Marvell Technology fell 7.7% in after-hours trading on Wednesday, to $232.39 from a $251.65 close, after Bloomberg reported the chipmaker is in advanced talks to buy Lumentum Holdings. Lumentum went the other way, up about 10% to roughly $1,008 from $913.95. Neither company has confirmed anything. No price, premium or payment mix has been reported. An announcement could come as soon as next week.
The two moves did not cancel out. Marvell shed about $17 billion of market value in one evening while Lumentum added around $9 billion on a diluted count, so the pair lost roughly $8 billion of combined value on a deal that does not yet exist. Markets usually wait for terms before judging a merger. This one judged the funding first.
The Cash Is Not There
At Wednesday’s close and a diluted share count of about 101 million, Lumentum is worth roughly $92 billion before any premium. A conventional 20% to 30% premium puts the price between $111 billion and $120 billion. Marvell ended its last quarter with $3.93 billion of cash and $4.96 billion of debt.
On October 22, six days before the report, Marvell swapped its $1.5 billion revolver for a $3 billion facility that lets its leverage ceiling rise from 4.0 to 4.5 times in the quarter it closes an acquisition worth at least $500 million, and for three quarters after. That’s the housekeeping a company does before a large deal. It isn’t deal money. Even at 4.5 times the combined operating earnings of both companies, new borrowing covers perhaps $25 billion. The rest is Marvell stock.
Run it at a 25% premium with $25 billion of debt and Marvell issues about $90 billion of shares at the after-hours price. That’s close to 390 million new shares on top of about 921 million diluted shares outstanding, leaving Marvell holders with around 70% of the combined company. Push the premium to 30% and borrow less, and they slide toward 65%.
Wednesday’s after-hours loss is roughly what a 20% premium would hand to Lumentum holders with zero credit for synergies. Marvell’s shareholders priced the dilution before they priced the strategy.
Why Marvell Wants a Laser Company
The strategy deserves more credit than the tape gave it. Marvell’s franchise in AI networking is the signal-processing chip inside pluggable optical modules, the business it bought with Inphi in 2021. That DSP sits between the switch and the light. Co-packaged optics, which Nvidia and Broadcom are pushing into switches from 2027, moves the optical engine into the switch package and removes the DSP from that link. The laser stays. It moves outside the package as an external light source, because lasers run hot and fail more often than silicon, but every optical link still needs one.
Marvell has spent this year arguing that its attach business survives a hyperscaler’s decision to design its own accelerator. That holds. What the attach business doesn’t obviously survive is co-packaged optics. Lumentum is the hedge. Marvell would own the DSP for pluggables, Celestial AI’s photonic fabric for scale-up, the lasers that feed both, and Lumentum’s optical circuit switches on top. Broadcom already sells lasers and co-packaged switches together. Marvell can’t offer that today.
Lumentum is no low-margin component shop. Its fiscal fourth quarter brought $1.006 billion of revenue, up 109%, with $649 million from components, mostly lasers. Non-GAAP operating margin was 36.6%, identical to Marvell’s latest quarter. Guidance for the September quarter is $1.225 billion to $1.275 billion at a 39.5% to 40.5% operating margin. Management has described its pump lasers as effectively sold out. Nvidia paid $2 billion in March for a Lumentum stake that came with capacity access rights to advanced lasers, and signed a similar $2 billion deal with Coherent the same day. When the biggest buyer in the market pays for the right to buy, the asset is scarce.
What Marvell Would Be Taking On
The moat Marvell is buying is real for as long as indium phosphide laser capacity stays short. It’s a different kind of moat from the one Marvell already owns. Marvell’s is design, run fabless on capex of about 4.3% of revenue. Lumentum runs its own fabs, spent roughly 15% of fiscal 2026 revenue on capex and is building a new plant in Greensboro, North Carolina. A combined company would earn about the same operating margin with far more equipment on the balance sheet. Lumentum’s two largest customers took 26.6% and 15.0% of revenue last year, and 79% of shipments went outside the US.
There’s a channel problem too. Marvell sells DSPs to module makers. Lumentum’s Cloud Light unit builds 800G and 1.6T modules and competes with those same customers. Some of Marvell’s buyers would wake up to find their supplier selling against them.
Then there’s the clock. China’s market regulator cleared Marvell’s Cavium and Inphi deals and Lumentum’s Oclaro purchase, but each review took five to nine months. Chinese module assemblers are among the largest buyers of both optical DSPs and lasers. A review of this size is unlikely to be faster.
The Number That Settles It
Marvell had nearly tripled this year before Wednesday. It closed at $287.01 on October 6, the day its investor day lifted the fiscal 2028 revenue target to about $20 billion and set a fiscal 2031 range of $70 billion to $90 billion, and it had already given back about 12% before the report. Both companies trade at similar multiples of guided earnings, so an all-stock deal at no premium is roughly neutral to Marvell’s earnings per share. Every point of premium is dilution that only Lumentum’s faster growth can repay. And only while lasers stay short past 2028, when the capacity now being built comes online.
This is the consolidation road that took Broadcom to a trillion-dollar valuation, and it runs on shares. A bigger market cap reached by issuing stock is worth nothing to the holders who paid for it. The test is whether Marvell shares, issued at well over 50 times guided earnings, buy laser earnings that grow faster than that.
Lumentum reports on November 5 and may say something about the talks. The figure that matters more is the exchange ratio, and how much of the price arrives as cash.
