RALLİ AVCISI

$LITE just delivered an incredible quarter. Revenue reached $1.263 billion, up 136.5% year over year. Non-GAAP EPS was $4.57 and non-GAAP operating margin hit 42.3%. The next-quarter outlook was even stronger: revenue near $1.6 billion, operating margin above 45%, and a second major NPO/CPO win backed by more than $5 billion in purchase commitments. Demand for lasers and optical interconnects is clearly not the issue. The market is focused on the deal structure. The $MRVL acquisition uses a fixed exchange ratio of 4.5 Marvell shares for every Lumentum share. The $1,132 figure was the implied value when the deal was announced, not a guaranteed cash value. With $LITE around $536 on Friday, the gap mostly reflects the drop in Marvell’s share price, the time until closing, and deal risk. The business is still performing. Holding Lumentum now also means holding a claim on future Marvell shares, so the value moves with $MRVL.

A lot of people are asking the same questions about $COHR’s new supply agreement with Guangzhi Technology’s subsidiary: Why this supplier? Why six-inch wafers? And how many companies can actually produce six-inch InP substrates at the required quality and scale? Jim Anderson gave a useful clue on the earnings call. In ultra-high-power lasers, especially 400 mW CW lasers, Coherent sees only one real competitor and very few others operating at the same level. That helps explain why six-inch InP matters so much. Coherent’s advantage is the depth of its photonics stack. It covers light generation, lasers, fiber transmission, optical assemblies, and detection. Very few companies control that much of the signal chain. The agreement is still a letter of intent, not a final purchase contract. Guangzhi’s controlled subsidiary, Xianrui Technology, expects to supply at least $601 million of six-inch InP substrates to Coherent’s plants in Switzerland and Sweden from 2027 through 2029. Final terms, orders, and export permits still matter. That is why this six-inch platform is strategically important for both sides.

$COHR jumped more than 13% in one session. Three things drove the move. First, the quarter and outlook beat expectations. Revenue reached $2.37 billion, up 50% year over year. Data Center & Communications grew 79%. 800G shipments kept scaling, 1.6T continued to ramp, and the next-quarter revenue outlook moved up to $2.55–$2.75 billion. Second, Coherent signed a major five-year PhotonLink agreement with an estimated value of about $20 billion. The program covers a broader optical solution anchored by 400 mW CW lasers, with revenue expected to start ramping in early 2028. Alongside the $NVDA relationship, it gives the market a clearer view of Coherent’s value as an independent photonics platform. Third, the stock was coming off a sharp selloff while the capacity barrier remained intact. Internal InP output doubled year over year and is still not enough to cover demand. Coherent plans to more than double it again by the end of 2027. That vertical integration supports both supply control and future margin improvement.

$NIXT’s return can be split into two pieces: gains from valuation recovery and gains from revenue and margin improvement. Based on the attribution used here, more than half of the operating-improvement contribution came from $LITE. Lumentum accounted for 7.6 of the 14.5 percentage points in that bucket, or roughly 52%. Against the fund’s reported 2026 return in this analysis, that works out to about 4.7 points of contribution. The key is that NIXT did not chase Lumentum after the rally had already started. It owned the stock when the market had largely given up on it, then held through the recovery. 1. Lumentum fell out of the top 1,000 U.S. companies by market value during 2023–2024. That fits the fund’s deletions strategy, which targets companies removed from major large-cap universes after a steep decline. Fiscal 2024 revenue fell 23%, the company lost about $550 million, and operating cash flow dropped to roughly $25 million. The stock fell from about $106 at the end of 2021 to around $44 in April 2024. That is exactly the kind of setup the strategy looks for: a neglected company that has been pushed out of a major index. 2. Lumentum also passed the quality screen. The rules remove the weakest 20% of candidates. Buybacks helped its shareholder-distribution score, while convertible issuance worked against it. The detailed company score is not public, but Lumentum stayed in the eligible group. 3. The fund owned it from the start. In the first disclosed portfolio dated September 30, 2024, Lumentum had a 0.91% weight. 4. The rebalance rules allowed the position to ride. Lumentum was still below the size cutoff at the April 2025 rebalance and remained in the portfolio at roughly an equal weight. Over the following year, AI optical demand took off and the stock rose from around $59 to roughly $902. Because the strategy rebalances only once a year, it did not trim the winner along the way. By the time Lumentum no longer qualified at the 2026 rebalance, its weight had grown sharply. The stock later pulled back, so the exit timing also helped. How should investors think about NIXT? The strategy is not designed to identify one future superstar. It owns a broad group of companies the market has discarded and waits for a few of them to catch a powerful recovery cycle. That is how a deletions strategy is supposed to work, but nobody knows in advance which holding will become the next Lumentum or how far it will run. Even when Lumentum made a strong contribution, the fund could still lag the broader market in other periods. The simple takeaway: NIXT is a diversified small- and mid-cap value portfolio with some built-in upside from turnarounds. The long-term record is still short, so it is worth following rather than treating one great winner as proof the strategy always works.

