The stock dropped.

And it is now back to the level I initially bought in.

I first wrote about this company a few weeks ago. At the time, I framed it as a smaller, overlooked optics business with real revenue growth, real margins, positive EBITDA, positive cash flow, short lead times, and exposure to the AI-driven transport buildout. The stock had already moved, so I sized it as a starter. I wanted earnings to either confirm the thesis or expose a problem.

After this quarter, the thesis is stronger.

The company grew revenue almost 60% year over year. Gross margin held near 48%. EBITDA more than doubled. Book-to-bill was considerably above 1. Management said they could have shipped more if they had more production capacity.

That is the type of problem I want to see in this part of the market.

My read is that the stock sold off because expectations had moved quickly, Q2 has some production-move timing noise, and the company is now being judged like a higher-quality growth asset. But the underlying business looks better than it did when I first wrote about it.

And after a 25% gain, it is now back to where I first bought it.

So, I want more.

Inside, I am going through the Q1 results, what changed from my original thesis, why the AI transport angle is getting clearer, what I expect from Q2, and my updated 2026 and 2027 financial framework. None of this is financial advice or a recommendation to purchase shares.

Read more

The subtitle was clear: Optics, Copper, Defense. MACOM does it all.

After this quarter, that framing still works. But the main update is that my model was too conservative.

When I wrote the original report, MACOM was guiding for data-center revenue to grow 35% to 40% in FY2026. After this quarter, management raised that to over 60%. That is a serious reset.

Revenue came in at $289M, up 6.4% sequentially and over 22% year over year. Adjusted EPS was $1.09. Adjusted gross margin was 58.5%, up 90 bps sequentially. Book-to-bill was 1.5x, the largest bookings quarter in company history.

The segment numbers were strong across the board too. Industrial and Defense revenue was $120.7M. Data Center revenue was $98.2M. Telecom revenue was $70.1M. All three grew sequentially, but data center was the standout, up 14.5% from Q1.

The Q3 guide was the part that really changed the model. Management guided total revenue to $331M to $339M, adjusted gross margin of 59% to 60%, and adjusted EPS of $1.31 to $1.37. Data center should grow about 35% sequentially in Q3, which means it could be running at roughly $130M+ per quarter next quarter.

That is a major step-up from where I was modeling the company two months ago.

The Thesis Update

My original MACOM thesis was that the company sits in an attractive part of the AI infrastructure stack, making the analog and photonic components inside high-speed optical links: TIAs, drivers, photodetectors, lasers, and equalizers.

That thesis got stronger this quarter.

Current growth is being driven by 800G and 1.6T PAM4 products, pluggable optical modules, optical cable production, and rising demand for 200G/lane photodetectors. The important part is that this growth is showing up before the bigger CW laser opportunity becomes a model driver. The current business is being carried by products already in production. Real customer demand. Real backlog.

CW lasers remain the longer-term upside layer.

Paid readers get the full model reset below: the CW laser timing correction, the photodetector and DFB laser read-through, copper and coherent light updates, the defense and LEO setup, the IQE supply-chain angle, my updated FY2026/FY2027 EPS framework, and the new valuation range versus my original March report.

Read more

If you were not there for the last call a few months ago, AAOI was trading around $4.5 billion with roughly $500 million in full-year revenue. On that call, management laid out a path toward roughly $378 million in monthly transceiver revenue by mid-2027. The market loved this and has since re-rated the company to roughly $12 billion in market cap.

Now, management did more than confirm that target. They increased the framework by almost $100 million per month.

So now we are facing a potential situation where management sees visibility to $471 million per month of datacenter transceiver revenue by mid-2027. If you layer in more than $325 million of annual CATV revenue, the total framework approaches $500 million per month, or roughly $6 billion annualized.

That is not formal FY2027 guidance. It is a mid-2027 run-rate framework. But it is still massive.

The stock is valued at roughly $12 billion.

This sounds too good to be true, right?

The fact that the market cap sits at only $12 billion tells us that this is far from the whole story. A long list of things still has to go right:

• Capacity needs to expand dramatically and very quickly.

• Capex needs to ramp, and dilution is on the table.

• Margins need to improve materially.

• Orders need to get qualified.

• Qualified orders need to turn into on-time shipments.

• Customers need to diversify.

• Management needs to prove their execution.

• They have to prove they are a crucial player in CPO.

• 800G and 1.6T revenue need to show up in a real way.

This report unpacks all of it: what changed, what still has to happen, how I am updating the financial model, how I think the market will treat the stock, what my biggest concerns are, and how I am thinking about the risk/reward from here.

No financial advice is being provided. Please do your own research and consult with a financial professional.

My Quick Thoughts

The headline earnings were pretty week. The call was more bullish.

Revenue came in near the low end of guidance. Gross margin was soft. Q2 gross margin guidance stayed soft. If you only looked at the press release, you would definitely expect the stock to get hit.

But the call changed the read. Management raised the full-year 2026 framework. They quantified 800G revenue for the first time. They confirmed 800G is now shipping in volume. They increased the mid-2027 monthly transceiver revenue framework from $378 million to $471 million per month. And the CEO explicitly clarified that the $471 million number is revenue, not capacity.

Q1 proved the ramp has started. The rest of 2026 has to prove scale.

The Actual Print

They reported Q1 revenue of $151.1 million, up 51% year over year and 13% sequentially. That was inside the prior guide of $150 million to $165 million but near the low end. Non-GAAP gross margin was 29.2%, also inside the 29% to 31% guide but again near the low end. Non-GAAP EPS was a loss of $0.07, within the prior range.

