What did Marvell agree to buy?
Marvell Technology said on Nov. 2 that it had signed a definitive agreement to acquire Lumentum in an all-stock transaction. Lumentum shareholders will receive 4.5 Marvell shares for each Lumentum share, implying about $1,132 per Lumentum share at the Oct. 28 closing price.
Marvell put the implied enterprise value at $115 billion, a 23.9% premium to Lumentum's unaffected closing price on Oct. 28, the last trading day before speculation about a deal began. Reuters reported the equity value at $102.54 billion, with the larger figure including debt. Marvell shareholders will own about 67% of the combined company, and Lumentum holders about 33% on a fully diluted basis.
Lumentum is a less familiar name than Marvell, but its products sit at the same chokepoint. The San Jose company makes high-performance lasers, optical transceivers and optical circuit switches, and it runs specialized compound semiconductor manufacturing, including indium phosphide. Those components turn electrical signals into light and route it across data centers. Marvell's chips process and route the data; Lumentum's optics carry it across distances.
Both boards approved the deal unanimously. Matt Murphy will remain chairman and chief executive, while three Lumentum directors, including CEO Michael Hurlston, will join Marvell's seven-member board. The companies expect to close in the first half of calendar 2028, subject to shareholder and regulatory approvals.
The headline premium is already out of date. Because the exchange ratio is fixed, the value of the offer moves with Marvell's stock. Reuters reported that Marvell had fallen about 22% since talks were first reported. At Friday's close of $196.40, the ratio was worth about $884 per Lumentum share, while Lumentum closed at $840.70. Because $884 is below the $914 unaffected close, the announced 23.9% premium had already turned into a slight discount in current value.
The market reaction was mixed. Marvell rose 2.8% in premarket trading, while Lumentum's shares were halted, Reuters reported. The 23.9% premium was measured against the price before deal speculation, not against the prices investors actually traded on Friday.
Why is optical networking the next AI bottleneck?
AI data centers are limited by how quickly data can move between chips. Copper cables work over short distances, but they lose signal and consume more power as speeds rise. Optical links convert electrical signals into light, travel farther and carry more data per fiber.
A modern optical link has several parts. A digital signal processor shapes the signal, a laser converts it into light, and a transceiver packages the optics. An optical circuit switch can then redirect light between paths without converting it back into electricity. Marvell sells DSPs, switches, retimers and custom AI chips; Lumentum sells lasers, transceivers and optical circuit switches.
The industry is moving optics closer to the chip. Near-packaged optics and co-packaged optics place optical engines beside the processor or switch, shortening the electrical traces between them. Lumentum said it secured its second major NPO/CPO win in the first quarter, backed by purchase commitments exceeding $5 billion. It also said demand visibility for optical circuit switching had expanded and could support a multibillion-dollar annual revenue stream.
The engineering tradeoff is not just distance. As lane speeds rise, copper's usable reach shrinks, and the chips that equalize the signal consume more power. Optics can carry more bandwidth per lane over longer distances, but the laser, modulator and photodetector add cost and complexity. Co-packaging helps by removing the pluggable module and the long electrical path to it, but it makes the optics part of the switch or processor package. That raises the bar for manufacturing yield and for co-design between the chip and optics teams.
The economics are large. BNP Paribas estimated that DSPs and lasers together account for about 75% of a traditional PAM4 transceiver's bill of materials, Benzinga reported before the deal was announced. The bank expects the data-center optical transceiver market to grow from about $35 billion in 2026 to $91 billion by 2030. Those are forecasts rather than guarantees, but they explain why Marvell wants both halves of the link.
Marvell's press release frames the shift as a move from discrete components toward integrated optical systems. Murphy said AI infrastructure is undergoing "a fundamental architectural shift toward optical connectivity." The deal is intended to let the combined company sell pre-validated optical subsystems, not just individual chips.
How fast is Lumentum's optical business growing?
Lumentum's fiscal first-quarter results, released the same day as the deal, show why it is an attractive target. Revenue was $1.26 billion in the quarter ended Sept. 26, up 136.5% from $533.8 million a year earlier and 25.5% from the prior quarter.