I just listened to Lumentum's CEO discuss the AI selloff and Marvell's all-stock acquisition. The timing could hardly have been worse. Lumentum fell close to 40% last week as investors worried about data-center capex and then had to digest the $MRVL stock deal at the same time. The CEO still believes the combined company can create long-term value. The strange part is that customer demand and order visibility continue to improve while the public market is selling AI-related names aggressively. The biggest complication is the structure: each $LITE share converts into 4.5 $MRVL shares. As Marvell's stock falls, the value received by Lumentum holders falls with it. The companies expect the deal to close in the first half of 2028, subject to shareholder approval and regulatory reviews in the U.S. and other jurisdictions. His conviction has not changed. Lumentum reviewed Marvell's financials and new programs during due diligence and liked the long-term opportunity. That is why the board preferred stock consideration and continued ownership in the combined business. Both CEOs have experience completing difficult regulated acquisitions. Even so, the next major test is the long approval process.

Here is my view, plain and simple: No matter which party is in power, the U.S. cannot afford to slow down on AI. This is a global race, and any pause gives competitors in Europe, India, and elsewhere more room to catch up. When these big tech stocks rise, issuing debt becomes easier and financing gets cheaper. That gives the companies more money to keep building compute. I do not expect AI spending to stop. The selloff across optical networking, storage, and semiconductors looks heavily driven by fear and leverage. The moves over the last two days have been brutal. Time will tell. Maybe it takes a month or two, maybe six. Looking back, this semiconductor pullback could end up feeling a lot like the tariff-war selloff. My approach: I am trimming exposure in other areas and putting more focus on AI hardware, but only at a pace I can handle. I am also trying to build more cash flow, keep spending under control, and use this pullback carefully instead of chasing every bounce. ⚠️ Not financial advice.

$LITE just reported a huge quarter, but the cash-flow picture still deserves a closer look. Revenue reached $1.263 billion, up 136.5% year over year. Profit and EPS looked great on paper. But the company is still spending heavily to expand capacity. Cash, cash equivalents, and short-term investments fell by about $759 million during the quarter. Some of last year’s large GAAP loss also came from non-cash accounting tied to convertible-note conversions, not a collapse in the core business. This is the expensive part of the cycle. Revenue is growing fast, but factories, equipment, inventory, and working capital absorb a lot of cash. That is a common issue across the optical-hardware group. The stock has also been hit hard after a huge run, as the market quickly cuts the valuation it is willing to pay. There is another layer now: $MRVL has announced a definitive all-stock deal to acquire Lumentum. That raises real questions about customer relationships, supply-chain choices, dilution, and regulatory approval. The market’s view is pretty simple: the growth is real, but perfect execution is already expected—and that is a high bar. ⚠️ Not financial advice.

Near-term issues are real. $AAOI feels like the market is treating the capacity delay as weaker demand. U.S. expansion is getting slowed by power and construction, so the year-end 800G/1.6T monthly capacity target got cut from around 650k to 500k+. Demand on the other side didn’t change. They actually raised the end-of-2027 target from about 930k to over 1M units per month. The LTA actually improves 1.6T visibility. Right now the biggest question to me is when the capacity catches up.

$CRDO Credo just cleared an important milestone. Two optical-telemetry specifications developed with $ORCL were approved by the Open Compute Project. In plain English, the standards help operators spot weak optical links before they cause an AI cluster outage. That matters as networks move toward NPO and #CPO, where every link has to stay reliable at much larger scale. Credo is also moving ahead with 1.6T transceivers and the SRMO short-reach optics workstream for high-speed links inside and between racks. That keeps it closely tied to the hyperscaler ecosystem. The market is not rewarding the news right now. Investors are focused on whether cloud providers slow their capacity buildouts and push long-term orders further out, so the whole optical group remains under pressure. ⚠️ Not financial advice.