For Q2, management guided revenue of $180 million to $198 million, non-GAAP gross margin of 29% to 30%, and non-GAAP EPS of -$0.03 to +$0.03.

Revenue is ramping. Margin leverage has yet to show up. That is why the print and the call have to be read separately.

Read more

We saw firmware progress. 800G orders. A $200M+ 1.6T order. A larger Texas footprint. A ~$20M state grant. More laser capacity. More hyperscaler interest.

Now we need the next layer.

Orders need to become qualified and then become shipments.

Shipments need to become recognized revenue.

Revenue needs to improve margin.

Capacity needs to become actual production.

One thing worth keeping in mind going into this call is how management handled the last one. They acknowledged the 800G delay, then quickly redirected us toward Q2 and the full-year ramp. They may do something similar here if Q1 still has limited 800G revenue. The market may not like that, but if Q2 language is very specific rather than directional, that could help.

AAOI already guided Q1 revenue of $150M to $165M, adjusted gross margin of 29% to 31%, and adjusted EPS from a $0.09 loss to breakeven.

Last quarter they did $134.3M of revenue and 31.4% adjusted gross margin, so they already gave us a step-up in the rev guide (and a slight dip in gross margins). But the actual stock reaction will depend more on the Q2 guide and commentary than the Q1 print, in my opinion.

Management already told us 800G was supposed to dominate datacenter revenue starting in Q2. Q1 was always intended as a bridge quarter. If Q1 is fine but Q2 commentary is weak, that is a problem. If Q1 is fine and Q2 confirms the 800G ramp, the story gets stronger.

This call has all the fixings to be explosive. In both directions. We are up around 3x where we were in the last call, and expectations are running high.

In this post I will dive into exactly what I am watching, what I think would read bullish/bearish, and what my own strategy is. Also, for paid subscribers I will also be doing a live chat during the call, highlighting what I am hearing. None of this is financial advice.

Read more

Most of our companies are following the same trend so far.

Sell 10-15% on headline, then base or recover over the coming days.

I have no prediction on what happens in the short term with this stock.

I know what levels I would personally be a buyer, and I have confidence in where this will be trading at the end of 2026 and 2027.

So, I think Coherent’s Q3 was moderately bullish, but the call was stronger than the print.

The headline numbers were good. The Q4 guide was better. But the reason I came away more constructive was the call itself. Management addressed the exact questions I cared about: 6-inch InP, EML yields, 1.6T, CPO timing, OCS revenue, LTAs, and the 2027 growth setup.

The stock can still trade choppy because expectations were high and gross margin remains around the 40% zone. But the thesis itself got stronger after this report. Coherent confirmed demand, guided to another strong sequential step, gave more detail on 6-inch InP, pushed back against the EML bear case, gave better timing around CPO, showed OCS is revenue today, and added more 2027 growth layers through Multi-rail and thermal solutions.

This was the kind of call I wanted to hear.

The debate now shifts from demand to conversion. Can Coherent turn this demand into 40%+ gross margins first, and then 42%+ over time? That is the next test.

The quick read

Q3 revenue came in at $1.806B.

Non-GAAP EPS came in at $1.41.

Non-GAAP gross margin was 39.6%.

Datacenter & Communications reached $1.362B, up from $1.208B last quarter, and now represents 75% of total revenue.

That last part is one of the biggest pieces of the report. Coherent is becoming more of a pure AI datacenter and communications company. The industrial business still has value, but the growth is clearly coming from the part of the company we care about most. So transceivers, 1.6T, DCI, OCS, CPO, optical components, InP, EMLs, CW lasers, photodiodes, and silicon photonics.

Q4 revenue is guided to $1.91B to $2.05B. The midpoint is $1.98B, which is another large sequential step higher from Q3.

Q4 non-GAAP EPS is guided to $1.52 to $1.72, with a midpoint of $1.62.

Q4 gross margin is guided to 39% to 41%.

So IMO the print was good, the guide was better, and the call was the strongest part.

So let’s unpack everything.

The rest of this post is for paid subscribers. We discuss the entirety of the call from the 6-inch updates, their capacity expansion, the demand they are seeing, what OCS will look like, when CPO is ramping, what 2027 is brining, my financial forecasts and what my strategy will be. None of this is financial advice.

Read more

This is one of my largest positions.

After last earnings, the stock dropped 20% immediately.

I bought the dip because I felt like the market was over looking the real story.

We are now 100% higher.

Will today repeat?

We know demand is going to be there.

So can Coherent turn bookings into massive shipments?

Can 6-inch indium phosphide show up in gross margin?

Can 1.6T move from early ramp language into real revenue?

Coherent sits across a large piece of the AI optical stack

800G and 1.6T transceivers, EMLs, CW lasers, silicon photonics, photodiodes, ZR/ZR+ modules, DCI, OCS, CPO, external laser sources, industrial lasers, engineered materials, and early AI thermal-management materials.

That breadth is what made me comfortable allocating a significant amount of money here. Less torque than Lumentum, but a wider rev base to support.

But that goes both ways. The burden is in focus now.

A company this broad has to prove it can push capital into the highest-return parts of the portfolio and deliver the ideal product mix.

The rest of this post is for paid subscribers. In it you will see where Coherent stands today, what I am watching for, what my strategy is, and what levels I am watching. None of this is financial advice or a recommendation to purchase a security. This is solely educational.

Read more

Overall, stock reaction was negative to start and then right now overnight it’s roughly flat.

