Every number was a record. The stock still went dark.
Coherent just posted its first $2 billion quarter, beat on both lines and guided above consensus — and the shares trade 38% below their June high, under the 200-day, and below the lowest of 23 published price targets. The market is not disputing the earnings. It is disputing one thing: whether an indium-phosphide shortage is a moat or a rent. Six analyst lenses, three scenarios, four horizons.
Twenty months of history, five years of dispersion
One continuous axis. To the left of the grating, the actual tape: a collimated beam from $94.73 in January 2025, through the April 2025 crash to $45.58, up more than nine-fold to $440.00 on June 3, 2026, and back to $272.03. To the right, the beam disperses into three scenarios. Log scale — the range spans 20×.
History plotted through dated closing prices sourced in Phase 1 (see footer); the line interpolates between those anchors and is not a full daily series. Forward paths are illustrative frameworks built from the EPS ladder in section 09 times an exit multiple — not forecasts. Horizons are anchored to Coherent’s June fiscal year-ends.
Disagree with the weights? Move them.
The desk runs 30% bear / 45% base / 25% bull — deliberately bear-tilted because free cash flow is negative and the bottleneck that creates today’s pricing power is being dismantled by Coherent itself. Drag the sliders: the blended target, the dotted line on the chart above, and the clay row inside every scenario card all recompute live.
Scenario probability re-weighter
Bear and bull are yours; base is whatever is left over. The pair is capped at 100%.
Eight numbers that frame the argument
Reported August 12, 2026. Revenue beat the $1.98bn consensus; non-GAAP EPS of $1.74 beat $1.62. Then the stock fell 23% over the next fourteen sessions.
Sources: Coherent Form 8-K exhibit 99.1, August 12, 2026 (revenue, margins, EPS, segments, cash flow); S&P Global Market Intelligence via StockAnalysis, checked September 2, 2026 (forward P/E). Net leverage as stated by management on the Q4 FY26 call.
Six ways to read the same photon
Each lens was reasoned through on its own terms before reading the others. They land between $175 and $455 — a spread wider than the stock’s entire market cap. The average, $323, sits 19% above spot and 22% below the street.
The photon budget
Every incremental AI cluster converts copper to optics, and Coherent owns the scarcest input at the exact moment the industry doubles lane rates. Consensus FY28 EPS of $13.95 is a floor, not a ceiling.
- Datacenter & comms +58.6% y/y in Q4 and accelerating from +40.6% in Q3; 79% of revenue.
- Q1 FY27 guided to $2.2–2.4bn vs $2.13bn consensus; management targets a first $3bn+ quarter by end of FY27.
- Order coverage through calendar 2027 called “exceptional,” with customer orders extending into calendar 2028.
- Non-GAAP opex is already below the 18%-of-revenue target at the Q1 guide midpoint — EPS compounds faster than revenue.
The cash ledger
A company that generated $80m of operating cash on $7.1bn of revenue, burned $1.0bn of free cash and diluted holders 26% is not yet a quality compounder. It is a capital-intensive supplier being paid in inventory.
- Operating cash flow fell 87.5% to $79.5m; free cash flow −$1,023m; FCF yield −1.9%.
- Inventory rose 79.5% to $2,581m; working capital consumed $1,102m in twelve months.
- ROIC of 6.74% sits far below a 15.14% WACC: the P&L compounds, the economics do not.
- Credit where due: leverage 2.0× to 0.7×, tangible book per share from −$13.06 to +$18.61, Altman Z of 4.1.
The allocation rent
Margin expansion is rent on an indium-phosphide shortage that Coherent is spending over $1bn a year to eliminate. When the bottleneck clears, price and mix normalise into a far bigger fixed-cost base.
- The moat is the shortage: internal InP output doubles year-over-year by September 2026, a quarter ahead of plan. Supply relief is self-inflicted.
- Silicon photonics already covers 50–70% of the 800G/1.6T market and routes around InP; LPO strips the DSP and cuts 800G module power to roughly 10W.
- Share is moving the wrong way: InnoLight led 2025 global transceiver shipments near 23.4% against Coherent close to 16% (LightCounting).
- NVIDIA is shareholder, anchor customer and holder of capacity rights to 2030 — that is leverage over Coherent, not for it.
