Marvell's $115B Lumentum Deal: Why Both Stocks Sank on Optical Consolidation

This piece answers one question: Marvell calls this deal "strength into strength," so why did both the buyer's and the target's shares drop hard on the same day, and what risks is the market pricing? Data as of the Nov 2, 2026 close (ET), with after-hours quotes as of 5:37 pm ET. Sources: SEC 8-Ks, yfinance, Eastmoney/Tencent quotes, X (via Grok search), public media.

1. What happened

Before the open on Nov 2 (ET), Marvell (MRVL) announced an all-stock deal to acquire optical component maker Lumentum (LITE). Each LITE share converts into 4.5 MRVL shares, implying an enterprise value of about $115B. Based on both companies' unaffected Oct 28 closes, that works out to about $1,132 per share, a premium of about 23.9%. Lumentum also reported a very strong quarter the same morning.

The market didn't buy it. MRVL closed down 12.72% at $171.41 and LITE closed down 18.59% at $684.40, and the two tied for the top of Aya's US heat board. The longer view is starker. Bloomberg first reported the talks after the close on Oct 28. Measured from that close through Nov 2, MRVL fell 31.9% and LITE fell 25.1%, while the semiconductor ETF (SMH) fell 10.5%. Both stocks did far worse than the sector.

One number makes it concrete. At Monday's closes, 4.5 MRVL shares were worth $771.35, 12.7% above LITE's close. In other words, the stock LITE holders will eventually receive is now worth 31.9% less than the $1,132 headline. After hours, the two stocks recovered 1.22% and 1.55%, which didn't change the picture.

Why would a deal billed as "number one plus number one" get both sides sold? We answer three questions below: why the stocks fell, whether the synergies can be delivered, and whether the deal can close. We then look at the effect on Coherent, Ciena and Chinese optical companies.

2. Background: why a chip company wants to own lasers

First, what each company does. Marvell is a data center chip company. Its products include the DSPs inside optical modules, custom AI chips (custom ASICs) designed for cloud companies, and switch chips. A DSP is a digital signal processor that "cleans up" signals during high-speed transmission. Marvell built most of its portfolio through acquisitions: Fabricated Knowledge (2022-12-18) notes that buying Cavium, Inphi, Innovium and others turned it from a storage-controller company into a networking infrastructure company. Lumentum makes the most upstream parts of optical communications, mainly indium phosphide laser chips, which optical modules use to turn electrical signals into light. It also makes pump lasers and optical circuit switches (OCS), which redirect light paths with micro-mirrors instead of converting the signal back to electricity first.

Management's core argument is the "copper wall." On the merger call, Marvell said copper's reach shrinks quickly as data rates rise, so optics has to move steadily closer to the compute chip. The end state is CPO (co-packaged optics: the optical engine sits in the same package as the switch or compute chip) and NPO (near-packaged optics: the optical engine sits right next to the chip but in a separate package). At that point, the chip design, silicon photonics, lasers and packaging need to be "co-architected and delivered by a single supplier." Tech analyst @PatrickMoorhead summed it up well on X: Inphi brought the DSPs, Celestial AI the photonic fabric, and Lumentum the lasers, "the scarcest part of the AI network." His verdict: "Right strategy, big price."

Lumentum's same-day results show this is no distressed asset. In the quarter ended September 2026 (FY27 Q1), Lumentum's revenue was $1.26B, up 136.5% year over year. Its non-GAAP operating margin was 42.3%, versus 18.7% a year earlier. It guided next-quarter revenue to $1.575B–$1.625B. It also announced its second major NPO/CPO win, backed by purchase commitments of more than $5B. Management puts the combined company's annualized revenue at about $17.7B, rising to more than $21B within one quarter.

This chart is the starting point for everything that follows: Marvell is buying a company at the peak of a supply shortage, with revenue that doubled in a year. Both the upside and the controversy come from that.