I think the market didn’t love the revenue print. But the margins were great and the story is intact.

Revenue came in at $808.4M, above the midpoint of the prior guide and up 90% year over year.

The stronger part of the report was profitability. Non-GAAP gross margin reached 47.9%, non-GAAP operating margin reached 32.2%, and non-GAAP EPS came in at $2.37, above the prior guide.

Then guidance moved sharply higher. Lumentum guided Q4 FY26 revenue to $960M–$1.01B, with non-GAAP operating margin of 35%–36% and EPS of $2.85–$3.05. At the midpoint, Q4 revenue would be $985M, putting Lumentum close to a $1B quarterly revenue run-rate.

Lumentum is already approaching $1B of quarterly revenue while OCS and scale-out CPO are still modest contributors, and scale-up CPO remains in its infancy.

None of this is financial advice or a recommendation

The Main Read

It’s all about supply.

Management said the EML supply-demand imbalance is now greater than 30%, compared with the prior 25%–30% framework. Customers are still asking for more than Lumentum can ship, even after the company has been aggressively expanding output. The real question is whether Lumentum can add, qualify, and ship enough capacity across several constrained product lines at the same time.

So far, execution looks strong. But the execution burden is rising.

Revenue Was Solid, Margins Were Excellent

Q3 revenue grew from $665.5M last quarter to $808.4M. Components revenue reached $533.3M, up 20% sequentially and 77% year over year. Systems revenue reached $275.1M, up 24% sequentially and 121% year over year.

The mix is important as Lumentum is getting strength from both sides of the portfolio: laser chips, EMLs, pump lasers, narrow linewidth lasers, transceivers, OCS, and early UHP/CPO contribution.

Operating leverage was a really powerful financial signal. Revenue grew by roughly $143M sequentially, while non-GAAP operating expenses increased by only about $11M. That is how operating margin jumped from 25.2% to 32.2% in one quarter.

The print was bullish even though Q3 revenue landed near the guide midpoint. Lumentum showed that constrained supply, richer mix, pricing discipline, and better utilization can create a very different earnings profile.

Q4 Guidance

At the midpoint, Lumentum is guiding for roughly 22% sequential revenue growth from Q3 to Q4. The company also expects operating margin to move from 32.2% to 35%–36%.

That puts Lumentum much closer to the $1.25B quarterly revenue target management discussed at OFC. The path to that first target now looks more credible.

Management said more than half of Q4 sequential growth should come from Components, with the rest driven by Systems, primarily high-speed transceivers and additional OCS contribution. The guide is broad. EMLs, scale-across components, transceivers, and OCS are driving the near-term guide, while CPO remains early but on schedule for more meaningful December-quarter revenue.

EMLs

LITE shipped twice as many laser chips as it did in the same quarter last year and set new EML shipment records, with 200G EML revenue more than doubling sequentially.

The 200G transition helps both units and mix. Higher-speed EMLs carry better ASPs, and Lumentum is benefiting from both volume growth and richer mix. Even with this production growth, management said the supply-demand imbalance is now greater than 30%. Lumentum expects EML unit growth of more than 50% by the December quarter of 2026 versus the December quarter of 2025, yet demand continues to move faster than supply.

Scale-Across Became a Bigger Part of the Thesis

This was the biggest new theme on the call for me.

Lumentum spent much more time discussing scale-across than I expected. Scale-across is the optical connectivity layer between data centers, buildings, campuses, and distributed compute domains. As hyperscalers run into power and space limits inside individual buildings, they need to link compute across multiple physical locations. Lumentum is monetizing this through pump lasers, narrow linewidth laser assemblies, WSS (Wavelength Selective Switches) and emerging multi-rail technologies.

Management said pump lasers allow scale-across architectures to amplify light across 4, 8, or 16 fiber pairs. Narrow linewidth laser assemblies provide the precision required for 1.6T speeds and higher-order modulation. WSS keeps traffic in the optical domain while supporting high port counts for massive fiber routing between buildings.

The growth numbers were really strong. Narrow linewidth laser assemblies grew for the ninth consecutive quarter and rose over 120% year over year. Pump laser shipments grew 80% year over year. Management said these components remain effectively sold out for the foreseeable future.

Then management said scale-across components are probably more constrained than EMLs, especially pump lasers and narrow linewidth lasers.

Before this call, most of the Lumentum discussion centered around four growth drivers: cloud transceivers, OCS, optical scale-out, and optical scale-up. Scale-across now deserves its own place in the story. It is already contributing to gross margin improvement, and management is working on long-term agreements to help offset the CapEx required to add capacity.

Multi-Rail Means More Optical Content

Multi-rail means more parallel optical paths across large AI networks. More rails require more amplification, more pump lasers, more narrow linewidth lasers, more WSS, and more optical traffic management. Management said multi-rail increases content because more pumps are needed per deployment. Wupen Yuen went further, saying the multi-rail opportunity could be “huge,” while also noting the company is still working toward fuller quantification.

This is a broader read-through for the AI infrastructure stack. More scale-across and multi-rail means more physical fiber routes, higher fiber counts, more optical amplification, more switching, and more connectivity. I will cover the fiber and Corning read-through separately. For Lumentum, scale-across is becoming a real growth and margin lever.

Transceivers Are the Revenue Bridge

Cloud transceivers were the largest driver of Systems growth.

Management said cloud transceiver shipments grew over 40% sequentially, helped by the expanded manufacturing footprint in Thailand. Lumentum is also poised to ramp 1.6T transceiver shipments in Q4 FY26, with a portion of that volume using internal CW lasers. Transceivers are the near-term bridge to the larger OCS and CPO story.