Vertical integration
The advantage is not the module. It is owning the InP epitaxy, the laser, the silicon-carbide substrate and the packaging under one roof — the one thing a module assembler cannot import around if Washington keeps tightening.
- Coherent sits in the 200G-per-lane EML oligopoly that literally gates 1.6T delivery schedules industry-wide.
- Six-inch InP yields in Texas and Sweden already exceed the legacy three-inch lines across all devices — a cost step-down, not just capacity.
- Regulatory asymmetry: a proposed FCC restriction on Chinese transceivers would transfer share to the US number two.
- Threat to the moat: CPO migrates dollar content toward switch silicon, leaving Coherent selling lasers rather than modules.
The broken tape
Price sits below the 50-day and the 200-day, relative strength is 42, and the stock has round-tripped 38% from its June high. A high-beta momentum name in a confirmed downtrend, however good the print was.
- $272.03 against a 50-day of $314.44 and a 200-day of $274.45 — the short average is above price and rolling over.
- Beta 2.10. Shares fell 23.5% in the fourteen sessions after an 8.3% beat-and-raise pop to $355.72 on August 12.
- Peak-to-trough from June 3 ($440.00) to July 29 ($222.05) was −49.5%. Realised volatility is extreme in both directions.
- No squeeze fuel: short interest is 3.34% of shares at 1.5 days to cover, and insiders sold only — the CFO sold 3,000 shares at $324 on August 18.
The capex derivative
Coherent is a levered derivative of hyperscaler capital spending and US–China trade policy — two variables that have nothing to do with Coherent’s execution and everything to do with its multiple.
- August’s drawdown was sector-wide and headline-driven — rising Treasury yields and AI-capex payback doubts, not company news.
- Policy cuts both ways: an FCC ban on Chinese transceivers is a large upside catalyst; germanium, gallium and yttrium controls inflate input costs.
- At beta 2.10 and 28.9× forward, duration is long — a 100bp move in the ten-year is worth several multiple points on its own.
- S&P 500 membership cuts both ways too: a passive bid on the way up, forced selling in a de-rating.
These lenses are synthesized analytical frameworks written for this report. They are not real individuals, not real firms, and not real ratings. Section 06 shows what actual sell-side analysts publish.
The stock is trading below the lowest target on the street
Consensus is $415.36, a median of $420, a range of $280 to $500, and a rating split of 14 strong buy / 4 buy / 5 hold / 0 sell. Every one of those targets was raised after the August 12 print. At $272.03 the market is trading 2.9% below the most bearish published number — when a stock sits under the entire distribution, either the distribution is stale or the tape knows something.
Individual targets and ratings via TipRanks / S&P Global Market Intelligence, dated August 13–31, 2026: Northland (market perform, raised from $230), B. Riley (neutral, from $309), Morgan Stanley (equal weight, from $330), Deutsche Bank (buy, initiated Aug 31), Jefferies (from $375), Needham (from $380), Raymond James (strong buy, from $435), J.P. Morgan, Rosenblatt (from $425). William Blair maintains a hold with no published target. Aggregate consensus checked September 1, 2026.
What each path is actually worth
Every price below is an EPS number times an exit multiple; both are stated in the assumptions drawer. The clay row is live — it follows the sliders in section 03. Bars are normalised within each card so bull reads full width.
1-year
JUN 2027 · FY272-year
JUN 2028 · FY283-year
JUN 2029 · FY295-year
JUN 2031 · FY31▸ Show the assumptions & math
Anchor. Spot $272.03 (Sep 1, 2026 close). Diluted share count ~202m and rising; every EPS figure below is non-GAAP diluted, consistent with how Coherent and the street quote it. Reported history: FY25 $3.53, FY26 $5.61. Consensus: FY27E $9.41 (range $8.26–$10.26, 18 analysts), FY28E $13.95.
Bear — 30%. “The rent expires.” Coherent’s own six-inch InP ramp, plus silicon photonics and LPO substitution, clears the bottleneck through FY28. Pricing normalises, industrial stays weak, and the multiple compresses toward the high teens that component cyclicals earn.
Base — 45%. “The ramp holds, the multiple normalises.” Consensus is roughly delivered through FY28, then growth decelerates to the mid-teens as 1.6T matures. Free cash flow inflects positive in FY28 as the inventory build stops. Multiple settles in the mid-20s to low-30s.
