
- ○The business, counting debt, costs about 25 times its filed operating profit, whether the profit rises or the price falls.about ≈$111.64 today, 61% below the last close
- Reopens the researchIt stops being third in its own market
It needs:
- ○An independent ranking puts Coherent back in second place, or its data-centre growth matches the market's two years running.watched as it happens
We would also be wrong if
- Free cash turns positive for a full year while plant spending stays above $1B.
What is coming
- November 2026 2026 proxy statement and annual meeting: The fiscal 2026 bonus payout against its revenue and EBITDA goals, the first ownership table since Bain sold out, and how shareholders vote on pay after the special awards.
Every dated event ahead
- November 2026 2026 proxy statement and annual meeting: The fiscal 2026 bonus payout against its revenue and EBITDA goals, the first ownership table since Bain sold out, and how shareholders vote on pay after the special awards.
- 4 November 2026 Q1 fiscal 2027 results: Whether operating cash flow recovers as inventory converts and adjusted gross margin reaches the 39.5% to 41.5% guide, the pre-registered next test.
- February 2027 Q2 fiscal 2027 results: The first quarter in which PhotonLink revenue is meant to ramp, and whether datacenter growth closes the gap to its market.
- August 2027 Q4 fiscal 2027 results and the fiscal 2027 10-K: Whether a quarter above $3B of revenue, gross margin above 42% and operating expenses at 18% of revenue arrive as targeted.
Since we wrote this
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 1 OctThesis reviewA buyer that builds its own switch added. View heldNo print or filing since the thesis. Google designs its own optical switch, one of the new lines Coherent's growth plan needs; the shares have repeatedly fallen with the optics group on doubts about cloud spending.
- 8 SepThesis writtenUnattractive, medium convictionWritten from the fiscal 2026 annual report and an independent ranking that put a Chinese rival ahead of Coherent, in a record year that produced almost no cash.
Also checked 4 times with nothing new to act on.
Sources 26
Sources - 26
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-08-14the business description and the industry-standard-products statement, the two-segment realignment and the segment profit reconciliation, the income statement and cash flow statement, the balance sheet, the major-customer note, the NVIDIA agreement and private placement notes, the earnings-per-share reconciliation and share counts, the tax note and cash taxes paid, the export-control inquiry, the competition section, and the risk factorsbacksC01C02C03C07C09C11C12C13C14C15C16C17C18C19C20C22C23C25C26C32C33C38C39C40C42C43C44C45C61C63C64C68
- S02primarycompany filing2026-08-12the Q4 headline figures on both bases - GAAP gross margin 38.5% against non-GAAP 40.2%, GAAP earnings per share $1.19 against non-GAAP $1.74 - and the fiscal 2027 first-quarter outlookbacksC09C14C31
- S03primarytranscript2026-08-12the indium phosphide capacity constraint and doubling plan, the 6-inch cost structure, the backlog and long-term-agreement descriptions, the greater-than-$3B quarterly revenue target, the 42% gross margin and 18% operating expense targets, the addressable-market figures for optical circuit switching, co-packaged optics and multi-rail, and the segment growth rates. Fifteen analyst turns from eight analysts - Blayne Curtis, George Notter, Vivek Arya, Karl Ackerman, Meta Marshall, Michael Mani, Ryan Koontz and Simon Leopold - which are a separate origin from the answers
Our data center transceiver demand is absorbing every bit of capacity. That we have and then some. ... Indium Phosphide capacity continues to be our primary constraint. ... we have got 800G is still growing very robustly on a year over year basis, and then 1.6T is ramping incredibly fast. ... fiscal 27 is basically completely booked out We are booked really through the end of calendar 2027. ... they also have sort of minimum demand guarantees from our customers or sometimes you refer to those as take or pay agreements. ... we now expect to achieve our first quarter with over $3 billion of revenue by the end of fiscal 27. ... We remain on track to double our internal Indium Phosphide output capacity year over year by the end of the current quarter, 1 quarter ahead of our original plan. ... we remain on track to more than double our internal Indium Phosphide output capacity again by the end of calendar 27. ... what we are seeing now is, you know, customers now looking into calendar 2028. ... I talked about the 80% year over year growth that we saw in laser production. In our June quarter
backsC08C26C27C28C29C30C31C44C46C58C62 - S04secondaryvendor data2026-09-08Research pack frozen beside this thesis, 2026-09-08content/COHR/research-pack.jsonprice and the 252-day return, market capitalisation and enterprise value, the trailing valuation ratios against their six-year medians, the analyst split, the twice-monthly short-interest settlement, the Form 4 insider record, the measured price reaction to each news item, the theme co-movement series, and the reported-against-adjusted earnings-per-share gap for each printbacksC14C33C35C36C37
- S05primaryindustry data2026-05-31the independent supplier ranking and the statement that Coherent was passed into third place, the two Chinese rivals' revenue and growth, the estimated growth of the total transceiver market, the Ethernet share of it, and the explicit caveat that the ranking counts transceiver sales alone
Innolight's revenue in 2024 reached $3.3 billion, a 114% increase year-on-year. The company reported $5.3 billion in revenue for 2025 - up 61% and just above of our estimates of 58% growth in the total transceiver market last year. The company is the leading supplier to all the major cloud companies - the hottest segment of the optical transceiver market. It reported 207% growth in revenues for Q1 2026, compared to Q1 2025, benefiting from a sharp ramp in shipments of 1.6T transceivers. Eoptolink reached record revenues in 2024 at $1.2 billion and again in 2025 at $3.5 billion - up 175% and 189%, respectively. The company is a key supplier of 400G and 800G optics to Amazon, and it has qualified with other large companies in the U.S., including Nvidia. The company reached the #2 rank for the first time in 2025, surpassing Coherent. ... The ranking accounts for shipments of Ethernet, DWDM, FTTx and wireless fronthaul transceivers, but the Ethernet market accounted for 75% of the total market in 2025. ... Please note that the list accounts for sales of optical transceivers only, excluding all other products, which explains lower rankings of Coherent and Lumentum.
