On 2 November 2026, chipmaker Marvell Technology announced it will buy optical-components maker Lumentum in an all-stock deal with an implied enterprise value of $115 billion — the largest semiconductor acquisition on record. Under the fixed exchange ratio, each Lumentum share converts into 4.500 Marvell shares, valuing Lumentum at about $1,132 per share, a 23.9% premium to the unaffected closing prices on 28 October. That is an enormous price for a company most people have never heard of. So why would Marvell hand over roughly a third of itself to buy a laser supplier? The short answer: the AI boom runs on light — and this deal is a bet that whoever owns the light owns the future of AI data centres.

The bottleneck nobody talks about: moving data, not processing it

AI models are trained on clusters of thousands of GPUs — read how GPUs actually work if you want the full picture. Those chips are astonishingly fast, but they constantly need to swap data with each other: model weights, training updates, results.

The problem is that electrical signals weaken and slow down over distance. Moving data a few centimetres across a circuit board with copper wires is fine. Moving it hundreds of metres between racks — or kilometres between buildings — with electricity is slow, hot, and power-hungry.

The fix is optical interconnects: convert the electrical signal into pulses of laser light, shoot them through glass fibre, and convert back to electricity at the other end. Light loses far less energy over distance, and a single fibre can carry far more data than a copper wire. This is the same principle as the data centres powering AI — the fibre runs, not the copper ones, that matter.

Who does what in the light business

The deal makes sense once you see how the optical-connectivity stack splits between the two companies:

Marvell Lumentum
What it makes Networking chips, digital signal processors (DSPs), custom processors Lasers, optical transceivers, optical switching, specialist manufacturing
Role in the stack The brains — designs chips that encode data for travel and decode it on arrival The eyes and voice — the lasers and optical parts that turn data into light and back
Why AI needs it Every AI cluster needs massive switching capacity Every optical link needs lasers — and Lumentum's fiscal Q1 2027 revenue was $1.26 billion, up 136.5% year on year

Put together, one company would own both ends of the optical link: the chips that drive the data and the lasers that carry it. Marvell calls the combined opportunity a $550 billion total addressable market by calendar 2030.

The deal's terms, in plain English

  • All stock, no cash. Lumentum shareholders get Marvell shares, not money. That means the headline "$115 billion" is an implied enterprise value based on the October 28 share prices — it can move with Marvell's stock. It is not $115 billion in cash changing hands.
  • Fixed ratio. 4.500 Marvell shares per Lumentum share, no matter what. Current Marvell shareholders will own about 67% of the combined company; Lumentum holders about 33%.
  • Close expected in the first half of 2028, pending shareholder votes and regulatory approvals — and for a deal this size, regulators in multiple countries will take a long look.
  • Leadership stays. Marvell chairman and CEO Matt Murphy continues in both roles; three Lumentum directors, including CEO Michael Hurlston, join the board.
  • The payoff target: more than $1 billion in annual run-rate cost savings, mostly in cost of goods sold, with the deal expected to add to Marvell's non-GAAP earnings per share in the first year after closing.

Why investors flinched anyway

Despite the logic, both stocks fell on the news: Marvell dropped about 13%, and Lumentum sank nearly 19% to close at $684.40 — well below the implied $1,132 per-share value.

Investor nerves usually come down to three things with mega-deals:

  • Integration risk. Stitching two large companies together — different cultures, different product lines, different manufacturing — is famously hard. The synergies are projections, not guarantees.
  • Paying up at the top. The deal values Lumentum at a premium based on AI-infrastructure hype. If AI spending slows, $115 billion could look expensive in hindsight.
  • The waiting game. The deal won't close until the first half of 2028. A lot can change in a year and a half of AI — including the price of light itself.

What this tells you about where AI is heading

The biggest lesson is not about these two companies. It is about where the AI industry's real constraints now sit. The headlines in 2026 have been all about models and GPUs, but the infrastructure story is about everything around the chips: power, memory, and — above all — the plumbing that moves data between chips.

Nvidia's dominance taught the industry that the prize goes to whoever owns the scarcest input. For a while that was compute. Then it was memory (high-bandwidth memory is now a strategic battleground). Marvell is betting the next scarce input is bandwidth — the ability to move unimaginable volumes of data, cheaply, between ever-bigger AI clusters.

Takeaway checklist

  • ✅ Marvell agreed on 2 Nov 2026 to acquire Lumentum in an all-stock deal valued at $115 billion — the largest semiconductor acquisition on record.
  • ✅ Each Lumentum share converts to a fixed 4.500 Marvell shares (~$1,132/share, 23.9% premium); ownership splits 67%/33%.
  • ✅ Closing is targeted for H1 2028, subject to shareholder and regulatory approvals.
  • ✅ The logic: AI data centres move data as light, and the combined company would own both the chips (Marvell) and the lasers (Lumentum) in that chain.
  • ✅ Investors sold both stocks on the news — mega-deals carry integration and timing risk.
  • ✅ The bigger signal: in 2026, the AI race is increasingly about infrastructure bottlenecks — bandwidth, memory, and power — not just better models.

Hamza

Founder & Editor, TechToday

Hamza is the founder and editor of TechToday. Hamza writes about artificial intelligence, computing, security and the technology shaping everyday life, with a focus on explaining complex ideas in plain, honest language — clear answers for curious readers, without the hype.