3. Question one: why did buyer and target fall together?

Our view: the market isn't worried about dilution as such. It's worried that Marvell is trading a third of the company for earnings that may be at a cyclical peak, and that this can't be tested for 18+ months. Add the mechanical selling from merger arbitrage and a sector-wide sell-off, and you get both stocks falling at once.

Start with dilution. LITE has about 101.7M diluted shares. At 4.5x, Marvell has to issue about 458M new shares, or 52% of its current share count. The company says Marvell's existing holders will own about 67% of the combined company. One X critique was that "the target has no earnings and the acquirer is using its multiple," but that's wrong: Lumentum earned $3.54 in GAAP EPS this quarter. As a rough run-rate check, LITE's next-quarter non-GAAP EPS guidance midpoint, annualized, works out to about $5.38 per newly issued MRVL share. Marvell's own current-quarter consensus EPS, annualized, is about $4.40. At today's earnings, the deal adds to EPS on paper, consistent with the company's claim that it will be accretive to non-GAAP EPS in the first year after closing.

So the dispute isn't the arithmetic. It's how long those earnings last. One analyst on the call asked directly: the deal values Lumentum at about 1x its 2030 addressable market, while Marvell itself trades below 0.5x. Is Marvell buying "at the peak of its shortage environment"? Murphy answered that the supply shortage is a symptom and the real driver is "a shift in the architecture inside the data center." X account @RJCcapital relayed that BofA downgraded MRVL and cut its target to $225 from $400, saying $1B of cost cuts "falls well short of offsetting 50% share count dilution," and questioning whether Lumentum's valuation holds once laser shortages ease. We couldn't verify this downgrade publicly and treat it only as a market view. Korean industry KOL @jukan05 went further: the only way to read Lumentum management accepting the deal is that they have no faith in the current AI cycle and expect the market to collapse in two to three years. The post drew 1,606 likes. That's his speculation, with no evidence behind it, but it captures the market's concern: a seller willing to swap into the buyer's stock at the top is itself read as a signal.

Next, the selling pressure built into the deal timeline. The deal isn't expected to close until 1H 2028, an 18–20 month wait. The standard arbitrage on an all-stock deal is to buy LITE and short 4.5x as much MRVL, so after the announcement those funds keep shorting MRVL. MRVL traded 61.05M shares on Nov 2, against a three-month average of 21.72M. LITE fell even more than MRVL that day, and the drop splits into two parts: MRVL's own decline, and a wider deal spread. At the Oct 30 close the spread, measured at the 4.5x ratio, was only 5.1%; on Nov 2 it widened to 12.7%. One caveat: the ratio hadn't been announced on Oct 30, so the 5.1% is backed out from terms disclosed later.

Finally, the sector. This was already a sell-off week for AI hardware: Intel crashed 28.58% on Oct 30, and on Nov 2 Ciena fell 11.34% and Astera Labs 11.22%. Not all of the decline belongs to the deal. Roughly netting out the sector, MRVL fell about 21.4 percentage points more than SMH, worth about $47B of market value. That's only a rough estimate, and MRVL's beta of 2.33 means it normally swings more than the sector anyway.

Two things stand out in the chart. MRVL dropped 13% on the day of the report, so the market disliked the "all-stock, big premium" structure from the start. LITE barely moved for two days, then fell sharply once formal terms came out, because from that moment its price was tied to MRVL through the 4.5x ratio.

4. Question two: can the $1B in savings and "copper to optics" pay off?

Our view: the strategy has orders behind it, but $1B in cost savings is too small to support the valuation. The real value depends on when CPO ships at scale, and that can't be tested until 2027–2028.

Start with the size of the savings. The company says it can save more than $1B a year, "primarily from COGS," or about 5.6% of the combined company's $17.7B annualized revenue. CFO Dan Durn explained that many of Marvell's current solutions still use components from suppliers other than Lumentum, and those can be switched to in-house parts after the merger. One implication is easy to miss: part of the savings comes from taking share from other laser suppliers, such as Coherent, rather than appearing out of thin air. On valuation, one analyst (Korea Invest Insights, on Substack) values the synergies at about $21.3B using a 25x after-tax multiple. Meanwhile, the two companies together lost about $91B of market value from Oct 28 to Nov 2, far more than that figure even after netting out the sector. The market clearly isn't pricing the deal on management's synergy numbers.