$AAOI Q3 revenue came in at $269.9M, up 128% year over year. Data center revenue was $155.6M, up 254%. Back to GAAP profitability. Q4 guide is $385M–$420M. Demand is finally showing up as actual revenue. Right now hyperscalers care more about whether the 800G and 1.6T modules actually get delivered. AAOI just signed a multi-year strategic agreement with a Tier 1 hyperscaler, which improves long-term visibility. Management was asked if customers are willing to pay a premium for U.S. manufacturing. The answer was yes. That’s the main reason I’m watching the U.S. capacity expansion. They’re aiming for over 500k units/month of 800G/1.6T by the end of 2026, and over 1M units by the end of 2027. They also said the 2027 U.S. expansion will be slower than originally planned because of power and construction issues. That could keep pressure on the stock for a while.

Let’s talk plainly about the $MRVL deal for $LITE. A lot of shareholders on both sides are unhappy, just for different reasons. Some Marvell holders think the company is paying too much. Lumentum holders will own about 33% of the combined company after closing, so the dilution for existing Marvell owners is substantial. The promised synergies also have to be proven; putting two strong businesses together does not automatically create the returns management expects. Regulation is another major risk. The deal needs shareholder approval and multiple regulatory clearances. Because Marvell supplies DSPs while Lumentum owns important laser and optical capacity, regulators may look closely at access, bundling and whether competitors can still buy key components on fair terms. Lumentum shareholders have a different complaint. The deal is all stock, with no cash. Each $LITE share gets a fixed 4.5 shares of $MRVL, so the number of shares is locked but the dollar value is not. When the deal was announced, that ratio implied about $1,132 per LITE share. If MRVL falls, the value offered to Lumentum holders falls with it. There is no fixed cash floor. The wait is long too. The companies expect closing in the first half of 2028, subject to approvals and other conditions. That leaves plenty of time for markets, regulation and both businesses to change. So the doubts are understandable: a rich price, meaningful dilution, unproven synergies, regulatory uncertainty, a moving stock-based value and a long road to closing. That does not mean the deal will fail. It means the spread is there for a reason. This is a summary of the market debate, not investment advice.

The market is waiting for $AAOI’s Q3 report after the close today. The company’s angle is different from many optical-module suppliers: it is making a big bet on U.S. manufacturing. In October, AAOI completed a $600M ATM stock offering and raised about $588M net. The company said the money may support debt repayment, working capital and capex, with a major focus on expanding production for data-center optical products. Its Texas expansion is aimed at increasing 800G and 1.6T module capacity. That matters as large cloud customers look for more U.S.-based supply and try to reduce tariff, regulatory and delivery risk. Management has said demand is running ahead of capacity and that customer visibility extends through at least the second half of 2027. Tonight, I am watching two things. First, whether 1.6T qualification and orders are still moving on schedule. Second, whether the Texas production ramp is keeping up with the plan. The risk is straightforward. Raising money through new shares dilutes existing holders. If factory output slips, yields disappoint or a large customer spreads orders across more suppliers, the story gets harder to support. $AAOI is not only competing on cost. A big part of the bet is whether customers will pay for a more local and dependable supply chain. Not investment advice.

$COHR really did blow past expectations this quarter, so let’s keep the story simple. Fiscal Q1 revenue reached $2.37B. Reported growth was 50% year over year, or 56% on a pro forma basis. Data Center & Communications revenue jumped 79% and made up 82% of the total. AI customers are still pulling in more optical hardware than the supply chain can easily deliver. PhotonLink is the biggest new piece. This is not just another pluggable optical module. It is Coherent’s integrated platform for CPO, NPO and chip-to-chip links as optics move closer to GPUs and switches. The company disclosed a five-year PhotonLink agreement worth about $20B, with revenue expected to start ramping in early 2028. PhotonLink revenue from other programs is expected to top $100M in the March 2027 quarter and reach more than $250M a quarter before the end of calendar 2027. The bottleneck is still InP. It is the key material behind high-speed lasers, and qualified capacity takes time to build. Coherent doubled internal InP output a quarter ahead of schedule and plans to more than double it again by the end of 2027. Management expects fiscal 2027 pro forma revenue growth above 60%. For fiscal Q2, the official guidance is $2.55B–$2.75B in revenue, not $3B. It also expects its first quarter above a 45% non-GAAP gross margin before the end of calendar 2028. With $LITE now set to be acquired, Coherent remains one of the clearest independent bets on the next CPO/NPO buildout. The risks are just as clear: InP capacity has to ramp on time, yields have to hold, and customers can still move their deployment schedules. Not investment advice.