The internal CW laser integration matters. Management said roughly 20% of modules in the Q4 guide should use internal CW lasers, with that share expected to rise over time. That helps both supply control and margin. This also has read through for AOI, which we will discuss in our pre-earnings watchlist for them on Thursday.

Lumentum’s transceiver margins still trail peers, according to management, but 1.6T structurally carries better margins than 800G. The company is improving yields, reducing scrap, and securing more laser supply internally. Lumentum appears to be ahead on 1.6T but remains supply constrained. Management said the company could have shipped quite a bit more in Q3 and could ship quite a bit more in Q4 if supply were available.

OCS Is On Track, But It Is the Tightrope

OCS remains one of the most important parts of the Lumentum story and they really highlighted that today.

Management reiterated that the multi-year, multi-billion-dollar OCS agreement supports sustained growth. The ramp is largely on track, although the pace and slope are gated by supply chain constraints. Management believes the previously discussed $400M of OCS shipments in the back half of calendar 2026 is under control, but 2027 requires another major step up. Michael Hurlston called OCS the biggest single “tightrope” the company is walking.

Bullish because demand is clearly strong. Risky because OCS is a complex ramp, and supply chain pacing can shape quarterly revenue.

The most encouraging comment was around new OCS opportunities. Management said additional opportunities are substantial and in the same order of magnitude as the 2027 backlog discussion. If Lumentum can execute, OCS could become much larger than the first major agreement.

We will discuss OCS more with Coherent tomorrow, and I want to do a new OCS general report this weekend involving Huber + Shumer as well.

CPO and UHP Lasers Remain On Schedule

The CPO story remains intact.

Lumentum said the ultra-high-power laser chip ramp for CPO is proceeding according to plan. The company achieved sequential growth in Q3 and remains on schedule to deliver meaningful revenue in the December quarter and fulfill the multi-hundred-million-dollar purchase order scheduled for the first half of calendar 2027.

Management also said development work continues with multiple CPO customers through collaborations using Lumentum laser chips inside a pluggable turnkey ELS module solution. ELS is important because it moves Lumentum from selling only laser chips toward selling a more complete external light source module. On the call, management said ELS opportunities are getting closer and that significant wins feel “around the corner.” That could become a major 2027 driver.

The sequence is becoming clearer: UHP lasers first, ELS next, scale-up CPO later.

Scale-Up CPO Is Still the Largest Future Driver

The most important long-term line from the call was that current numbers and guidance include only modest initial contributions from scale-out CPO and OCS, while the largest single growth driver, scale-up CPO, remains in its infancy. Management said this gives them confidence that Lumentum remains on track to reach the $2B quarterly revenue goal discussed at OFC.

Lumentum is guiding to almost $1B of quarterly revenue before the largest future driver has meaningfully arrived. If scale-up CPO begins contributing in 2027, and if OCS plus scale-out CPO continue ramping, the $2B target becomes more credible.

Capacity Is the Limiting Factor

The main challenge now is capacity.

Lumentum is ramping EMLs, pump lasers, narrow linewidth lasers, transceivers, OCS, UHP lasers, internal CW lasers, ELS, and the Greensboro InP conversion simultaneously. The Greensboro acquisition is important, but it is a 2028 story. Management said Greensboro sits outside the current numbers and could support more than $5B of incremental revenue if executed properly. Significant contribution is still roughly six quarters away.

Nearer term, Lumentum is relying on existing fabs, contract manufacturing, Thailand expansion, higher utilization, internal CW laser integration, and aggressive CapEx. The balance sheet gives them room. Cash and short-term investments increased to $3.17B, mostly due to NVIDIA’s direct investment. Inventory rose by $62M sequentially to support expected cloud and AI revenue growth, and Q3 CapEx was $125M, primarily focused on manufacturing capacity for cloud and AI customers.

Updated Forecast Framework

I am using calendar year-end quarterly run-rate estimates because Lumentum’s fiscal year ends in June. These are my working ranges, separate from management guidance, based on the Q4 guide and management’s $2B quarterly revenue target.

Read more

Lumentum is now tied to almost every major theme I care about in photonics: cloud transceivers, 1.6T, EML lasers, indium phosphide capacity, optical circuit switches (OCS), co-packaged optics (CPO), external light sources (ELS), and optical scale-up.

This call is about whether Lumentum can earn it’s multiple, expand it’s capacity enough to meet demand, and capture the massive TAM that is infront of it.

I will be listening to this call live and providing notes real time for paid subscribers.

At the end of this article I also lay out my strategy going in to the call and what levels I am watching for a potential add.

And at the end of the the day today I will post a full write-up on the call.

Where the Story Stands

Last quarter was already a major inflection point.

Lumentum reported $665.5M of revenue, up more than 65% year over year, and guided the next quarter to $780M–$830M. The midpoint of that guide is $805M, which would be another company revenue record. Non-GAAP operating margin was 25.2% in Q2, and management guided Q3 operating margin to 30%–31%.

We already know demand is going to be strong, so the question now is conversion: shipments, backlog, pricing power, margins, and guidance.

The challenge is that we already know the March quarter should be strong as well. The stock reaction will likely depend more on Q4 guidance, backlog conversion, and whether management reaffirms the $2B quarterly revenue model.

The Supply Constraint Is the Thesis

The most important part of the Lumentum story is that they are selling into markets where demand is moving faster than supply. Fabrinet reaffirmed yesterday that there is still a major supply constraint, especially in lasers.

