A $115 billion transaction has just placed cables, lasers, and optical components at the center of the global race for artificial intelligence. Marvell Technology announced on November 2 a definitive agreement to acquire Lumentum in a deal entirely paid for with stock. The figure immediately attracts attention, but the industrial stakes are even more significant: next-generation data centers need to handle ever-increasing volumes of data, faster, and with better energy efficiency.
According to documents filed with the U.S. Securities and Exchange Commission, the transaction assigns Lumentum an implied enterprise value of $115 billion, or approximately $1,132 per share. Lumentum shareholders would receive 4.5 Marvell shares for each share they hold. They would own about 33% of the new entity, compared to 67% for current Marvell shareholders. The deal remains subject to shareholder votes and regulatory approvals, with a target completion in the first half of 2028.
Why This Announcement Could Change the AI Battle
Artificial intelligence does not solely rely on the most powerful processors. It also depends on the connections that allow thousands of chips to work together. As models grow larger, the movement of data becomes as strategic a problem as its computation. Traditional copper links face limitations in distance, bandwidth, and energy efficiency. Optics, which transport information using light, thus becomes an essential piece of the architecture.
Marvell designs semiconductors and connectivity solutions for digital infrastructures. Lumentum manufactures, among other things, photonic components, lasers, and optical technologies used in networks and data centers. Their merger aims to integrate electronics and photonics more closely. In other words, the group wants to control a larger portion of the data journey, from the chips that process it to the light links that transport it.
A Spectacular Amount, But a Still Fragile Agreement
The announced price represents a premium of about 23.9% over the unaffected share prices on October 28, before speculation about a deal spread. A high premium may convince the target’s shareholders, but it also increases pressure on the buyer. Marvell will need to demonstrate that future revenues, promised savings, and market growth justify such a large-scale operation.
The companies project more than $1 billion in annual synergies over time and estimate that their combined potential market could reach $550 billion. These projections remain goals, not guaranteed results. The integration of two tech groups requires aligning teams, roadmaps, customers, and supply chains. It can also lead to functional redundancies, although no detailed restructuring plan should be assumed before official announcements.
The stock market reaction reflects this caution. Marvell’s stock fell after the announcement, in a session already challenging for several tech stocks, according to the Associated Press. A decline in the acquirer’s stock does not mean the project will fail. However, it shows that investors are assessing the risk of dilution, the price paid, and the unusual duration separating the announcement from the anticipated closure.
Broadcom, Nvidia, and the New Map of Tech Power
Reuters presents the project as an attempt to strengthen Marvell against Broadcom, a major player in custom chips and networks for AI. The comparison is revealing. Competition is no longer limited to selling an isolated component: major suppliers are seeking to offer coherent sets capable of connecting accelerators, switches, memory, and optical networks in gigantic infrastructures.
Nvidia remains the dominant reference for AI accelerators, but the market’s expansion opens new fronts. Cloud giants are developing their own chips, operators are looking to reduce their dependence on a single supplier, and component manufacturers want to occupy essential junction points. In this landscape, optical connectivity becomes a form of industrial power: even the best chip loses value if data does not arrive quickly enough.
What Europe and the Rest of the World Should Watch
The agreement is American, but its consequences will be global. Data centers are multiplying in Europe, the Middle East, and Asia, while governments seek to attract investments in cloud and AI. The cost of energy, the availability of electrical networks, and access to advanced components are becoming decisive criteria. Improved efficiency of optical interconnections could support this expansion, but it does not alone resolve the pressure data centers exert on electricity, water, and land.
For France and the European Union, this mega-operation also highlights the persistent concentration of strategic technologies outside the continent. Europe has champions in semiconductor equipment, photonics, telecommunications, and research, but still struggles to develop groups of comparable size across the entire AI chain. European authorities may examine the effects of the merger on customers, innovation, and access to components, depending on the jurisdictions involved.
The Signal That No One Can Ignore
The message from Marvell is clear: the next phase of AI will be played out as much in connections as in processors. After years dominated by computing power, the market is discovering that bandwidth, latency, and the energy consumed to move data can determine the actual performance of a system. Lumentum brings precisely expertise in this often-invisible physical layer to the public.
The size of the operation also makes it a test for the financial enthusiasm surrounding AI. If demand continues to grow at the expected pace, the new entity could become an essential supplier. If investments slow down, if regulators impose concessions, or if integration disappoints, the price paid will seem much heavier. Between these two scenarios lies the most likely reality: a long industrial battle, monitored by customers, competitors, and markets.
The acquisition is therefore not yet complete. It will require votes, approvals, and nearly eighteen months of preparation if the announced timeline is respected. But the decision is already a marker. By betting $115 billion on the light that connects machines, Marvell asserts that the hidden infrastructure of AI is now worth as much as some of the most visible companies in global tech.