One easy-to-miss detail with $AAOI: 800G and 1.6T modules share the same production lines. The main difference comes near the final testing stage. That is why management usually reports combined 800G and 1.6T capacity instead of giving a separate number for 800G. It also explains why the widely repeated figure of 90,000 800G units per month often does not match the company’s own numbers. The definitions are different. AAOI is targeting combined capacity of about 650,000 800G and 1.6T units per month by the end of 2026, with roughly 30% produced in Texas. By the end of 2027, the target is around 930,000 units per month, with more than half of capacity in Texas. There is another important point: installed capacity is not the same as sellable output. In Q1 2026, 800G revenue was only $4.6 million, or 5.6% of data-center revenue. The company kept full-year revenue guidance near $1.1 billion and repeatedly said demand was not the problem. Capacity and the supply chain were. Management expects demand for 800G and 1.6T to stay ahead of capacity through at least mid-2027, with orders extending into Q2 2027. Execution still matters. AAOI has missed capacity and yield targets before. It also buys high-speed EML chips and DSPs from outside suppliers. A finished factory does not guarantee that usable yields will follow on time. Even if capacity reaches 930,000 units per month by the end of 2027, the scale would still be below the largest Chinese module makers. Market estimates put Innolight above two million units per month and Eoptolink around 1.3 million to 1.4 million. High-speed modules are also heavily customized, and qualification takes time. Extra capacity alone does not let one supplier take market share overnight. One last caution: several of these capacity figures come from earnings-call summaries and market estimates, not formal company filings. Most were published between May and November 2026. If management updates the outlook, the latest company guidance should take priority. The standalone 90,000-unit figure is especially easy to misread because the reporting basis is unclear.

$INTC has fallen all the way from the $90s. After earnings, it suffered its worst one-day drop since 1974. The stock looks worse than the balance sheet. Intel ended Q3 with about $50.1 billion in total debt and $54 billion in cash and short-term investments. It also raised $22.6 billion by issuing new shares, which shows how much cash this turnaround still needs. The server business is still growing, and the foundry business is making progress. At this price, I think Intel is worth researching and building a position in slowly. CEO Lip-Bu Tan recently bought about $9.7 million of shares. The Pelosi household also disclosed purchases of Intel shares and call options earlier this year. Neither guarantees the turnaround will work, but both give me a little more confidence. I usually don’t tell people to buy stocks 😅 But with $INTC under $60, honestly, I think it makes sense to start small and give the turnaround time. This is the kind of price where I’d be comfortable buying a little and holding it for years. Intel still has a lot to prove. Now it needs time.

A closer look at $AAOI and the production ramp for 800G and 1.6T optical modules. This is an important stretch for Applied Optoelectronics. Its 800G modules are already shipping in volume, while 1.6T is nearing the end of qualification and preparing to ramp. Management has been pretty direct: output will decide how much revenue the company can generate. Demand is running ahead of capacity, so the ceiling depends on how quickly production scales, yields improve, and customer qualification is completed. Start with 800G. Volume shipments are already underway, and the process is more mature. On the Q2 call, the company said 800G shipments more than doubled sequentially. Management also said demand for 800G and 1.6T should remain above capacity through at least mid-2027, with orders extending into Q2 2027. The problem is no longer finding buyers. It is delivering enough product. The industry has had more time to improve 800G assembly, so yields and production stability are better. AAOI has already made it through qualification, purchase orders, and the first part of the capacity ramp with a major cloud customer. Now 1.6T. Qualification is close to completion, initial shipments are expected around the end of Q3, and management is targeting a bigger ramp in Q4. The company expects more than $70 million of 1.6T revenue in Q4 2026. It also expects 800G sales to move above older 400G products later this year. The hard part is yield. Overall 1.6T assembly yield is still below the more mature 800G process, and DSP supply and yield remain constraints. Most leading 1.6T DSPs use 3nm, while the earlier 5nm generation struggled with power. The entire industry is still working through that ramp.