Management said indium phosphide capacity remains fully allocated, and even after capacity additions, the company is still under-shipping customer demand by roughly 25%–30%. EML capacity is effectively spoken for through calendar 2027 under long-term agreements.

When a company has scarce capacity, long-term agreements, pricing leverage, and customers asking for more than contracted volumes, the model can change very quickly. That is exactly what happened last quarter. Gross margin moved to 42.5%, up 1,020 basis points year over year. Operating margin expanded by 1,730 basis points year over year.

That is why Lumentum has started to trade more like a strategic capacity bottleneck.

The Four Growth Engines

Management has been clear that the future of the company is built around four drivers, each with a different timing profile.

1. Cloud Transceivers

This is the current revenue bridge.

Lumentum acquired Cloud Light to get more direct exposure to cloud transceivers, and the early execution was messy. Management has been candid about that. But the tone has changed. Last quarter, they said the transceiver business had reached a “turning point.” They are now in the lead pack as customers move toward 1.6T, and 1.6T demand came in stronger than they expected 90 days earlier.

The important detail is that 1.6T carries a better margin profile than 800G. Transceivers can grow while becoming less of a drag on overall company margins.

The key question is whether 1.6T shipments remain on track for this summer, and whether customer demand has strengthened again.

2. Optical Circuit Switches

OCS is a major near-term upside lever.

Lumentum originally expected its first $10M OCS quarter in fiscal Q3. They hit that level one quarter early. OCS backlog is already well above $400M, with most of that expected to ship in the second half of calendar 2026. At the Nokia/OFC event, management said they had signed a new multi-year, multi-billion-dollar agreement with a large OCS customer and expect OCS to ramp beyond $1B in calendar 2027.

OCS felt far away a couple quarters ago. Lumentum is now talking about backlog, customer expansion, and a billion-dollar revenue run rate. The key question is whether the $400M-plus backlog still converts in 2H CY26, whether OCS ramping beyond $1B in calendar 2027 stays intact, and whether additional customers are engaging. Any increase in backlog or customer count would be very bullish.

3. Scale-Out CPO

Scale-out CPO is the first CPO revenue layer, mainly tied to ultra-high-power lasers and external light source modules.

Lumentum said it secured an incremental multi-hundred-million-dollar purchase order for UHP lasers supporting optical scale-out, with shipments expected in the first half of calendar 2027. At OFC, management said they expect the first $100M scale-out CPO / UHP laser revenue quarter around late calendar 2026, followed by delivery on the multi-hundred-million-dollar commitment in early 2027.

CPO has been discussed for years. Lumentum is now talking about specific orders, shipment windows, and capacity planning. That is a meaningful shift. We are hearing this across the stack too. CPO is already beginning in early stages.

The other piece I am watching is ELS, or external light source. Rather than selling only a laser chip, Lumentum can sell a more complete external light source module. Management said this creates roughly 2x to 2.5x the revenue content of standalone lasers while preserving strong margins. ELS could let Lumentum move up the value chain while helping customers that lack deep internal optical engineering teams.

4. Scale-Up CPO

Scale-up is the real 2027-plus option, and this is where the optical story gets much larger.

Today, copper still dominates short scale-up links inside racks and clusters. But copper becomes harder as bandwidth rises and cable distances increase. Lumentum expects first scale-up CPO shipments in the second half of calendar 2027. Management has said the first scale-up opportunity could be 3x–4x larger than initial scale-out CPO, and later in-rack optical adoption could create an even larger lane opportunity.

If AI architectures move from rack-scale to multi-rack scale-up domains, the optical content per system can move much higher. For Lumentum, that means demand for UHP lasers, external light sources, InP capacity, OCS, and high-speed EMLs. They are adding capacity now for demand that could show up in 2027, 2028, and beyond.

The Capacity Plan

The company is working on several layers at once.

First, Lumentum is squeezing more output from its existing indium phosphide footprint. EML output has increased 8x since FY23, and management expects indium phosphide output to increase another 50% from Q4 CY25 to Q4 CY26. On the Q2 call, management also said the earlier 40% capacity expansion plan was front-loaded, with more than half of that already delivered in the December quarter, and additional capacity expected through Sagamihara, Caswell in the UK, and Takao in Japan.

Second, Lumentum is shifting more capacity toward ultra-high-power lasers for CPO. San Jose is the first UHP ramp. Caswell is becoming the second UHP capacity block. Greensboro is the third, larger step. Management originally viewed Caswell as EML overflow, but the UHP opportunity became large enough that they are now leaning more of that site toward scale-out and scale-up CPO demand.

Third, Lumentum closed on the former Qorvo fab in Greensboro, North Carolina. Management described this as its fifth indium phosphide fab, primarily focused on UHP lasers, with 4-inch and 6-inch compatibility. Shipments are expected in 2028. This fab is important because it supports the next leg of scale-out and scale-up CPO demand, but it sits outside the current $2B quarterly revenue model.

Fourth, backend manufacturing is also part of the capacity story. Lumentum has been expanding through Thailand and leaning more heavily on contract manufacturers. Management said factory capacity is also tight, which means the bottleneck is wider than front-end InP wafers alone. Even with strong wafer output, assembly and manufacturing execution still need to keep pace.

Fifth, the NVIDIA investment gives Lumentum more flexibility to fund strategic CapEx, working capital, and potential vertical integration. Management also said additional capacity through existing fabs or new capacity by acquisition remains an active discussion.