Bull — 25%. “The bottleneck is the moat.” Optical circuit switching, co-packaged optics, multi-rail and silicon-carbide thermal substrates stack on top of transceivers rather than replacing them. Coherent holds its EML position, converts the $15bn integrated-optics TAM, and keeps a premium multiple.
Blend at desk weights (30 / 45 / 25).
Why bear-tilted. The default 25/50/25 was shifted to 30/45/25 for one reason: the asset that produces today’s pricing power — scarce indium phosphide — is being deliberately doubled by Coherent itself, on management’s own schedule, a quarter ahead of plan. Combined with −$1.02bn of free cash flow and 26% annual dilution, the downside distribution is fatter than a simple growth screen implies. Note the blend still lands 30% below the street’s $415.36 one-year consensus.
These are illustrative frameworks, not forecasts. They exist to make the arithmetic of the debate explicit and falsifiable — if you disagree with an exit multiple, the drawer shows you exactly which number to change.
The year the cash flow inverted
Revenue is deliberately not on this chart: at $7.1bn it would flatten every cash-flow bar into a line. What matters is the crossing. For three years operating cash flow ran comfortably above capital spending. In FY26 they swapped places — capex $1,103m against operating cash flow of $79.5m — and free cash flow went to −$1.02bn. Debt, meanwhile, kept falling. That is the whole argument in one picture.
FY23–FY26 are reported figures; free cash flow is operating cash flow minus capital expenditure on a consistent basis (FY26: $79.5m − $1,102.9m = −$1,023m, per the August 12, 2026 8-K). FY27E is estimated: free cash flow of roughly −$0.55bn is in line with the −$0.57bn consensus levered figure from S&P Global; capex of ~$1.9bn is this desk’s estimate, anchored to management’s guidance that capital spending rises sequentially again in Q1 FY27; operating cash flow of ~$1.35bn is the arithmetic consequence of those two. Debt for FY27E assumed roughly flat. Treat the FY27E group as a sketch, not a report.
Every price target in this report is one of these bars times a multiple
Non-GAAP diluted EPS, fiscal years to June. Two reported, two consensus. The street’s $415.36 is roughly FY28E $13.95 at 30×; this desk’s blended one-year $293 is FY27 at about 30×. The entire debate is whether the two clay bars arrive.
Reported figures from Coherent’s Q4 FY2026 release, August 12, 2026 (table 8, non-GAAP diluted EPS). FY2027E and FY2028E are S&P Global Market Intelligence consensus, last updated August 31, 2026. Coherent’s Q1 FY27 guidance alone is $1.85–$2.05, against $1.00 in the year-ago quarter.
The operating business is accelerating. The cash flow is collapsing.
Latest year-over-year growth, sorted low to high. The two terracotta bars at the top are the bear case in numerical form; everything below the zero line is the bull case. Both are true at the same time — which is precisely why the stock is 38% off its high while the sell side is unanimously positive.
Q4 FY26 year-over-year unless stated. Non-GAAP gross profit $822.6m vs $582.2m; non-GAAP operating income $445.8m vs $275.1m; non-GAAP EPS $1.74 vs $1.00; datacenter & communications $1,615.0m vs $1,018.3m; industrial $430.5m vs $511.1m — all per the August 12, 2026 8-K. Operating cash flow is full-year FY26 ($79.5m vs $633.6m). InP laser output growth and the FY27E consensus revenue figure are from the Q4 FY26 earnings call and S&P Global respectively.
Is the shortage a moat, or a rent?
Both columns are built from the same August 12, 2026 filing. That is the point: this is not a disagreement about facts, it is a disagreement about which facts are durable.
The bull case
- A seventh consecutive record quarter, and the first $2bn one. Q4 revenue $2,045.5m, up 33.8% year-over-year and 13.3% sequentially, with datacenter & communications up 58.6% — an acceleration from +40.6% in Q3.
- Beat and raise on both lines. Revenue topped the $1.98bn consensus, non-GAAP EPS of $1.74 beat $1.62, and Q1 FY27 was guided to $2.2–2.4bn revenue and $1.85–$2.05 EPS — each above the street.
- The demand is contracted, not hoped for. Management called order coverage through calendar 2027 “exceptional” with orders extending into calendar 2028, on top of NVIDIA’s multibillion-dollar purchase commitment and capacity rights running to 2030.
