Marvell's $115 Billion Lumentum Deal: Is AI's Next Bottleneck Light, Not Chips?

Marvell's purchase of Lumentum points to a contest beyond GPU speed, in data transfer and power efficiency. Whether the price is right is not yet confirmed.

The 20-second read

1
Marvell has announced it will acquire optical networking company Lumentum. Nvidia invested $2 billion in each of the two companies this year. It is a sign that data connectivity technology is becoming more strategically valuable in the AI industry.
2
In AI data centers that link thousands of GPUs, overall performance depends not only on the computing power of the chips but also on communication bandwidth, latency, and power efficiency.
3
Lumentum's purchase price is about 33 times expected core operating profit over the next 12 months. That is double the peer median of about 16 times. The growth of the optical networking market and the economics of the acquisition need to be judged separately.

Does connecting more GPUs make AI proportionally faster?

Diagram: GPU (compute) <-> network (data movement) <-> GPU (compute). Overall AI system performance depends on the balance between computing and data movement. Optical links do not replace every copper connection.

Think of an AI data center as a giant factory.

GPUs are the production equipment that does the computing. But when thousands of GPUs train a single AI model, they must exchange data constantly. However fast the machines are, output does not rise if the aisles that carry parts are jammed.

The same holds in a data center. If a GPU that has finished its calculation waits longer for the next batch of data, expensive chips go underused.

That is why AI infrastructure competition is widening from raising GPU performance to how fast and efficiently GPUs are linked together.

Copper-based connections have advantages in price and power efficiency over short distances. As speeds and distances grow, though, signal loss and power problems get harder to manage.

Optical communication converts electrical signals to light to carry large volumes of data. It is an important technology for expanding communication capacity in large GPU clusters.

What is Marvell buying for $115 billion?

  • Enterprise value including debt: $115 billion
  • Acquisition premium: 23.9%
  • Lumentum holders' stake in the combined company: about 33%
  • Expected closing: first half of 2028

On Nov. 2, Marvell and Lumentum announced an all-stock acquisition agreement.

Lumentum shareholders will receive 4.5 Marvell shares for each Lumentum share they hold. The $115 billion figure is Lumentum's enterprise value including debt, not the sum of the two companies' values. Its equity value is about $102.5 billion.

Marvell has custom AI chips, networking chips, and electrical and optical interconnect technology. Lumentum is strong in lasers, optical transceivers, optical circuit switching and specialty semiconductor manufacturing.

The logic for combining is clear: to design and supply together the chips that process data and the components that carry it as light.

Marvell bought optical connectivity company Celestial AI in February. The Lumentum deal looks less like entry into a separate business than a larger extension of its existing optical strategy.

Why did Nvidia invest in both companies?

Nvidia invested $2 billion each in Marvell and Lumentum in 2026, $4 billion in total.

That suggests building good GPUs alone is not enough to complete a competitive AI system.

For Nvidia, the more GPUs customers install, the more the networking and optical technology that makes those GPUs work properly matters.

If the combination improves the performance and supply of optical components, Nvidia's AI system ecosystem could benefit. That does not mean Nvidia led the merger or has an exclusive supply relationship.

Marvell, meanwhile, faces a larger rival in Broadcom, which holds a strong position in custom AI chips and networking.

What Marvell is after is not one more optical part. It is a platform that integrates everything from chip design to optical connectivity for customers.

Why did Wall Street not welcome the deal?

Chart: enterprise value to expected core operating profit. Lumentum deal price: 33x. Peer median: 16x. Source: Reuters / LSEG, based on forward 12-month estimates.

The problem is price.

By Reuters' calculation using LSEG data, Lumentum's acquisition valuation is about 33 times expected core operating profit over the next 12 months. The peer median is about 16 times.

Paying more than peers does not by itself mean failure. But for the deal to succeed, results must show not only Lumentum's growth but also cost savings after integration, stronger products and higher cash flow.

The companies expect more than $1 billion a year in cost synergies and a rise in non-GAAP EPS in the first year after closing. Both are management targets, not current results.

On Nov. 2, Marvell shares fell more than 12% and Lumentum fell more than 15%. More important than one day's move is the question the market raised.

Even if AI optical networking grows strongly, can profits grow by more than the price being paid now?