backsC04C05 - S06primarycompany filing2026-07-23one buyer's own filed capital spending, as a measurable rate for the driver rather than a forecast: purchases of property and equipment for the six months to 30 JunebacksC41
- S07primaryother2026-09-08NVIDIA Silicon Photonics product page - co-packaged optics switches NVIDIA Corporationthe attacker's own description of what its co-packaged optics switches replace, the named first adopters, and the production status
Replacing pluggable transceivers with silicon photonics on the same package as the ASIC, NVIDIA CPO innovations provide 5x better power efficiency and higher network resiliency resulting in 5x sustained AI application runtime. ... Design simplicity and power efficiency drives CoreWeave, Lambda, Meta, Microsoft, and Oracle Cloud Infrastructure as first adopters. ... Spectrum-X Ethernet Photonics Ramps to Full Production
backsC07 - S08primarycompany filing2026-08-12the quarterly revenue split by segment in dollars, which appears in neither the 10-K nor the release text, and the line-by-line GAAP to non-GAAP reconciliation of gross margin and operating expensesbacksC47C48
- S09primaryother2022-08-22a buyer's own engineers describing the optical circuit switch they designed and run in production themselves, and the transceivers they co-designed for it
implementation details of our internally developed 3D MEMS-based OCS
backsC49 - S10tertiarynews2026-07-28the objection stated on the day of the July 2026 sell-off, in the coverage's own words
The catalyst is growing investor skepticism about whether hyperscaler AI capex is sustainable and whether richly valued optics suppliers can keep re-rating higher
backsC51 - S11tertiarynews2026-09-14the objection stated on the day of the 14 September 2026 sell-off
Weekend AI lab commentary calling for slower frontier development hit NVIDIA's optical supply chain
backsC52 - S12tertiarynews2025-08-14the objection stated the day after the fiscal 2025 results
fiscal fourth-quarter non-GAAP operating margin came in at 18.0%, narrowly missing the FactSet consensus estimate of 18.2%
backsC50 - S13secondaryvendor data2026-09-30the daily closes around each repricing episode: the session moves, the 2 June 2026 high of $426.89, the 29 July 2026 low close of $222.05 and the 30 September 2026 close of $287.81backsC50C51C52
- S14primarycompany filing2026-02-23IPG Photonics Corporation Form 10-K for the fiscal year ended December 31, 2025 IPG Photonics Corporationa rival's own filed list of who competes in materials-processing lasers, naming Coherent beside Trumpf and two Chinese fiber-laser makers, and its statement that competitors are cutting pricesbacksC53
- S15primaryindustry data2025-12-05Cignal AI press release - Optical Circuit Switching Market to Exceed $2.5B in 2029, 5 December 2025 Cignal AIan independent research firm's 2029 size for the external optical circuit switch market and its estimate of the leading rival's share; the free release of a paid report, read through ir_fetch
Cignal AI forecasts the external OCS market to grow beyond $2.5 billion in 2029 - approximately 40% higher than the initial January 2025 forecast. ... Lumentum has provided public guidance to roughly $100 million per quarter in OCS revenue by the end of 2026, implying the company could account for around half of the OCS market next year if current trajectories hold.
backsC54C55 - S16primaryindustry data2026-02-02the raised 2029 outlook and the statement that most optical circuit switch spend stays captive to Google's own designs through the decade; read through ir_fetch
Cignal AI now projects that the 2026 addressable market is roughly three times larger than its earlier estimate, with the 2029 outlook more than 40 percent higher than the forecast published in December. ... most near-term spend will remain captive to Google's internal designs, with a smaller but rapidly growing portion available to third-party vendors.