- Margins are structural. Non-GAAP gross margin has risen in eight of the last nine quarters to 40.2%; SG&A fell 205bps to 8.2% of revenue; FY26 non-GAAP EPS grew 58.9% on 22.5% revenue growth.
- The balance sheet was repaired in a single year. Net leverage 2.0× to 0.7×, $513m of debt repaid, tangible book per share from −$13.06 to +$18.61, $1.99bn of cash and short-term investments.
- New revenue stacks on top of transceivers. Optical circuit switching ($4bn TAM), integrated optics and CPO ($15bn incremental), multi-rail ($2bn by 2030), PhotonLink unveiling at ECOC on September 21, 2026, and 300mm high-thermal-conductivity silicon carbide now sampling to AI chip partners.
- The price has decoupled from every published estimate. $272.03 is below the lowest of 23 sell-side targets, at 28.9× forward earnings against Lumentum’s roughly 50× — and only 6% above the $256.80 NVIDIA itself paid in March 2026.
The bear case
- The profit boom produced no cash. FY26 operating cash flow fell 87.5% to $79.5m and free cash flow was −$1,023m — against $805m of reported net earnings. Cash conversion was effectively zero.
- The gap is inventory. Stock on hand rose 79.5% to $2,581m and working capital consumed $1,102m in twelve months — into a market where 400G is set to become “mature declining” by 2027 and 800G peaks around 2026–27.
- Capital intensity is exploding. Capex went from $441m to $1,103m and management guided it higher again. ROIC of 6.74% sits less than half of a 15.14% WACC — growth is currently value-destroying on a returns basis.
- Shareholders paid for it. Shares outstanding rose 26.26% in a year through the preferred conversion and NVIDIA’s $2bn placement. Per-share economics now have to run twice as hard for the same result.
- Coherent is dismantling its own moat. It is doubling internal indium phosphide output year-over-year a quarter ahead of plan, while silicon photonics already covers 50–70% of the 800G/1.6T market and LPO removes the DSP to cut 800G power to roughly 10W.
- Share is going the wrong way. InnoLight led 2025 global transceiver shipments near 23.4% versus Coherent close to 16% (LightCounting), and Chinese suppliers account for roughly 60% of datacom transceiver revenue.
- A fifth of the company is shrinking. Industrial revenue fell 15.8% year-over-year in Q4 to $430.5m and 10.3% for the full year — the cyclical, low-multiple half of the story.
- The tape sides with the bears. Down 23.5% in the fourteen sessions after an 8.3% post-earnings pop, below the 50-day ($314.44) and 200-day ($274.45), with insiders selling only — the CFO sold 3,000 shares at $324 on August 18, 2026.
Ten ways this goes wrong
Cells heat with severity: hottest at likely × high. The genuine tail risk sits bottom-right — low odds, and if it lands, it changes what Coherent is rather than what it earns.
- Industrial segment drag
- Working-capital & FCF burn persists
- Capex over-build into a demand plateau
- Chinese module share gains
- Rate-driven multiple de-rating
- Silicon photonics / LPO substitution
- NVIDIA concentration
- AI capex digestion
- Critical-materials export controls
- CPO value migrates to switch silicon
Capex over-build
Capacity commissioned in FY28 lands just as 800G peaks and 1.6T pricing normalises. Fixed costs sit idle, gross margin gives back the 215bps it just gained, and the $1.9bn-a-year spend has no revenue to carry it.
Free cash flow stays negative
Another year of $1bn-plus inventory and receivable build keeps free cash flow underwater, forcing a return to the debt or equity market and undoing the leverage repair from 2.0× to 0.7×.
Industrial keeps shrinking
The segment fell 15.8% year-over-year in Q4 and is now 21% of revenue. It dilutes group growth, drags the blended multiple toward a cyclical rating, and management has no visible catalyst for it.
Silicon photonics and LPO route around InP
Coherent’s pricing power rests on indium phosphide scarcity. Silicon photonics already serves 50–70% of the 800G/1.6T market at high yield, and linear-drive optics strip out the DSP entirely. If the industry designs around the bottleneck, the rent disappears.
NVIDIA concentration
NVIDIA is simultaneously a roughly 4% shareholder, the anchor customer, and the holder of capacity rights through 2030. A roadmap change, a second-sourcing decision, or in-housing the optical engine resets the entire model — and Coherent has little leverage to resist.