Will optical networking completely replace copper connections?
  • Not in every segment. Copper has cost and power-efficiency advantages over short distances.
  • How fast optical networking expands depends on connection distance, bandwidth requirements, and technology cost.
Doesn't this create a competitor for Nvidia?
  • Marvell is a partner of Nvidia and also a competitor in some areas.
  • Advances in optical networking can be an opportunity for both Nvidia GPUs and other companies' AI chips.
Can Marvell catch up with Broadcom?
  • The Lumentum acquisition helps Marvell expand its optical technology and product portfolio.
  • But it will need sustained results in winning customers, design competitiveness, production capacity, and profitability.
If the optical networking industry grows, should I buy Marvell stock?
  • Even a company in a growing industry can see lower investment returns from a high acquisition price and the share dilution caused by issuing stock.
  • It makes more sense to evaluate post-integration cash flow per share and valuation together, rather than the stock price alone.
What is the most important failure scenario for this deal?
  • Optical networking demand grows, but competition pushes down selling prices and profit margins.
  • Expected synergies are delayed, or deal approval and integration costs increase.

What to watch

Optical networking revenue growthWhether rising AI network investment translates into actual sales growth.
Gross margin and operating marginWhether the integrated high-value products raise profitability.
Cost synergies of over $1 billion a yearHow quickly the announced savings are realized, and what integration costs come with them.
Free cash flow and earnings per shareWhether revenue growth and improved non-GAAP EPS turn into actual cash and shareholder value.
Shareholder and regulatory approvalHow realistic the plan to close the deal in the first half of 2028 is.
AI competition is widening from GPU performance to the technology that connects GPUs, but whether Marvell's price for Lumentum is justified is not yet confirmed.

Insight Times Editorial Desk

More by topic

Starting with this story's topics, then what readers are following now

Investor psychology, decision-making and books on investing

Data Centers

This story's topic93 storiesSee all →

A building that houses servers and networking gear and supplies them with powe…

US AI Data Centers Are Headed for 90GW. Gigawatts Now Matter More Than Dollars

Goldman Sachs expects US data center capacity to grow from 64GW at the end of 2026 to 90GW at the end of 2027. The AI race is moving from buying GPUs to building power plants and transmission lines alongside them.

Tech

Why the White House Swapped "AI" for "Super Intelligence"

Executive Order 14434 renames AI across the federal government without defining a new technology. The real story is about roughly $700 billion in capex, free cash flow and electricity.

Tech

Nvidia's Next Target After GPUs: Controlling What AI Agents Are Allowed to Do

OpenShell and Sentry are not about policing what an AI says. They are about controlling, at the server infrastructure level, which files and APIs an AI can touch, what it can execute, and who can actually pull the plug when something goes wrong.

Tech

2027 Investing Needs Alignment, Not Prediction

A new investing book argues the hard part was never picking the next winner. It's staying aligned with where value and bottlenecks are moving while keeping old human weaknesses in check.

Investor Mind

AMD Crossed $1 Trillion, and the Market Started Looking at CPUs Again

AMD's stock topped a $1 trillion market cap for the first time. The rally isn't just about being an Nvidia alternative. It reflects a reassessment of AMD as a full-stack AI infrastructure vendor, one that bundles CPUs, GPUs and networking, as AI shifts from chatbots to agents and reignites data center demand for CPUs.</deck> <parameter name="oneline">As AI moves from chatbots to agents, the market has started counting the CPU sitting next to the GPU again.

Tech

Why Nebius Is Keeping Some 2027 AI Capacity Off the Shelf

Nebius mixes long-term contracts, short-term premiums and auctions. Holding back capacity pays only if the servers stay busy, and it is not yet clear the pricing will last.

Tech

DeepSeek's $15 Billion Raise: Model Makers May Feel It Before Nvidia Does

China's DeepSeek is closing a round about twice its original target. For US investors, the question is less the size of the check than where the money goes and whose profits it could erode.

Tech

NVDA

This story's topic87 storiesSee all →

Amazon Goes After Nvidia's Strongest Moat, and It Is Software, Not GPUs

Amazon's open-source TorchNeuron aims to let existing AI code run on its Trainium chips. The contest comes down to the total cost of AI compute and the return on Amazon's $220 billion in annual capital spending.

Tech

The US-China Summit Didn't Erase Taiwan Risk for Chip Stocks

Washington's Taiwan policy held, but roughly $14 billion in arms sales is still on hold. For AI chip investors, the summit reads less like risk removed and more like a fragile balance extended.

Macro

AI's Next Bottleneck Is Money, Not Chips

Broadcom is lending Anthropic up to $42 billion, and Nvidia wants to build a $500 billion market for loans backed by GPUs. The AI race may turn on what rate buyers pay to borrow.

Tech

ASML Is Discussing 2030 Orders. How Long Can the AI Chip Supercycle Last?

Despite bubble worries, ASML sees equipment shortages through 2028. Investors should watch future production capacity, not just today's orders.