GAAP gross margin was 51.0%, and non-GAAP gross margin was 53.2%. Non-GAAP operating margin reached 42.3%. Components revenue, which includes lasers and pump products, rose 98.9% to $754.2 million and accounted for 59.7% of total revenue.
The systems side grew even faster. Systems revenue rose 228.8% to $508.4 million and made up 40.3% of the total, so it is growing faster than components and now represents a large share of the business. Margins expanded alongside volume: GAAP gross margin rose from 34.0% a year earlier to 51.0%, while GAAP operating margin went from 1.3% to 28.5%. Those gains show operating leverage, but they also raise the question of how much further they can stretch.
Guidance points to more growth. Lumentum forecast second-quarter revenue of $1.575 billion to $1.625 billion, a non-GAAP operating margin of 45% to 46%, and non-GAAP diluted earnings per share of $5.90 to $6.20. Hurlston said the company expects about $350 million of sequential growth, more than the entire company's quarterly revenue two years earlier.
The guidance is a company forecast, not a guarantee, and it reflects a period of exceptional AI spending. The key question for Marvell is how much of that demand is durable capacity buildout and how much is a cyclical rush.
Why does Marvell want Lumentum now?
Marvell has spent years moving from storage and networking chips into custom AI silicon. Reuters noted that revenue rose from $3 billion in fiscal 2021 to $8.2 billion in fiscal 2026, and that its market value has climbed above $170 billion. In August, Marvell signed a deal to supply chips alongside Google's AI processors, worth up to $120 billion in sales through fiscal 2033, with Google entitled to buy a $12.2 billion stake.
The company already bought Celestial AI for $3.5 billion in February and Inphi for $10 billion in 2021. Lumentum would be its largest purchase since Inphi. The logic is vertical integration: combine Marvell's electrical design, switching and custom silicon with Lumentum's lasers, indium phosphide manufacturing and optical switching.
The target is Broadcom. Reuters describes Broadcom as Marvell's larger rival and Google's long-standing custom-chip partner, with a broader menu of AI data-center hardware. BNP Paribas analyst Karl Ackerman argued that Broadcom is the only supplier combining compute and switch silicon with an in-house laser portfolio, according to Benzinga. Buying Lumentum would give Marvell a similar combination.
The combined portfolio would also span more of the network. Marvell's electrical interconnect, switching and custom XPU (accelerator) attach cover the chip-to-chip and rack-to-rack links. Lumentum's lasers, transceivers and optical circuit switches cover the longer optical spans. Marvell says that reach runs from connections inside an advanced semiconductor package to links between distributed data centers. Marvell argues that customers increasingly want a validated system rather than individual components, which is the strategic case for paying a premium.
Nvidia is entangled with both sides. It invested $2 billion in Marvell in March and $2 billion in Lumentum earlier this year, Reuters reported. CNBC said the Marvell investment came with a collaboration on silicon photonics and AI telecommunications infrastructure. BNP Paribas described the investments as a way to secure optical-component supply, Benzinga reported. A combined Marvell-Lumentum would also be a more powerful supplier to Nvidia's optical roadmap.
Marvell estimates the combined addressable market at $550 billion, with more than $1 billion in annual cost synergies and accretive non-GAAP earnings per share in the first year after closing. Those are management targets. They depend on closing, integration and continued demand.
What are the risks of a $115 billion all-stock deal?
Dilution is the first risk. Lumentum holders would own about a third of the company, and Marvell had only $3.9 billion of cash, MarketWatch reported. An all-stock deal avoids new debt but issues a large number of shares. Stifel analyst Tore Svanberg told MarketWatch that the deal would be dilutive near term, though cost savings could pay off later.
The fixed ratio cuts both ways. If Marvell's stock falls, Lumentum holders receive less value; if it rises, they receive more. The recent 22% drop already reduced the implied value to roughly $884 a share, below the $914 unaffected close. That could make the deal harder to sell to Lumentum holders, even though their board approved it.
