Coherent (COHR): an AI-optics winner with a diamond-cooling option — priced, not cheap

I bought Coherent originally for a thesis I still believe in: synthetic-diamond thermal management — “diamond cooling,” and yes, diamond is the gem name for the same material engineers call synthetic or CVD diamond — would matter more and more as AI data-center power density climbs, and the hyperscalers would chase it. That direction is right. But as a stock in 2026, the diamond story is not what’s driving the price, and conflating “great technology” with “buy at any multiple” is the most common way to lose money on a good company.

Here’s my whole call in one box, then the work behind it.

How to read this. I separate what’s known (the filed numbers, the NVIDIA deal, the product launches) from what’s inferred (how the inventory and order cycle behave) from what’s opinion (my call and sizing). Figures are as of 2026-06-26/27 and come from my own work-up; treat them as a dated snapshot, not gospel. This is me thinking in public, not a recommendation. Size your own risk.

Signal tiering: is COHR an AI-optics company or a diamond-cooling company?

Right now, it’s an AI-optics company — overwhelmingly. That matters because it tells you what’s actually being re-rated.

The first driver of COHR’s re-rating isn’t diamond — it’s AI datacenter interconnect: lasers, InP and optical devices, 800G/1.6T, and the road to CPO/NPO. NVIDIA’s $2B investment and multi-year, multi-billion purchase commitment is the order-visibility anchor underneath the move. The diamond thermal line is real and high-quality, but it is not, today, the estimate the market is paying for.

Is LITE stronger? Pick a metric

The natural question for anyone in optical interconnect: is there a “lite” that’s stronger? There is — Lumentum (LITE) — and on today’s operating metrics it genuinely is. But “stronger operator” and “better stock” are not the same thing once you put the valuation next to the growth.

COHR vs LITE — pick a metrictap to switch

Revenue growthGross marginFree cash flowP / S (FY26E)P / E (FY26E)

COHR

+20.5%

LITEstronger

+90.1%

LITE is the stronger operator today (growth / margin / cash flow); COHR is the larger, cheaper, broader platform. LITE is not a cheaper COHR.

LITE is the higher-beta, purer, faster optical play; COHR is the larger, broader platform that’s cheaper on sales and earnings. Both are richly priced. If the question is “who’s executing harder right now,” it’s LITE. If it’s “where’s the better risk/reward after a pullback,” I lean COHR — because COHR still carries un-financialized options (diamond thermal, CPO/NPO, OCS, multi-material) that LITE’s price has less of.

Cash flow is the real swing factor, not the P/E

This is COHR’s sharpest tension: the revenue and order logic is strong, but the cash flow hasn’t validated it yet.

Line (9-mo FY26)	Figure	Read
Operating cash flow	$10.06M (vs $503.3M a year ago)	Collapsed — almost all of it is working capital.
Capex	$547.2M	Heavy build-out ahead of demand.
Free cash flow ≈	−$537M	The honest weak spot of the bull case.
Inventory	$1.438B → $2.127B	Up sharply “to support higher revenue.”
Receivables	$964M → $1.188B	Growing with the top line.
Cash + ST inv.	$1.593B + $825M	Not financially fragile; total debt ≈$3.194B.

Two readings of the inventory build: good inventory (stocking for AI-interconnect orders) or bad inventory (demand linearly extrapolated, product pulled forward). I lean good-inventory, ~60% — D&C grew +41% and Q4 is guided higher — but it’s very falsifiable. The falsifier: next print shows operating cash flow still deeply negative, inventory north of ~$2.2–2.3B, and D&C decelerating to low single digits sequentially. That would say the order visibility is priced too full.

Diamond thermal: leader, or just one of the leaders? And is anyone using it yet?

The technology is real, not a slide. COHR’s Bondable Diamond (Jan 2026) bonds directly to Si, SiC, GaN, AlGaN, GaAs and InP and claims to cut interface thermal resistance by up to 99%. Its Thermadite 800 cold plates (Mar 2026) run ≈800 W/m·K — roughly twice copper — drop chip temperature by >15°C, and are ≈60% less dense than copper.

But two things keep this an option, not a current earnings line:

• It isn’t a monopoly. Element Six offers CVD diamond heat spreaders and copper-diamond composites targeting hyperscale GPUs, HPC, AI accelerators, CPUs and co-packaged optics, with conductivity quoted up to 1000–2200 W/m·K. COHR’s edge isn’t “highest number on the spec sheet” — it’s the system integration (material + surface prep + bonding + cold plate + existing laser/optics customer relationships). That’s worth a lot if NVIDIA / AMD / Broadcom / hyperscaler ASIC / advanced-packaging chains design it in; it’s worth little if it stays at the launch-deck stage.
• Adoption is ≈ zero-to-low-single-digit today. I can’t find auditable public data putting a percentage on diamond cooling in AI accelerators, and COHR’s own materials put new thermal revenue in H2 2027+, with a 2030 SAM around $2B — smaller than its datacenter interconnect and CPO/NPO pools. Mainstream AI cooling is still copper cold plates, liquid, TIM and package-level spreading; diamond is migrating in from RF, lasers and wide-bandgap, where heat flux is already extreme.