Next, the evidence for "copper to optics." The supporting side has three points. First, orders: Lumentum's second NPO/CPO win comes with more than $5B in purchase commitments, and management describes OCS as a future "multi-billion-dollar annual revenue stream." Second, technology: SemiAnalysis (2026-07-02), covering ECTC 2026, reports that Marvell showed an optical multi-chip interconnect bridge (OMIB) that puts six photonic chips and one compute chip on a single organic interposer, reaching a bandwidth density of 1.8 Tbps/mm². TrendForce (2026-07-03) also sees Broadcom and Marvell as early CPO leaders. Third, the gap is real: Marvell has DSPs, silicon photonics and packaging but no indium phosphide laser capacity of its own, while Lumentum holds "the world's largest indium phosphide manufacturing footprint."

The evidence against deserves equal space. First, new architectures often spread more slowly than expected. Fabricated Knowledge (2023-12-17) looked back at LPO (linear-drive optical modules that drop the DSP): it was once seen as a threat to Marvell, but it reached commercial use much more slowly than promoted, and by the time it matured the industry had moved on to the next speed generation. CPO could see a similar delay. Second, customer conflict. Marvell sells DSPs to optical module makers, Lumentum sells lasers to the same module makers and also makes modules itself. After the merger, those customers face an upstream supplier that is also a competitor. Murphy said on the call that the ecosystem runs on "co-opetition" and supply will continue, but there's no data yet on whether customers will move orders to a second source. Third, the cycle: if the laser shortage eases in 2027 and prices fall, whether Lumentum can hold its current 42% operating margin is the single biggest variable in the deal's return.

5. Question three: what do regulatory approval and the deal spread tell us?

Our view: the 12.7% spread mainly reflects time, hedging cost and uncertainty, not a simple probability of completion. The real regulatory risk most likely sits in China.

From the published terms, the deal needs US antitrust clearance (HSR) plus approvals under "certain other specified antitrust or foreign investment laws." The full list of required approvals is in a disclosure letter that isn't public. The public text doesn't name China, and the call didn't discuss Chinese approval. The merger agreement sets three termination fees. If Marvell's board changes its recommendation or Marvell takes a superior proposal, Marvell pays Lumentum $5.8B; in the mirror situations, Lumentum pays Marvell $2.9B. If the deal fails for regulatory reasons, Marvell pays $3.2B, or about $31.5 per LITE share.

Why China? First, the size of the deal and the nature of the industry. Several of the world's top optical module makers are Chinese, such as Innolight and Eoptolink, and they all buy upstream lasers and DSPs. The merger concentrates that upstream supply, which is exactly what antitrust reviewers focus on. Second, there's precedent. Fabricated Knowledge (2024-04-08) recounts how the Silicon Motion acquisition was terminated amid geopolitical factors, leaving the buyer with both a break fee and a credibility hit. The long timeline to a 1H 2028 close also suggests both sides expect a lengthy review.

How should the spread be read? At the close, LITE traded 12.7% below its exchange value. If the deal closes in 17–20 months, that's about 7.4%–8.8% annualized. We couldn't find LITE's borrow-cost data, so this return is not net of borrow and hedging costs and is only an estimate. Backing out a completion probability requires first assuming what LITE would be worth if the deal broke. At a break price of $600, the implied probability is about 49%; at $500, about 68%. The two assumptions are nearly 20 points apart, which shows how much the result depends on the assumption. There's another wrinkle: Lumentum's earnings are growing fast, so its standalone value after a break may not be below today's price. On the other side, arbitrageurs have to stay short a beta-2.33 stock for a year and a half, and that risk alone calls for extra compensation. One analyst calculated an 8.2% spread at an earlier point; it widened to 12.7% by the close, which suggests the formal terms, especially the timeline, made the market more cautious.