AI capex digestion
The August 2026 selloff was driven by exactly this fear. Backlog “into calendar 2028” is only as good as the hyperscaler budgets behind it; orders in this industry have been cancelled before, and $2.58bn of inventory would become the problem.
Chinese module share gains
InnoLight and Eoptolink out-scale Coherent on cost with domestic wafer capacity and shorter lead times. Coherent’s roughly 16% share erodes further at the module level even if it keeps the component position.
Rate-driven de-rating
Beta of 2.10 at 28.9× forward earnings is long duration. A sustained move higher in the ten-year yield compresses the multiple by several turns with no change whatsoever to Coherent’s earnings.
CPO moves the value to switch silicon
The genuine tail. If co-packaged optics wins decisively, the optical engine collapses into the switch package and the dollar content migrates to Broadcom and NVIDIA. Coherent stops selling $2,000 modules and starts selling lasers — a structurally smaller, lower-margin business, no matter how well it executes.
Critical-materials export controls
Germanium, gallium and indium restrictions cut the other way: the same policy environment that could hand Coherent share via an FCC transceiver ban could also choke its own compound-semiconductor substrate supply and inflate input costs.
What the words mean
Terms marked with a dotted clay underline throughout this report are defined here.
- Indium phosphide (InP)
- The compound semiconductor wafer that high-speed transceiver lasers are grown on. Coherent’s scarcest input and, for now, the industry’s binding constraint on 1.6T shipments.
- EML
- Electro-absorption modulated laser. The light source that sets the data rate ceiling in a transceiver. The 200G-per-lane EML supply is an oligopoly of roughly three vendors, Coherent among them.
- Silicon photonics
- Building optical circuits on ordinary silicon wafers instead of exotic compound semiconductors. Cheaper and easier to scale — and the main route by which competitors design around an InP shortage.
- LPO
- Linear-drive pluggable optics. A transceiver design that deletes the digital signal processor chip, cutting power and cost. A threat to vendors whose margin comes from complexity.
- CPO
- Co-packaged optics. Placing the optical engine directly beside the switch chip rather than in a removable module. Saves power, but moves the dollar value toward whoever owns the switch silicon.
- OCS
- Optical circuit switching. Routing light directly between ports without converting to electricity, used to reconfigure AI cluster topologies. Coherent sizes this at a $4bn market.
- 800G / 1.6T
- Transceiver speed generations, in bits per second. 800G is the current volume product; 1.6T is ramping through 2026–27. Each generation roughly doubles the optical content per port.
- Non-GAAP EPS
- Earnings per share excluding share-based compensation, acquired-intangible amortisation, restructuring and one-offs. Coherent’s FY26 figure was $5.61 against $4.12 on a GAAP basis — the gap is real cost, just not cash cost.
- Free cash flow
- Operating cash flow minus capital expenditure — the cash actually left for owners after the business is maintained and expanded. Coherent’s was minus $1.02bn in FY26.
- FCF yield
- Free cash flow divided by market capitalisation. A negative yield, as here at −1.9%, means the company consumed cash rather than generating it for shareholders.
- EV/EBITDA
- Enterprise value (market cap plus net debt) over cash operating profit. It compares the whole business, debt included, rather than just the equity. Coherent trades near 38.5×.
- ROIC vs WACC
- Return on invested capital against the weighted average cost of capital. When ROIC (6.74%) is below WACC (15.14%), each incremental dollar invested destroys value — however fast revenue grows.
- PEG ratio
- The P/E multiple divided by the expected earnings growth rate. Below 1.0 is conventionally cheap. Coherent’s 0.55 says the stock is inexpensive if the growth rate is real.
- Working capital drag
- Cash swallowed by rising inventory and receivables as a business scales. Coherent’s consumed $1,102m in FY26 — the single reason profit did not become cash.
- Pro forma revenue
- Revenue restated to exclude businesses that were sold, so growth compares like with like. Coherent’s FY26 growth was 22.5% reported but 28% pro forma, after divesting aerospace & defence and its Munich tools business.
- Exit multiple
- The P/E the market is assumed to pay at the end of a forecast period. Every scenario price in section 07 is an EPS estimate multiplied by one of these — it is the single most subjective number in the model.