Understand How Coherent and Lumentum Bypass an AI Bottleneck With Optics

Дата публикации: 13-08-2026 16:51:42

Copper cables are choking AI clusters, and two optical companies just reported blowout quarters solving that problem in completely opposite ways. One bets on owning every layer of the stack, the other retreats to the single ingredient nobody else can replicate.

Основное содержимое страницы с новостью.

AI clusters have a wiring problem. Copper cables cannot move enough data between GPU racks, so the industry is switching to light. That shift is why Coherent (NYSE:COHR | COHR Price Prediction) and Lumentum (NASDAQ:LITE) just posted the kind of quarters optical suppliers rarely see. Coherent reported Q4 FY2026 on August 12, 2026, and Lumentum reported a day earlier. Both beat. Their paths diverged.

One Owns the Module. The Other Owns the Laser Inside It.

Coherent booked $2.045 billion in revenue, up 33.7% year over year, with the Datacenter & Communications segment producing $1.615 billion, or 79% of sales, growing 59% on a pro forma basis. Non-GAAP EPS came in at $1.74, a 7.60% beat. CEO Jim Anderson framed the setup plainly: “As AI datacenter architectures increasingly transition from copper to optical connectivity, we believe Coherent’s broad photonic technology portfolio and manufacturing scale uniquely position us to deliver accelerating growth.” Translation: own the whole module stack, from indium phosphide wafers up.

Lumentum went the other direction. Revenue hit $1.006 billion, up 109.3% year over year, non-GAAP gross margin reached 50.4%, a 1,260 basis point jump, and operating margin expanded 2,160 basis points to 36.6%. CEO Michael Hurlston said the company is “reaching our target model more than a quarter ahead of schedule.” That is the “pick and shovel” laser engine playing out in real time.

Full Stack vs. Rare Ingredient

Lens	Coherent	Lumentum
Core Bet	Own module stack and InP scale	Own high-margin laser engines
Q4 Gross Margin	40.2%	50.4%
Anchor Partner	NVIDIA at Sherman, TX	Hyperscaler CPO order, H1 CY2027
Next Catalyst	CPO/NPO H2 2026, OCS H1 2027	OCS backlog above $400M

Coherent is doubling internal InP output by year-end and doubling again by 2027, plus up to $50M in CHIPS Act funding. Vertical integration is the moat to capture massive top-line revenue at scale. Lumentum retreats up the value chain to own the rarest, most indispensable material-science bottleneck, and that shows in a P/S of 25.66 versus Coherent’s 9.74.

What I Am Watching Into FY27

Coherent guided Q1 FY27 revenue to $2.20B to $2.40B. Lumentum guided to $1.225B to $1.275B with operating margin of 39.5% to 40.5%. I want to see whether Coherent’s 205% one-year rally converts into CPO revenue on schedule, and whether Lumentum sustains margin now that the target model arrived early.

Where I Come Out on the Two

If I want breadth, US manufacturing, and the NVIDIA (NASDAQ:NVDA) relationship anchoring my thesis, I lean Coherent. If I want the steepest margin ramp and the most concentrated exposure to the laser bottleneck itself, Lumentum fits better, even at a 43 forward P/E. Both look expensive after a 679% one-year run on LITE, so the next two quarters of CPO and OCS revenue will be the key checkpoints to watch before the story is proven.

Contact [email protected] for any questions or corrections.

Схожие новости

#	Наименование новости	Тональность	Информативность	Дата публикации
1	Lumentum Puts The Insane Opportunity for Companies like Marvell and Ciena Into Focus With One Stunning Quote	0	7.28	13-08-2026
2	Optics Stocks Divide: Coherent and Cisco Drop After Earnings While Nokia and Ciena Soar	0	6.8	13-08-2026
3	GPUs And RAM Are In Short Supply, But The Real Bottleneck For AI Is Electricians	0	10	27-05-2026
4	The AI rush is hitting a bottleneck	-2	6	27-04-2026
5	Jim Cramer Says Skip Poet and Buy These 2 Proven Photonics Stocks for AI Data Centers in August	0	10.4	13-08-2026
6	Cisco Wins Over AI Customers With Merchant Silicon And Optics	0	10	18-05-2026
7	Salience Labs Wants To Scale Up AI With Silicon Photonics Optical Switch	0	10	22-07-2026
8	‘Tokenmaxxing’ is dead: How the hardware shortage is blowing up AI budgets	0	7	29-06-2026
9	OpenAI and Broadcom announce chip designed for LLM inference at scale	0	7	24-06-2026
10	Two chipmakers, one demo, and a signal for AI’s next phase	0	6.94	05-08-2026

Классификация: Пресс-релизы. Схожих патентов: 0. Схожих новостей: 10. Тональность: 0. Информативность: 8.52. Источник: 247wallst.com.