The chart shows that LITE holders' consideration has already lost about a third of its value, and the stock trades below even that. The regulatory break fee is only about $31.5 per share, so it does little to support LITE's price.

Key deal terms at a glance:

6. Effect on Coherent, Ciena and Chinese optical companies

Coherent (COHR): a beneficiary in theory, but the market isn't buying it yet. Once the deal closes, Coherent becomes the main laser maker not owned by a chip company. The media's argument is that customers who want a neutral supplier could send it more orders. The price action says otherwise. COHR fell 4.93% on Nov 2, less than Ciena, but it is down 16.5% since Oct 28, about 6 points worse than SMH. The market clearly isn't treating it as a winner; it looks like the same doubt about how long the laser shortage lasts is being applied to Coherent. Coherent reports on Nov 4, with consensus at $1.97 EPS and $2.30B revenue. What it says about laser demand and pricing will be the first hard data on the "peak shortage" thesis.

Ciena (CIEN): looks more like a sector sell-off, with no clear link to the deal. Ciena fell 11.34% on Nov 2 and is down 21.7% since Oct 28. We found no evidence tying its decline directly to this deal; some media over the same period attributed it to profit-taking across optical stocks, so we don't assign a cause here.

Innolight, Eoptolink, YOFC: their Nov 2 moves were not a reaction to the deal. The timeline matters here. The formal announcement came before the US open on Nov 2, which was the evening of Nov 2 in Beijing, after the China A-share and Hong Kong markets had closed. So Innolight's 0.54% decline, Eoptolink's 2.31% decline and YOFC's 11.49% gain that day (+11.7% on the heat board) can't be read as reactions to the terms; at most they reflect the deal rumors circulating since Oct 29. The first real reaction comes in the Nov 3 Asian session.

Over the longer term, module makers now face an upstream supplier that has become a vertically integrated potential competitor. And if CPO spreads, value shifts away from assembling pluggable modules toward chips and lasers. YOFC makes optical fiber and cable, so the extra fiber that "copper to optics" brings helps it no matter who wins the module layer. We couldn't find the specific reason for its jump on Nov 2.

7. What the market is betting on

Here's one way to read current prices. Per yfinance, MRVL trades at about 23.4x forward earnings and LITE at about 18.7x, so the target is actually the cheaper stock. On consensus numbers, one LITE share carries about $36.56 of forward EPS, while the 4.5 MRVL shares it converts into carry about $32.98. If consensus is right, Marvell is using its higher-valued stock to buy lower-valued, faster-growing earnings, which looks like a good trade on the numbers alone.

The market is still selling hard, which means it's betting on at least one of two things. First, LITE's consensus can't be trusted: today's earnings rest on pricing power from the laser shortage and will fall once the shortage eases. Second, Marvell deserves a lower valuation: about a third of the combined company would be capital-heavy, highly cyclical optical component manufacturing, which should carry a lower multiple.

Are these bets reasonable? The first has some basis: Lumentum's margin jumped from 18.7% to 42.3% in a year, and a slope that steep points to a strongly cyclical business. But there's counter-evidence too: more than $5B of CPO purchase commitments and the OCS orders aren't purely shortage-driven profit. The second is a question of valuation framework and can't be disproven in the short run. Also, the 44 analysts covering MRVL still have an average price target of $340.56, and most haven't updated for the deal yet, so that figure says little for now.

8. What to watch next

• Nov 3 Asian session: the first reaction from Innolight, Eoptolink and YOFC to the formal terms. If module makers fall broadly, the market is pricing upstream vertical integration squeezing the module layer.
• Nov 4, Coherent earnings: consensus $1.97 EPS and $2.30B revenue. Listen for whether laser prices are softening, and whether any customers are shifting orders for "neutral supply." The first bears on how durable Lumentum's earnings are; the second on whether Coherent actually benefits.
• Dec 1, Marvell FQ3 earnings: consensus $1.10 EPS and $3.15B revenue. This is management's first earnings call to detail synergies and dilution, and a check on whether Marvell's existing business growth is being obscured by the deal story.
• S-4 registration and proxy (dates not yet disclosed): these will list the full set of required approvals (including whether China's SAMR is on it), plus the earnings projections each management team gave its financial advisers. The projections speak directly to the "buying at the peak" question.
• The deal spread: a steadily narrowing spread means confidence in completion is returning; a widening spread means regulatory or fundamental doubts are deepening.
• Around Feb 2027, Lumentum FY27 Q2 results: whether it delivers on its $1.575B–$1.625B revenue guidance.