The path to $1.25B quarterly revenue and then $2B quarterly revenue is based on existing and near-term capacity. Greensboro is the next layer of optionality for 2028 and beyond. That means the current model already embeds a major capacity ramp, while the newer U.S. fab supports the next phase of UHP laser and scale-up CPO demand.

The $2B Quarterly Revenue Model

At OFC, Lumentum laid out a path to $1.25B quarterly revenue in 9–12 months, $2B quarterly revenue 9–12 months after that, and 40% non-GAAP operating margin at the $2B level.

The current quarter midpoint is $805M. If management is right, Lumentum could move from an $800M quarterly revenue base to a $2B quarterly run rate within roughly 18–24 months. That is a very aggressive target. The key question is whether management still has the same confidence in that path.

The rest of this post is for paid subs. If you subscribe you will see my exact watchlist, what would make me more bullish, what would make me bearish, and what my strategy + levels are.

Read more

They reported $1.214B of revenue, up 39% year over year and 7% sequentially.

Non-GAAP EPS was $3.72. Both came in above guidance.

Q4 guidance came in at $1.25B–$1.29B of revenue and $3.72–$3.87 of non-GAAP EPS.

That implies continued sequential revenue growth, EPS roughly flat to up sequentially, and about 40% year-over-year revenue growth at the midpoint.

So the business read is fairly bullish.

The stock setup is more complicated in my opinion.

This quarter confirmed that AI infrastructure demand is real, but it also showed why the market may still debate the pace of growth, revenue visibility, margin pressure, and the right valuation to pay today.

Datacom demand is stronger than reported revenue, DCI is accelerating, HPC is expanding beyond the original program, and capacity is being built for a much larger business.

Below is my full breakdown of what changed, what still needs to prove out, and the levels I am personally watching.

None of this is financial advice or a recommendation.

Read more

We had AIXA and LPK in the morning, and AXTI and COHU in the evening.

We had plenty of volatility across the board.

We also had major signal, and thesis reinforcing/breaking comments.

After hearing a couple of these calls, I want to build/increase my position size. And I also want to avoid one of them for the time being.

This article will be very comprehensive. I am breaking down all 4 calls.

I got to listen to AXT call live, but I had to read + analyze the other 3. It’s taken quite some time, but I am happy with this report.

If you read this, not only will you understand where these companies stand today and whether their thesis are in-tact or not, but you will also get signal for the overall optics trade.

Let’s dive in!

Read more

If you haven’t read my introduction piece on Viavi yet, please read it here:

Was taking live notes during the call today for subscribers, and the call was jam packed with alpha!

So now in this article I want to unpack things further, see just what went down, what is coming over the next few quarters, what signals it gives for the whole stack, my new financial model, and how I want to be positioned.

Subscribe to unlock!

Read more

If you didn’t hear, the Wall Street Journal reported that OpenAI missed recent user-growth and revenue targets, and that CFO Sarah Friar had raised internal concerns about whether revenue could scale fast enough to support future compute commitments. Reuters summarized the report and included a joint statement from Sam Altman and Friar saying they are “totally aligned” on buying as much compute as possible. Bloomberg separately reported OpenAI’s company-level pushback, with the company calling the report “prime clickbait” and saying its consumer and enterprise businesses are “firing on all cylinders.” OpenAI’s own recent enterprise update from CRO Denise Dresser adds the counterpoint thatenterprise is already more than 40% of revenue, APIs are processing more than 15 billion tokens per minute, and Codex has 3 million weekly active users.

Both things can be true I think. AI demand can be massive while investors still question whether revenue is broadening fast enough to justify the infrastructure being committed.

That tension is exactly what Microsoft, Alphabet, Amazon, and Meta have to answer this week. Between them, roughly $600 billion of AI-related capex and spending is on the table for 2026. We already know the number. What we want to know is whether real, broad-based demand is sitting underneath it.

So my central question is is AI infrastructure demand still broadening, or is too much of the market leaning on a small number of aggressive buyers?

There is also an important counterpoint to the OpenAI-specific bear case that I want to address. Even if OpenAI had to slow the pace of future compute commitments, that only damages the broader AI infrastructure thesis if the capacity has limited alternative buyers. Anthropic, Google, Amazon, Meta, sovereign AI programs, and enterprise AI workloads are all trying to secure more compute. Anthropic has already committed more than $100 billion to AWS over ten years, securing up to 5GW of capacity. It has also signed a separate agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity expected to come online starting in 2027.

So I think the real question here is if OpenAI slows, are there enough other buyers ready to absorb the capacity?

Subscribe now

What I Am Watching

Capex numbers and the language around them. A capex raise can signal conviction, capacity pull-forward, or higher buildout cost. A lower number can signal timing, efficiency, or a softer demand view. The number is the first read and then explanation gives more insihgt. High spend tied to visible demand is the strongest signal. High spend with softer language around utilization, pacing, or return timelines is more complicated. Read the number and the context together. Listen for “demand ahead of supply” and “capacity being absorbed as it comes online.” Be cautious if you start hearing “optimizing deployment” or “building ahead of demand.”

Cloud growth with margins intact. Revenue scaling fast enough to absorb GPUs, networking, power, cooling, and depreciation is the economics argument. Growth alone gives an incomplete answer. Growth with acceptable margins gives the stronger signal.

Inference spreading into real products. Inference is what makes the buildout recurring, because inference scales with usage. Every time a user gets a search result, a code suggestion, an ad recommendation, or an enterprise workflow runs an AI agent, that is inference consuming infrastructure. If management teams describe AI usage spreading across search, ads, coding tools, enterprise agents, and cloud APIs at scale, the demand base is broadening in the way the thesis requires. If the conversation stays concentrated on frontier model development, the demand base is narrower than the spending implies and the OpenAI concern may become more relevant.

