Coherent Q4 FY2026 Results (12 August 2026): $2.05bn Revenue Within $5m of the Guide Ceiling, EPS $1.74 Above It — and Operating Cash Flow Fell to $79.5m

Coherent reported its first $2 billion quarter after the close on 12 August: revenue of $2,045.5 million, within $5 million of the top of its own guided range, and non-GAAP earnings of $1.74 per share, two cents above the ceiling of that range. Nothing in the reported quarter missed. The number that deserves attention is on a different statement entirely — full-year operating cash flow of $79.5 million, down from $633.6 million, against $1,102.9 million of capital expenditure. This is what it looks like when a company funds a demand curve before the demand pays it.

What actually happened

The fourth-quarter and full-year fiscal 2026 release landed after the New York close on Wednesday 12 August, covering the quarter ended 30 June 2026.

The full year is the cleaner picture of the transition. Revenue of $7,118 million was up 22.5%; GAAP operating income reached $898 million on a 12.6% margin against $290 million and 5.0%; GAAP diluted EPS was $4.12 against a prior-year loss of $0.52; non-GAAP operating margin was 20.5%, up 262 basis points, and non-GAAP EPS was $5.61 against $3.53.

Chief executive Jim Anderson's framing in the release was that "Fiscal 2026 was an outstanding year for Coherent, with record revenue, significant margin expansion, and non-GAAP EPS growth that was more than twice the rate of revenue growth." That last clause is arithmetically correct and worth isolating: revenue grew 22.5% and non-GAAP EPS grew 58.9%. Operating leverage of that shape comes from mix and from holding operating expense growth below revenue growth, and both happened.

Of the three ways this print could have read, the one that landed was the first — capacity confirming, with mix improving as it went. Revenue at the ceiling, margin up 215 basis points, and a forward guide that steps up rather than consolidates. What the scenario map did not anticipate is that the strain would show up somewhere the income statement does not reach.

One segment is now effectively the whole company

The segment table is the most under-reported part of this release.

Datacenter and communications was 79% of fourth-quarter revenue, against 67% a year earlier. The industrial segment — lasers for materials processing, instrumentation, automotive — shrank for the year. So the consolidated 33.8% growth rate understates what is happening inside the business: one segment grew 58.6% while the other contracted 15.8%, and the blended figure is the average of a boom and a mild recession happening in the same company.

This matters for how the print should be read across. When Coherent's revenue growth accelerates, that is not a photonics cycle turning; it is AI infrastructure spending arriving through a single door while the industrial economy the company also serves stays soft. Anyone treating the consolidated number as a read on general manufacturing demand has it backwards. On the call, management guided first-quarter datacenter growth to exceed 80% year on year — which would push the mix further still, and further reduce whatever diversification the industrial book was providing.

The bottleneck this post flagged has moved — from Coherent's own fab

The original version of this piece argued that the constraint nobody previews sits below the module: indium phosphide, the compound substrate every competitive high-speed transceiver is built on, and where China's 2025 export licensing regime made the licence queue rather than demand the near-term governor of supply. It named a substrate supplier's capacity schedule as the thing to watch.

The answer arrived from inside Coherent instead. On the earnings call, management said June-quarter indium phosphide production rose approximately 80% year on year, and that the company expects to double internal output year on year by the end of the current quarter — one quarter ahead of plan. Chief financial officer Sherri Luther put the economics plainly: a 6-inch wafer provides four times the output of a 3-inch wafer at half the cost, and 6-inch yields now exceed 3-inch performance across multiple product types.

That is a genuine change in the picture rather than a restatement of it. A vertically integrated supplier that makes its own substrate is not standing in the licence queue for the part that matters most. It also reframes the trade-policy argument in the section below: the layer everyone assumed was the shared chokepoint is being widened unilaterally by one of the participants, at four times the output per wafer and half the cost.

Record earnings, $79.5 million of operating cash flow

Here is the part the guide did not describe and the headline did not carry.

GAAP net earnings for the year were $805 million. Operating cash flow was $79.5 million. The gap is working capital, and the balance sheet says exactly which line: inventories rose $1,143.4 million, receivables another $379.2 million. Against that, payables rose $1,058.4 million — meaning suppliers are financing a large share of the inventory build rather than Coherent funding all of it from cash. That is a meaningful mitigation, and it is also a dependency: it works while suppliers are willing to extend terms into a ramp they can see the orders behind.

The mechanism is not mysterious and it is not a warning sign by itself. A company whose fiscal 2027 is essentially fully booked, with purchase orders running through calendar 2027 and some customers ordering into calendar 2028, has to buy the substrate, build the wafers, staff the lines and stock the parts before any of that revenue is recognised. Inventory rising 79% ahead of a guide that steps revenue up as much as 17% sequentially is what pre-funding booked demand looks like on a balance sheet.

The funding came from outside the operating business. The cash flow statement shows $1,998.6 million of proceeds from the issuance of common shares, $1,250.0 million drawn on a Term A facility, $437.0 million from the sale of a business, and $1,723.4 million of payments on existing debt. Total equity rose to $11,238.2 million from $5,998.0 million, the $2,483.3 million of mezzanine equity on the prior-year balance sheet went to zero, and long-term debt ended slightly lower at $3,214.3 million. The company finished the year with $1,162.0 million of cash and $825.0 million of short-term investments.