9. Conclusion

This $115B deal has orders and technical groundwork behind its strategy, and on current earnings it adds to EPS on paper. Both stocks fell because the market is pricing three things: Lumentum's earnings may be at the top of a shortage cycle, the 18–20 month timeline brings steady arbitrage selling and regulatory uncertainty, and AI hardware is deleveraging as a sector. The $1B in cost synergies is nowhere near enough to offset about $91B of lost market value, and for now the market only gives credit for what can be verified. The 12.7% spread at the close is not a clean probability of completion; it's the price of time, hedging cost and China approval risk combined. Coherent's earnings, Marvell's earnings and the S-4 will provide testable data, in that order.

Appendix: Data & Sources

A1 Key financials & valuation (as of Nov 2, 2026)

A2 Formulas & inputs

A3 Timeline (ET)

A4 Aya heat data (Nov 2, 2026 close)

A5 Market views (opinions, not facts)

• @PatrickMoorhead (Nov 2): "Inphi brought the DSPs, Celestial AI the photonic fabric, Lumentum the lasers, the scarcest part of the AI network. Right strategy, big price."
• @jukan05 (Nov 2, ~1,606 likes): Lumentum management accepting the deal suggests they have "no faith in the current AI cycle."
• @RJCcapital (Nov 2), relaying: BofA downgraded MRVL and cut its target to $225, saying $1B in cost cuts falls well short of offsetting ~50% share dilution. Not verified publicly.
• @marketswithmay (Nov 2): "the acquirer is clearly using their multiple."
• Korea Invest Insights (Substack, Nov 2): argues the premium is already absorbed by the market and synergies are priced at zero; stance "lean neutral to mildly positive."

Sources

• Lumentum 8-K Ex. 99.1 (merger release), Nov 2, 2026: https://www.sec.gov/Archives/edgar/data/0001633978/000119312526442758/d163169dex991.htm
• Lumentum FY27Q1 results (8-K Ex. 99.1), Nov 2, 2026: https://www.sec.gov/Archives/edgar/data/0001633978/000162828026068655/lite_ex991xq1fy27.htm
• Marvell Form 425 (call content), Nov 2, 2026: https://investor.marvell.com/sec-filings/all-sec-filings/content/0001193125-26-443876/d178787d425.htm
• Merger agreement break-fee terms (WebSearch summary of the Lumentum 8-K), Nov 2, 2026
• Korea Invest Insights, "Marvell Already Paid for Lumentum," Nov 2026: https://koreainvestinsights.substack.com/p/marvell-already-paid-for-lumentum
• Oct 29 rumor coverage (24/7 Wall St., Motley Fool and others, via WebSearch)
• yfinance: MRVL, LITE, COHR, CIEN, SMH daily prices, ticker info and earnings calendars, through Nov 2, 2026
• Eastmoney / Tencent quotes: COHR, CIEN, Innolight, Eoptolink, YOFC, Nov 2, 2026
• X (via Grok): @PatrickMoorhead, @jukan05, @RJCcapital, @CenterPointSec and others, Nov 2, 2026
• Research corpus: SemiAnalysis (Jul 2, 2026, ECTC 2026); TrendForce (Jul 3, 2026); Fabricated Knowledge (Dec 18, 2022; Dec 17, 2023; Apr 8, 2024)
• Aya daily report "US & HK Market Update (2026-11-03)"

For information and research discussion only; not investment advice.

🎙️ Podcast script (Chinese, ~20 min, 泓观阳略): read it here