$COHR set a high bar last quarter. Revenue jumped, margins improved, and the comparison for this report is much tougher. Here is the fiscal Q4 2026 scorecard for the quarter ended June 30. Revenue reached $2.05 billion, up 34% year over year. Adjusted EPS was $1.74, up 74%. GAAP EPS came in at $1.19 versus a loss of $0.83 a year earlier. Adjusted gross margin rose to 40.2%, up 215 basis points. Adjusted operating income reached $446 million, up 62.1%, with a 21.8% operating margin. Data Center & Communications generated roughly $1.6 billion, close to 80% of total revenue. It remains the main growth engine. For the full fiscal year, revenue reached $7.118 billion, up 23%. Net income was $805 million, with diluted EPS of $4.12. Data Center & Communications revenue rose 40% to $5.275 billion, while segment profit climbed 47% to $1.33 billion. Industrial revenue fell 10% to $1.844 billion, although segment profit still increased 4% to $423 million. R&D and selling and administrative expenses also moved higher as the company kept investing. The midpoint of the latest fiscal Q1 2027 outlook calls for $2.3 billion in revenue, up about 36% year over year, and adjusted EPS of $1.95, up about 12%. That is the key setup for the report: revenue growth is accelerating, while EPS growth is slowing. The market may be willing to forgive a small revenue miss if demand stays strong. A profit or margin miss would probably be harder to explain.

For the optical group, I am watching one thing: whether NPO, #CPO, and 1.6T orders turn into real shipments and reported revenue. $LITE just agreed to be acquired by $MRVL, reported a strong quarter, won a second large NPO/CPO program backed by more than $5 billion in purchase commitments, and started to see OCS revenue scale. Its next-quarter outlook moved higher again. Investors are debating dilution from the deal, but those long-term commitments are hard to ignore. $COHR has long-term capacity commitments stretching into 2028 and is focused on CPO, silicon photonics, and 1.6T. Its next earnings report will come down to margins and the pace of large-customer deliveries. Along with Lumentum, it remains one of the key names in the move toward NPO and CPO. $AAOI is expanding its Texas manufacturing footprint to ramp 800G and 1.6T module capacity. It has large 1.6T orders from cloud customers and offers a U.S.-based supply option close to hyperscalers. It is also a smaller company, so the upside and the volatility can both be much larger. Expectations across optical networking are already high. An announced opportunity is no longer enough. Investors want volume shipments and recognized revenue. The sector is moving from concept stories into execution. A faster ramp can still surprise the market. A delay can turn good news into a sell-the-news event very quickly.

The big question: if Lumentum’s business is this strong, why sell to Marvell now? The deal is all stock. Each $LITE share will convert into 4.5 $MRVL shares. First, why is Marvell willing to pay so much? As AI servers pack in more GPUs, copper eventually hits physical limits in bandwidth, power, and distance. The industry calls it the “copper wall.” Optical interconnects are moving from data-center links and rack-to-rack connections into the rack and closer to the package. With pluggable modules, customers could buy Marvell’s DSP, Lumentum’s lasers, and have a manufacturer assemble the module. CPO and NPO change that setup. The optical engine sits much closer to the GPU or switch, so the electrical and optical pieces need to be designed together from the start and packaged as a heterogeneous system. Marvell is a leading optical-DSP supplier, but it is fabless and does not own InP laser manufacturing. That process takes decades to build, and a new facility can take three to five years to qualify. Buying Lumentum gives Marvell scarce InP capacity and a product chain that runs from switches and DSPs to laser sources. It also strengthens Marvell’s position against Broadcom. Lumentum’s logic is different: lock in value near the top of the cycle and reduce the risk of a future downturn. The implied enterprise value of about $115 billion is already close to Lumentum’s potential 2030 market opportunity. Taking Marvell shares lets Lumentum holders trade the volatility of a standalone optical-component company for ownership in a larger, more diversified business. Michael Hurlston also sold Finisar near an industry high. And as #CPO grows, standalone component vendors may have less influence over the final architecture. Combining with Marvell gives Lumentum a seat earlier in hyperscaler system design. Marvell does not have enough cash to fund a deal this large, so the consideration is entirely stock. The companies say the combination will expand their addressable market to $550 billion, create more than $1 billion in annual run-rate cost synergies, and add to Marvell’s non-GAAP EPS in the first full year after closing. The deal could reshape the optical supply chain. Competitors may look for new laser suppliers, which could raise the value of the remaining independent optical-chip companies. Module makers may spread DSP orders across more vendors, and neutral suppliers could become more attractive. Closing is expected in the first half of 2028, so regulatory review will remain a major variable. I’m watching three things: Whether Lumentum delivers its guidance and ramps OCS and the $5 billion-plus CPO program on schedule. Whether module makers shift DSP orders and whether independent optical-chip vendors win more hyperscaler business. Regulatory progress and how moves in Marvell’s share price change the value of the fixed exchange ratio.