Lumentum's Friday close of $840.70 was about 5% below the $884 implied value. That discount is a rough measure of deal risk: investors are pricing in the long regulatory review, the chance that Marvell's stock moves again and the possibility that the deal does not close. A narrower discount would suggest growing confidence; a wider one would suggest doubt.
There is also a customer-conflict problem. Lumentum is a key customer for Marvell's DSPs, and RBC analysts said the combination could prompt other customers to seek alternative suppliers. Marvell's DSP market share is between 60% and 65%, according to RBC. Owning a major customer could make rivals reluctant to depend on Marvell.
Integration is hard. Lumentum brings compound semiconductor fabs and indium phosphide processes, while Marvell brings CMOS design, switching and custom silicon. The combined company is supposed to add advanced packaging and deeper co-design across the electrical and optical signal chain. Turning that into co-packaged optical engines is a multiyear engineering project, and the companies must retain laser and photonics engineers in a tight market.
The deal is also a leveraged bet on AI capital spending. Marvell shares fell 22% in the days before the announcement, partly because Intel's weak forecast reignited doubts about AI demand. If hyperscalers slow orders, the premium and the synergies look worse. Evercore ISI analysts also questioned why Lumentum would sell when it has said its business is sold out through fiscal 2029, MarketWatch reported.
Will regulators approve the Marvell-Lumentum deal?
The companies need approval from Marvell and Lumentum shareholders, plus regulators in the United States and other jurisdictions. MarketWatch reported that the deal would be subject to Chinese and US reviews and that China's process can be lengthy. The expected close in the first half of 2028 suggests the companies are preparing for a long review.
China is the main uncertainty. Lumentum competes with Coherent and China's Zhongji Innolight, and Reuters noted that Lumentum could benefit from a proposed US ban on new Chinese optical transceivers. Such a ban could shift demand toward Lumentum, but it could also make Beijing more cautious about approving a deal that consolidates a key optical supplier.
An 18-month timeline is long for a US semiconductor acquisition, and it leaves room for the terms to change. The companies have not publicly detailed which regulators must clear the deal beyond the US and Chinese reviews. They also have not disclosed whether they would accept conditions or divestitures. Those details usually appear in the merger proxy and the Form S-4, which Marvell said it intends to file.
No regulator has announced a decision, and the companies have not disclosed conditions or remedies. The Form S-4 registration statement will include risk factors and a joint proxy statement. Until that filing, the market has little detail on the timing or the antitrust theory.
Approval is not assured. The deal would put a major laser and transceiver supplier inside a chip designer, a combination regulators are likely to examine closely. Investors should treat the 2028 close as a plan, not a certainty.
What should investors and customers watch next?
The first milestone is Lumentum's fiscal second-quarter results, which will test its guidance for $1.575 billion to $1.625 billion in revenue. Watch the NPO/CPO purchase commitments and optical circuit switching revenue, because those are the businesses Marvell is paying for.
Marvell's full fiscal third-quarter results will matter too. The company said preliminary revenue was consistent with guidance and narrowed its range to plus or minus 2%. Investors will look for custom AI chip orders and any signs of customer concentration.
The Form S-4 and shareholder votes will set the timetable. The implied deal value will move with Marvell's stock every day, so Lumentum's share price relative to 4.5 times Marvell's price is a live measure of deal risk.
Watch the direction of that gap, not just its size. A widening gap between Lumentum's price and the implied offer would signal that investors doubt the deal can close on these terms. A narrowing gap would signal that they expect the transaction to complete and are pricing the time value of waiting.
Customers will watch whether Marvell keeps Lumentum's optical components available to rivals. If Marvell restricts supply or bundles them with its DSPs, hyperscalers may accelerate work with Broadcom, Coherent or Chinese suppliers. If it keeps the business open, the combination could become a reference design for co-packaged optics.
The wider signal is AI infrastructure spending. Lumentum's 136.5% growth is evidence of demand, but the market's reaction to Intel's forecast shows how quickly sentiment can turn. The Marvell-Lumentum deal is a bet that optics will be the scarce layer of AI infrastructure. The next two years will show whether that bet is early, right or too expensive.