Backlog and customer commitments. Rising commercial RPO, cloud backlog, long-term agreements, and capacity reservations are the financial proof that demand is real and forward-looking instead of short-cycle purchasing that could slow quickly.

The Companies

Microsoft: The question is whether Azure and other cloud services AI demand extends meaningfully beyond OpenAI into the broader enterprise base across Copilot, GitHub, developers, and internal products. Last quarter commercial RPO reached $625 billion, up 110%, with roughly 45% tied to OpenAI, and Azure and other cloud services grew 39%. If that breadth is real and expanding, the concentration risk fades. If growth is heavily tied to one or two large relationships, the thesis gets narrower exactly when we may already be nervous about it.

Alphabet: Google Search is one of the few places where AI is already embedded in a product used by billions of people every day. If AI is strengthening Search economics and driving Google Cloud growth simultaneously, that is the closest thing this week has to a clean answer on whether AI creates revenue as well as cost. Guided 2026 capex to $175B to $185B. Google Cloud grew 48% last quarter. The bar is high for sure.

Amazon: AWS margins have been strong and Trainium plus Graviton hit a combined annual run rate above $10B last quarter with Trainium2 fully subscribed. Watch whether AI cloud demand is accelerating while profitability stays intact. If AWS can show both simultaneously, it helps answer the economics question.

Meta: Meta’s ad business is enormous and already growing. The question is whether AI is visibly improving ad performance metrics, recommendation quality, and engagement in a way that shows up in revenue. Meta guided 2026 capex to $115B to $135B while also guiding 2026 operating income above 2025. Meta’s AI payoff is partially measurable right now through ad revenue and engagement data instead of future cloud commitments. If ad performance is strong and the infrastructure cost is being absorbed, Meta becomes one of the clearest real-world proof that AI investment translates into business results.

The Read-Through For My Coverage

This is where I translate the calls back into the companies I follow.

The biggest signal is compute appetite. I want to hear whether Microsoft, Alphabet, Amazon, and Meta are still trying to secure more compute, bring more data-center capacity online, and expand AI infrastructure faster than current supply allows.

If hyperscalers still need more compute, demand flows through chips, servers, networking equipment, fiber, optics, power, cooling, and data-center construction.

I do expect these companies to speak one layer above the supplier level. They may say “technical infrastructure,” “servers and networking equipment,” “AI capacity,” “data-center capacity,” “cloud demand,” “supply constraints,” “power,” “cooling,” “depreciation,” and “customer commitments.” That is the language I will be translating into my coverage universe.

If management teams talk about larger AI clusters, regional capacity, AI backend networks, WAN connectivity, or data-center networking, that would be supportive for the optical stack. That is the language I am listening for across fiber, coherent optics, transceivers, data-center interconnect, switches, routers, and test equipment.

Power and cooling are the other side of the same buildout. If density, grid interconnect timelines, liquid cooling, energy costs, or power availability show up as limiting factors, that supports the broader power infrastructure and cooling trade.

The risk would be a more cautious tone around compute demand, data-center buildout pace, utilization, customer commitments, or return timelines.

I will be sharing the full post-earnings breakdown across the stack for paid subscribers. I will go over what the signals are for the general AI infrastructure stack and also for all the companies we track around optics/interconnect/compute etc.

Subscribe now

I did a live note taking / signal check during the call for subscribers

But my main question going into the quarter for me was was Meta a one-off agreement, or the first proof point of a broader customer-backed AI optical capacity model?

Corning answered it.

The company signed two additional large, long-term hyperscaler agreements, each similar in size and duration to the Meta deal.

Optical Communications grew 36% year over year.

Optical net income grew 93%.

Management also said they will upgrade and extend Springboard through 2030 at the May 6 investor event and introduce a new Photonics Market-Access Platform focused on Gen AI OEM customers.

The Q2 guide had solar noise. The stock is already up a lot. The market wanted a cleaner raise. But the core signal improved. Corning is moving from an optical products supplier into a customer-backed capacity partner for AI infrastructure.

So below, I want to break down what actually changed this quarter, why the Optical segment may deserve a higher-quality multiple, what the new hyperscaler agreements really mean, and why May 6 could be the next major catalyst for both $GLW and the broader optics sector. You’ll want to read this!

Read more

The company still looked like a sleepy manufacturer, but the business underneath was starting to shift toward the physical AI infrastructure buildout. Then ahead of earnings I laid out the three angles I cared about most: optics, power, and CPU/inference infrastructure.

Q2 gave us our first real answer. And it was a strong one.

The headline numbers beat across the board.

But the more important story was in the call itself…what drove the ZT upside, what kind of accelerated compute shipped, how the revenue mix moved, and what management said about 2027.

I listened to the call live and was dropping comments in a chat for the subs

After this call, the Sanmina thesis is progressing.

That is what I am breaking down below.

Behind the paywall, I’m breaking down the Q2 beat, the ZT surprise, the AMD connection, the optics confirmation, the power angle, the FY27 setup, the main risk in the Q3 guide, and how I’m thinking about my position after the move.

Read more

That was the core idea behind Nokia’s Third Act Is Starting to Come Into View

And the real question behind Nokia Earnings Preview — What to Watch.

Was the architecture thesis actually starting to show up in the financials, or was this still mostly a story investors had to piece together from product launches, OFC commentary, and management language?