So the honest summary is that fiscal 2026 was a record earnings year funded by the capital markets rather than by operations, deliberately, in order to build capacity against an order book. Whether that was the right call is answered in fiscal 2028, when the capacity either converts the backlog at the guided margin or does not. What can be said now is that the risk has changed shape: a year ago the question was whether Coherent could get enough substrate; now the question is whether $1.1 billion of annual capital expenditure and $2.6 billion of inventory find the revenue that was booked against them.

This is the same pattern running through the equipment and memory complex, where committed capacity arrives years after the demand that justified it — visible in the Samsung and SK hynix fab commitments, and in Applied Materials' free cash flow falling 80% while earnings hit a record — a company reporting the day after Coherent, with the same shape on the same statement.

Lumentum set the bar, and Coherent's guide answered it

Lumentum reported one day earlier: net revenue of $1,006.3 million, up 109.3% year on year, with non-GAAP gross margin of 50.4% and non-GAAP EPS of $3.23, and a first-quarter guide of $1.225–1.275 billion against consensus nearer $1.16 billion. The forward guide was the read-across that mattered, and the question left for Coherent was whether it would step up in the same direction.

It did. Guiding $2.2–2.4 billion against $2,045.5 million delivered is a 7.6% to 17.3% sequential increase, and the stated ambition of exceeding $3 billion in a quarter by the end of fiscal 2027 implies roughly 47% growth from the quarter just closed. Two overlapping suppliers to the same customers both guiding materially above their just-reported run-rates is information about the end market rather than about share — which is the one configuration that resolves the ambiguity the preview flagged. If only one had stepped up, the read would have been about qualification wins at particular customers. Both stepping up says the demand is broad.

The gap between them is margin structure, and it is wide: Lumentum at 50.4% non-GAAP gross margin against Coherent at 40.2%. That difference is product mix and vertical position, not execution. It is also why the two names carry different sensitivity to the module-versus-component question in the callout above.

Why a clean beat did not hold the tape

Coherent cleared its guide on revenue, beat the top of its own EPS range, expanded margin, and guided up. Shares fell roughly 6% in pre-market trading on 13 August, after rising about 9% on the 12th into the print.

That is not a contradiction, and it is the most transferable lesson in the release. The published guide is a bar set by the company; the bar that determines the share price reaction is set by whoever bought the stock in the run-up, and it is never published. Coherent had roughly doubled year-to-date before this week. A stock that rises 9% on the day of its own earnings has already priced a good print, so a good print pays nothing and anything short of exceptional costs something. The original version of this piece noted that pre-earnings position-trimming was "context for how steep the unpublished bar is relative to the published one, not a signal about direction." The direction resolved downward on a quarter with no miss in it.

The reasonable candidates for what the unpublished bar contained are the flat-to-slightly-higher Q1 gross margin guide against a large volume step-up, and the cash statement. Both are legible in the release. Neither is a deterioration in demand.

What it touches, and what it does not

Coherent and Lumentum both joined the S&P 500 in the March 2026 rebalance, Coherent effective 23 March. The optical complex trades as a high-beta expression of AI infrastructure spending, which is why these names appear in NAS100 and US500 commentary far in excess of their index weights — and why a 6% move in a company of this size registers in sector sentiment more than in index arithmetic.

The read-through travels further than the shares. Optical component makers sit between hyperscaler capital expenditure and the physical layer that spends it, which makes their order books and capacity commentary one of the few near-real-time reads on whether announced data-centre plans are converting into shipped hardware — the same question Cisco's fourth quarter addressed from the systems side on the very same evening. An order book booked through calendar 2027 with orders arriving for 2028 is a stronger version of that signal than any capex announcement, because a purchase order is a commitment and a press release is not.

What it is not is a currency event. There is no honest channel from a photonics fiscal quarter to the relative pricing of the majors, and the one real link runs the other way: revenue and costs booked in non-dollar currencies are translated at prevailing rates, so where the US dollar has been flatters or trims reported figures with no change in units sold. That dollar is scored on rates, growth, positioning, risk and commodities — five factors, eight currencies, a different clock entirely. More on how that framework fits together on the about page.

What would change the picture

Three things, in descending order of how much they would move the interpretation.

Cash conversion in the first half of fiscal 2027 is now the central question. Inventory of $2,581.0 million against a guided quarterly revenue of $2.2–2.4 billion means the company is carrying roughly a quarter's worth of revenue in stock. If revenue steps up as guided and inventory flattens, operating cash flow recovers mechanically and the fiscal 2026 figure reads as a one-year investment. If inventory keeps rising at this rate alongside capex that management has said increases sequentially again in Q1, the balance-sheet funding requirement continues, and that is a different company from the one the income statement describes.

The Q1 gross margin outturn against the 39.5–41.5% guide will settle the module-versus-component question. Margin holding near 40% on revenue up double digits sequentially would confirm that lower-margin assembled module content is growing faster than component content — gross profit dollars rising while the percentage stalls. Margin at the top of the band on that volume would mean the opposite, and would be the more valuable outcome.

And a published FCC rulemaking with defined scope would still convert a headline into a modellable event. The two definitional questions remain open — what counts as a "new" module, and whether a restriction follows the company or the country of manufacture — and they contain almost all of the economic content. Cignal AI's assessment, which this piece has cited from the start, is that non-Chinese suppliers lack the module throughput to fill the gap and would need years rather than quarters; it also notes that transceiver modules were excluded when tariffs were imposed last year. Coherent's substrate expansion changes the layer beneath that argument without changing the argument itself, because the constraint a trade restriction targets is assembly capacity, and that is not what Coherent just added.

Educational macro context only — not investment advice.