Tech

Tesla Energy's 13.7GWh Paradox: Expectations Ran Faster Than Deployments

Q2 deployments recovered but margins collapsed. Q3 set a record for deployments and still missed estimates. The issue is not the size of AI power demand but how fast it converts into Tesla revenue and profit.

Tech

Why Nvidia Is Happy to Help Make AI Models Nearly Free

Falling model prices can squeeze model makers. For Nvidia, the number that matters is the total volume of tokens generated worldwide and the compute that produces them.

Tech

The Number That Matters More Than How Smart AI Gets

This week's AI models got stronger and cheaper at the same time. The chart long-term investors should watch isn't the weekly benchmark leader, it's the falling cost of finishing a task successfully.

Tech

GPU

This story's topic63 storiesSee all →

A chip built to draw graphics, whose ability to run thousands of identical cal…

The Next AI Chip Race Is About How Chips Connect

Applied Materials will lift its stake in Dutch equipment maker BESI from about 9% to 20%. The bet is less about control than about a new way to link chips.

Tech

Musk Borrowed $40 Billion for GPUs. He Also Warned AI Cloud Firms Could Go Bust

SpaceX's giant financing shows a new dividing line in AI: who borrows matters as much as how much is spent. Firms with other revenue may outlast those that rent out GPUs alone.

Tech

AI Data Centers' Next Bottleneck: Getting Power From the Grid to the GPU

Securing a gigawatt of power does not mean a gigawatt reaches GPU compute. The next fight in AI infrastructure is happening inside the data center, over how little electricity gets wasted before it reaches the chip.

Tech

AI Stops Just Talking, Starts Working: The Second Infrastructure Cycle

The center of the AI race is shifting from chatbot performance to agents that finish real work. For investors, what matters is not model rankings but how a single query turns into dozens of inference calls, pulling demand beyond GPUs into CPUs, HBM, networking and power.

Tech

The AI Bubble Question Has Changed: By Late 2026, It's ROIC Over GPUs

There is no sign yet that AI investment is slowing. If anything, hyperscaler spending forecasts keep climbing. But as the scale of that capital has grown, the market's question has shifted from "is AI growing" to "when, and how much, does this capital come back as cash."

Tech

Nvidia Put $2 Billion Into Lumentum. Optics May Be the Next Bottleneck, but the Stock Has Already Run

Optical networking is emerging as the next constraint on AI data centers. Investors need to separate the industry's growth potential from the price they are already paying for it.

Tech

If GPUs Are the Muscle, the Network Is the Nervous System

AI data centers do not get faster just by buying more GPUs. When thousands of accelerators cannot exchange data on time, the most expensive asset in the building sits idle. That is the whole investment case for Arista.

Investing

Valuation

This story's topic34 storiesSee all →

Measuring where a share price sits relative to the company’s results.

Anthropic's IPO: The Number That Matters More Than $100 Billion Is $2 Trillion

Claude-maker Anthropic is growing at a stunning pace, but the real question for investors is not how fast, only how long, and how much cash is left after GPUs and data centers.

Tech

Oil at $110, 10-Year Yield at 5%: The Discount Rate on Growth Stocks Is Rising Again

Brent crude spiked to $109.97 intraday and the US 10-year Treasury yield tested the 5% threshold. What matters is not the numbers themselves but that oil-driven inflation fears are pushing bond yields higher across the globe at once.

Macro

ARK Sets Tesla 2031 Target at $4,400, Betting on 8.1 Million Robotaxis and Optimus

ARK Invest's new model implies a $17.5 trillion market cap for Tesla by 2031. It reads less like a forecast than a scenario that assumes robotaxis and robots earn real money.

Tech

AI Needs to Find $1 Trillion a Year From Somewhere to Pay Back Its Capex

Saying AI capex is big is no longer the point. The harder question is what end users have to pay for the buildout to pencil out, and Goldman Sachs puts that number at roughly $1 trillion a year.

Tech

Year Five of the Bull Market: Why U.S. Stocks Aren't Expensive Yet, and What AI Chips and Power Must Deliver

U.S. stocks have gained about 117% since the October 2022 low. What happens next depends less on the bull market's age than on whether expected earnings, which lean heavily on AI semiconductors, turn into reality.

Investing

AI Can Lift More Than Stocks: It May Push Long-Term Yields Higher Too

The AI boom can grow the numerator of corporate earnings while also growing the denominator of the discount rate. Investors watching long-term Treasury yields now need to add "AI capital demand" to the Fed, inflation and the deficit.

Tech

Lam Research Revenue Jumped 54%. The Stock Fell 12%. Here Is Why

Record results and a sharp sell-off at Lam Research show a paradox of growth investing. The key question is not how fast a company grows, but how much investors paid for that growth.

Tech