This quarter gave us the clearest answer yet. Nokia showed that AI and cloud demand is now large enough to move the numbers, move the guidance, and start changing how the company should be framed.

The parts of Nokia tied most directly to AI networking demand became too strong to keep treating as side angles.

So let’s break everything down that I think is important from this call!

What changed this quarter

Three things got pulled forward this quarter and together they explain why the story feels different.

The feared air pocket never showed up. Hoorah! Going into the print, the setup carried real risk. Telecom drag was real. The AT&T headwind was real. The concern was that Nokia might still be stuck in an awkward transition where the old business was slowing before the new one was large enough to carry weight. The quarter came through much cleaner than that.

AI and cloud demand showed up directly in the numbers. In current sales, current orders, and current guidance. Many companies tell a good future story. The ones worth owning show the demand in the current period.

The story broadened beyond optical though. Optical is still doing the heavy lifting and remains the clearest proof point, but IP is starting to move alongside it. Nokia becomes a much more valuable story when it is selling into multiple layers of the network stack simultaneously.

This was the quarter where the AI networking thesis move from being theoretical and started being financial.

The numbers behind all of this — the exact figures, the guidance raises, the order flow, the margin expansion, what I think it means for how Nokia gets valued from here, and what my strategy is for my investment — are below for paid subscribers.

Read more

Today we got the full earnings with the call.

There were lots of good nuggets in there.

Let’s unpack it.

The Financial Inflection & Balance Sheet

As anticipated, the actual Q3 income statement lagged the massive order flow. Q3 revenue came in at $10.3 million, down 44% YoY, primarily due to a lower volume of high-margin WaferPak shipments. WaferPaks are the contact hardware Aehr uses to test entire wafers. This less favorable product mix dragged non-GAAP gross margins down to 36.5%, compared to 42.7% a year ago.

The real story was the operational pivot and exploding backlog. CEO Gayn Erickson highlighted this shift: “We’re very pleased with the strong momentum in our business across multiple market segments, highlighted by more than $37 million in quarterly bookings and a book-to-bill ratio exceeding 3.5x”. Adding in the $12.2 million booked in the first five weeks of Q4, Aehr’s effective backlog has surged to a record $50.9 million.

One crucial update is that Aehr has officially tapped out its At-The-Market, or ATM, equity offering. They raised $10.5 million in Q3 and another $19.5 million shortly after the quarter ended. CFO Chris Siu noted that they have “now fully utilized the $40 million available under the ATM and have sold over 1.13 million shares at an average price of $35.38”. This leaves them with a bolstered balance sheet, ending Q3 with $37.1 million in cash and cash equivalents, with the current ATM now fully utilized.

Because of the massive $50.9 million backlog, management provided highly optimistic guidance for the remainder of Fiscal 2026.

Management expects full-year fiscal 2026 revenue “to be on the high side of the $45 million-$50 million range” and second-half bookings to hit the “high side of the $60 million-$80 million range”.

Gross margins are expected to improve in Q4 as scaled manufacturing better absorbs fixed costs. Consequently, Aehr expects “to return to profitability on a non-GAAP basis in the fourth quarter”. For the full year, they project a non-GAAP net loss between -$0.13 and -$0.09 per share.

One other timing detail worth knowing is that Aehr is also shifting its fiscal year-end from late May to late June. So after Fiscal 2026 ends on May 29, 2026, Fiscal 2027 will begin on June 27, 2026.

The Sonoma ramp + milestone

Aehr’s Sonoma system is its package-level burn-in platform. It tests chips after they have already been packaged, stressing them under heat and power to catch weak parts before they go into expensive AI systems. Sonoma is becoming a foundational tool for the AI processor industry, driven by its ability to handle extreme power loads that traditional test equipment simply cannot sustain.

During the quarter, they completed development of a next-generation fully automated higher-power Sonoma system supporting up to 2,000 watts per device.

Erickson gave useful context on what that means competitively: “We have some devices that we’re going to be testing this spring that are almost 2,000 watts per device, right? Everybody’s out there talking about how can you do what does it take to get to 1,000 watts? We’re jumping right past that.”

As AI ASICs and GPUs grow more complex, they draw massive amounts of power and generate extreme heat. Getting to 2,000 watts per device creates a meaningful technological moat in a market where the thermal demands of cutting-edge chips are outpacing what legacy test platforms were designed to handle.

To meet anticipated demand from a major hyperscaler planning to begin shipments in Q1 FY27, Aehr is scaling manufacturing capacity aggressively. They will begin shipping Sonoma systems from an additional contract manufacturer this quarter, adding capacity of more than 20 systems per month, which meaningfully expands its ability to support future growth.

Read more

The biggest improvement was the balance sheet. POET said it now has roughly $430 million in cash, which materially changes the setup versus the old story where financing risk sat over everything. That gives them room to scale manufacturing, hire, and pursue acquisitions without the same near-term capital pressure. Hopefully. I have felt this in the past as well and they raised more. But $430m is a lot of cash for a $850m mkt cap company.

The second improvement was execution visibility. Management laid out a more explicit 2026 production cadence with light-source products in Malaysia beginning in Q2 2026, followed by high-speed 800G optical engines in Q3 2026. That is more concrete than the older broad commercialization language.

The third improvement was strategic framing. POET leaned harder into the idea that its opportunity extends beyond standard transceiver engines, putting more emphasis on ELSFP, high-power applications, CPO, and even GPU-to-GPU and GPU-to-memory optical interconnects (wink wink Celestial/Marvell)

What was genuinely new

Read more