backsC54 - S17primarytranscript2026-08-11Lumentum Holdings Q4 fiscal 2026 earnings call transcript, held 2026-08-11 Lumentum Holdings Inc.the rival optical circuit switch supplier's own revenue run rate, capacity plans and claim to lead the merchant market; read through transcript_fetch (FMP)
After doubling shipments from fiscal Q3 to Q4, our guidance includes our first triple-digit OCS revenue quarter. The demand signal for 2027 continues to be incredibly strong, and we have started the initial work to add capacity with contract manufacturers ... Yes, Papa, look, what I'd say is we're definitely tracking to the $400 million. ... So I would expect sometime in early '27 that we cross over and be the #1 supplier ... So we're really the only merchant supplier today of OCS.
backsC55 - S18primaryindustry data2025-12-31LightCounting, '2025 was the year of CPO' - December 2025 AOCs, DACs, Linear Drive Pluggable and Co-Packaged Optics report LightCountingan independent analyst's 2030 value and port count for co-packaged optical engines; the free summary of a paid report, read through ir_fetch
We expect that both Broadcom and Nvidia will introduce scale-up switches, GPUs or XPUs with CPO in 2026. Marvell plans to have it as well, using Celestial AI's technology. Shipments of these products will start in 2027. The market value for CPO engines, including scale-up and scale-out applications, is now projected to reach $10 billion with close to 100 million of CPO ports shipped in 2030.
backsC56 - S19primaryindustry data2026-05-31the adoption assumption under LightCounting's co-packaged forecast and the upside it names; read through ir_fetch
Most of the market growth is expected from applications of NPO/CPO in scale-up networks, yet this forecast is based on an assumption that only 10-15% of all scale-up interconnects will migrate to NPO/CPO by 2030-2031.
backsC56 - S20secondarynews2026-09-07Semiconductor Today, 'As AI models scale 100x every two years, Yole maps shift from compute to communication', 7 September 2026, reporting Yole Group's 'Co-Packaged Optics for Data Centers 2026' Yole Group, as reported by Semiconductor Todaya second analyst's co-packaged engine forecast, an order of magnitude above LightCounting's; Yole's own pages answered HTTP 202 to ir_fetch, so this is the trade press's report of it
Yole's new market & technology report 'Co-Packaged Optics for Data Centers 2026' quantifies that shift: CPO optical-engine revenue is forecast to grow from $0.6bn (2026) to $112.1bn (2031), with scale-up applications alone expected to capture about 94% of that market.
backsC57 - S21primarycompany release2026-09-21the company's engagement counts and revenue timing for its integrated optics platform; GlobeNewswire refuses and no 8-K carries it, so read through ir_fetch from a verbatim copy
Coherent has more than ten CPO customer engagements and more than ten NPO engagements, with anchor customers and long-term agreements secured for both architectures. The company also has more than five engagements in emerging chip to chip connectivity. Revenue from the PhotonLink platform is expected to begin ramping in Q4 of CY2026.
backsC59 - S22secondaryanalyst2026-09-23Futurum Group, 'Coherent PhotonLink Raises Its TAM to $90 Billion. Will Bundling Win CPO?', 23 September 2026 Futurum Groupan analyst firm's account of what management said at the launch event: the raised 2030 addressable market, the two laser agreements and the content per chip; read through ir_fetch
Anderson told the ECOC audience that PhotonLink adds roughly $30 billion of integrated optics opportunity to the company's 2030 addressable market, raising the total from $60 billion for the existing portfolio to $90 billion. ... The company disclosed 2 long-term agreements with leading hyperscale AI datacenter customers for its ultra-high-power laser platform, spanning both CPO and NPO, and put its content opportunity at up to $15,000 per 100 Tbps switch or compute chip on 200G optical lanes ... These TAM and content figures remain vendor-supplied estimates that procurement behavior has yet to validate.
backsC59 - S23primarycompany filing2025-10-02beneficial ownership of directors and officers and of holders above 5% as of August 31, 2025, the fiscal 2025 bonus metrics, goals and payout, the long-term award design and the 2023 PSU payout, the fiscal 2026 bonus metrics, and the officers' start datesbacksC64C65C66
- S24primarycompany filing2026-08-31the target values, share-price hurdles, relative-return condition, four-year performance period and one-year holding period of the special awards granted August 27, 2026backsC67
- S25primarycompany filing2026-02-11Schedule 13D/A (Amendment No. 5) filed February 11, 2026 by Bain Capital's BCPE Watson (DE) BML, LP on Coherent Corp. common stock BCPE Watson (DE) BML, LP (Bain Capital)Bain's sale of its last 9,437,238 shares in a block trade on February 9, 2026, leaving it with nonebacksC65
- S26secondaryvendor data2026-10-04Earnings calendar saved by the daily routine, content/calendar/2026-10-04.json (FMP per-symbol earnings rows) Financial Modeling Prepthe date of the next print, Q1 fiscal 2027
"COHR": {"archived": false, "next": {"date": "2026-11-04", "eps_actual": null, "eps_estimated": 1.97
read, but no claim rests on it