The AI boom is easy to picture as a race to build faster chips and smarter software. But behind every dazzling demo is a less visible challenge: moving enormous amounts of data quickly, reliably, and without wasting too much power.
That’s why Marvell Technology’s agreement to acquire Lumentum matters. Announced November 2, the all-stock deal values Lumentum at approximately $115 billion in enterprise value, with an implied equity value of about $102.5 billion. It’s a headline-grabbing bet on the physical infrastructure that makes AI possible, and a reminder for entrepreneurs and creators that the most valuable part of a boom may be the essential layer nobody sees.
What Marvell is buying
Under the agreement, Lumentum shareholders will receive 4.5 shares of Marvell common stock for each Lumentum share. If the deal closes, Marvell shareholders are expected to own about 67% of the combined company, with Lumentum shareholders owning about 33%. Closing is expected in the first half of 2028, subject to shareholder and regulatory approvals and other customary conditions.
The all-stock structure means the deal’s implied value can move as Marvell’s share price changes. It’s not a cash payment at a fixed dollar amount for each Lumentum share. And while the companies have announced an agreement, the acquisition has not closed.
The strategic logic is about bringing two sides of connectivity together. Marvell works across semiconductors and electrical and optical interconnect. Lumentum makes photonics technology, including high-performance lasers and optical transceivers. The components that help data travel as light through fiber.
In simple terms, one company brings important tools for processing and directing data; the other brings expertise in turning signals into light and transmitting them. Marvell says the combination is intended to create a broad connectivity portfolio and support more integrated, optimized solutions for AI infrastructure.
The companies have already worked together on optical technology. In a 2025 demonstration, they paired Marvell’s 400G-per-lane technology with Lumentum’s indium phosphide optical transmitter. That work pointed toward a future where faster optical connections help data centers handle growing AI workloads.
AI doesn’t run on chips alone
Think about what happens when an AI model is trained or used at scale. Vast numbers of calculations take place across connected machines. Those machines need to exchange data, often constantly and at high speed. The more complex the system becomes, the more important it is to move information efficiently between processors, servers, and other parts of the infrastructure.
That’s where the “pipes” come in. Copper connections still have a role, but the companies argue that growing bandwidth needs and the physical limits of copper are accelerating a shift toward optical technologies. Light traveling through fiber can help move data across distances inside and between data centers.
It’s not just a technical detail. Connectivity can shape how much computing capacity an organization can use, how efficiently it can use it, and how easily it can expand. A powerful processor is less useful if the larger system can’t get data to it fast enough.
This helps explain the scale of the deal. Marvell and Lumentum describe the combination as a way to serve more of the connectivity needs emerging across AI infrastructure, from data-center links to connections closer to the chips themselves. Their goal is deeper coordination between electrical and optical technologies, rather than treating every component as a separate piece.
The announcement also signals a more ambitious position for Marvell in the fast-growing AI infrastructure market, where it is challenging established players such as Broadcom. The contest isn’t only over who can make the fastest chip. It’s also about who can provide the network of technologies that lets the whole system work together.
The biggest opportunities can hide in the supporting layer
For small business owners, entrepreneurs, and independent creators, a $115 billion transaction may feel far removed from everyday decisions. The useful lesson is closer than it appears: when a market changes quickly, the companies that quietly solve a critical problem can become indispensable.
The attention-grabbing layer of AI is often the interface, model, or product people use. But durable businesses also grow around what those products depend on: infrastructure, specialized services, tools, distribution, compliance, training, and support.
The same pattern shows up outside technology. A creator economy may depend on reliable editing workflows, booking tools, rights management, and audience analytics. A small retailer’s growth may rest on logistics and inventory systems. A new service business may win by solving the awkward, unglamorous handoff between a customer’s need and a finished result.
The takeaway isn’t “build a data-center company.” It’s to look closely at the layer your customers depend on, and ask whether it’s becoming more important as the market evolves.
Four questions to find your essential layer
1. What do customers need before they can get your main value?
A customer may come for a finished product, but their biggest frustration could be onboarding, scheduling, setup, or follow-through. Fixing that bottleneck can matter more than adding another headline feature.
2. What part of your business becomes more valuable as demand grows?
Some offerings get harder to deliver at scale. Others become more useful as more customers join. Identify the part that gets stronger with growth, and invest in making it dependable.
3. Which capability would be costly or difficult for someone else to replace?
A repeatable process, trusted expertise, a specialized supplier relationship, or a unique understanding of your audience can give a business resilience. Make that strength clear, and build it with care.
4. Are you watching the change underneath the trend?
A rising wave can create opportunities, but the deeper shift is often in customer behavior or infrastructure. Ask what has changed, what the change now requires, and which needs will remain after the initial excitement fades.
Build where the need is real
Marvell’s proposed acquisition is not a guarantee that every forecast will come true. The transaction still needs to clear its required approvals, and its anticipated benefits depend on execution. But the strategic wager is unmistakable: as AI systems expand, moving information becomes as important as processing it.
That’s a powerful reminder for anyone building something. Don’t only chase the visible trend. Pay attention to the unglamorous work that makes the trend possible. Ask where customers get stuck, what they rely on without thinking, and which capability becomes more essential as the market grows.
The next breakout opportunity may not be the loudest idea in the room. It may be the quiet, reliable piece that lets everything else move faster.