So your original instinct was forward-looking — but on the stock, diamond is a 2027–2030 call option, not a 2026 profit driver. It raises the long-run payoff; it shouldn’t move my entry from $345 up to $380–400.

What price actually pays you to own it

At ≈$380 COHR trades around 10.7× FY26E EV/Sales and ≈69× FY26E non-GAAP EPS — not a cheap multiple for a company whose operating cash flow hasn’t recovered. Anchoring off a ~$8.6 implied forward EPS, 40× lands near $345 and 35× near $300. That’s where the ladder comes from.

COHR build-zone ladderhover / tap for detail

1. ≥ $390Don't chase0–10%
2. $370–385Probe only≈$380 now10–20%
   ≈10.0–10.7×EV/Sales≈42–44×forward P/E≈ −13%off 52-wk high

   ≈ current price, and near the analyst average target of $379–384. A starter, not a full position.

3. $350–365Start building20–30%
4. $335–350Core build zone40–50%
5. $300–315High-payoff add25–35%
6. < $280Don't auto-bottom-fishcheck for break

$345 isn’t a number I picked from the air: it’s roughly where COHR is ≈9–10% below today, ≈40× forward EPS, and ≈20% off the $440 high — three conditions lining up at once. $300–315 is where the margin of safety gets genuinely comfortable; it may not come, but if it does, that’s a time to look hard, not to panic-sell.

Three scenarios

Scenario	Prob.	Trigger	Falsifier
Base	55%	Q4 meets guide; D&C keeps growing; inventory stabilizes; operating cash flow recovers off the lows.	Q4 OCF still deeply negative and inventory keeps climbing.
Bull	25%	1.6T / CPO / NPO pull-ins; FCF improves; thermal lands a named design-in earlier than H2 2027.	Only product news and spec sheets, no revenue in the numbers.
Bear	20%	AI-optics order slope slows; the inventory turns out “bad”; high-multiple AI hardware de-rates together.	D&C growth, margin expansion and FCF recovery all show up at once.

Base case is high-level chop where 350–365 becomes attractive and the build zone does the work; the bear case is a 25–35% de-rate even if the long-term logic is fine — which is exactly why I won’t pay the full multiple at the high.

The macro gauntlet and the real verdict date

Two-week macro + the next print Jul–Aug 2026

1. 07/01 · 10:00 ET US ISM Manufacturing PMI
   orders · inventory
2. 07/02 · 08:30 ET US June nonfarm payrolls
   rates · duration
3. 07/06 · 10:00 ET US ISM Services PMI
   price stickiness
4. 07/08 · 14:00 ET FOMC minutes (Jun 16–17)
   valuation gate
5. 07/10 TSMC June monthly sales
   AI / HPC demand
6. ≈ 08/12 COHR Q4 FY26 earnings
   the verdict

Risks and falsifiers

• Valuation. Trading-screen GAAP P/E is ≈180×; even FY26E non-GAAP P/E is ≈69×. If the next quarter doesn’t deliver revenue and cash flow together, the high multiple gets cut first.
• Cash flow / working capital. 9-mo FCF ≈−$537M, driven by inventory; if it keeps rising while growth slows, that’s the bad-inventory signal.
• Diamond slower than hoped. Thermal revenue is an H2-2027+ event; if 2027 arrives with no named design-in, mark the option down.
• NVIDIA is double-edged. The $2B and the capacity deal add visibility, but also bring capex, working-capital and customer-concentration pressure.
• Sector de-rate. If yields rise and AI capex expectations cool, COHR and LITE both get treated as long-duration hardware and sold together.

What would make me pay up: a print where operating cash flow turns clearly positive, inventory stops climbing, and management says diamond thermal has entered a major AI-accelerator, CPO, or hyperscaler reference design — with a revenue timeline. That’s the line that flips diamond from option to second estimate curve.

The honest version

The diamond direction you saw early is genuinely forward-looking. But buying COHR today, the first-principles question isn’t “is the material strong?” — it’s “how much has the market already paid for the AI-optics main line, and can diamond cooling become a financial contribution after 2027 rather than a press release?” I’d hold a low-cost core, build into 335–355, add hard at 300–315 if the fundamentals hold, and not chase the diamond story above 380. Position cap 6–8% of a normal book, 10% only if your cost is very low and you can stomach the volatility.

Opinions only. Data as of 2026-06-26. Size your own risk.