As artificial intelligence models evolve toward the era of massive intelligent agents, the primary constraint within data center infrastructure is shifting decisively from raw compute power to network connectivity. This transition is triggering a fundamental revolution in underlying hardware, moving the industry standard from copper cabling to optical fibers. During the second day of Computex Taipei, Matt Murphy, Chairman and CEO of Marvell, a dominant force in AI custom chips and optical interconnection, delivered a keynote address that underscored this paradigm shift. NVIDIA CEO Jensen Huang made a surprise appearance as a special guest, standing alongside Murphy to highlight the deep strategic partnership between the two technology giants. Upon taking the stage, Huang set the tone for the event with a singular declaration referencing Marvell, prompting immediate applause from the audience. This public endorsement follows NVIDIA's announcement months prior of a $2 billion strategic investment in Marvell, cementing their integration within the AI data center infrastructure landscape.

With the release of the previous quarter's financial report, market attention has intensified on Marvell's performance within the AI supercomputing cycle. In response, Murphy provided a stark illustration of this transformation: a decade ago, Marvell's data center business revenue accounted for less than 10% of total sales, but in the most recent quarter, this ratio exceeded 75% and is accelerating at an annual pace of approximately 40%. Data compiled by Woofun AI shows that behind this surge in performance, the core investment theme identified by both CEOs is the emergence of connectivity as the defining factor for system performance once compute and memory bottlenecks are resolved. The consensus between the two leaders is clear: Marvell is positioned at the heart of this infrastructure revolution, a sentiment reflected in the company's stock price which surged more than 16% in after-hours trading following the dialogue.

Murphy articulated the logical chain driving the criticality of connectivity, noting that AI infrastructure bottlenecks have appeared and been overcome sequentially. The first phase was compute power, led by NVIDIA becoming the world's first company with a market value exceeding $3 trillion, followed by the memory sector where high-bandwidth memory solutions recently emerged. The current phase is connectivity, where top hyperscale cloud service providers are redesigning their overall network architecture. Murphy stated that the expansion of AI infrastructure has become the primary connectivity challenge, a conclusion drawn not from personal opinion but from direct feedback from their largest customers. Woofun AI notes that Huang Renxun provided the most straightforward business logic during the discussion, asserting that useful AI has arrived and can now be profitable, including through token production. As token production becomes profitable, the demand to produce more tokens drives the high demand for Marvell's connectivity solutions.

Huang Renxun emphasized that AI is moving toward an 'Agent' model, requiring tasks to be broken down and deployed in a distributed fashion across massive computing clusters. When a computing problem is fragmented and distributed across an entire data center, connectivity becomes the most crucial element. Murphy reinforced this by stating that relying on a single processor is no longer sufficient for AI workloads; instead, millions of processors must work together. The scale of computation is fundamentally a connectivity challenge, with the industry having already addressed the compute bottleneck and currently tackling the memory bottleneck, leaving connectivity as the next limitation to push infrastructure to its limits. Woofun AI analysis suggests that the transition from copper to optical fibers is the immediate technical imperative driving this infrastructure evolution.

A key segment of the conversation focused on the timeline for transitioning from copper cables to optical fibers. Huang Renxun outlined a strategy framework where copper cables, while simple and low-cost, face physical limits regarding bandwidth and transmission distance. Once these boundaries are crossed, optical fibers take over to meet expansion needs between racks, between data centers, and across data centers. Murphy explained that the transmission distance of copper cables is inversely proportional to bandwidth, with distance halving for every doubling of bandwidth. The current fastest commercially available system features a single-channel speed of 224G, corresponding to a copper cable length of about 2 meters, while rack height is approximately 2 meters. Considering internal wiring, 2.5 meters represents the absolute limit. Each time the 'Copper Wall' moves to the right, the number of connections increases by at least an order of magnitude, directly triggering an explosion in optical communication demand.

To address these physical limits, Marvell is doubling down on Co-Packaged Optics (CPO) technology, which solves density and power consumption challenges by directly integrating optical fibers into the package adjacent to the compute chip. On the day of the conference, Marvell officially launched a new 100T Ethernet switch designed for AI data centers with industry-leading low power consumption and showcased a 51.2T switch based on CPO, completely eliminating copper trace routing at the board level. Murphy stated that this is not a future concept but is already being implemented. Once optical interconnectivity breaks the distance barrier, future data centers will no longer have rigid physical boundaries for compute and memory, allowing infrastructure to dynamically scale in response to AI model requirements. To meet these complex network architecture needs, NVIDIA's $2 billion strategic investment facilitates collaboration across optical communication, silicon photonics, and NV Link Fusion.

The emergence of NV Link Fusion aims to address customization pain points for Cloud Service Providers (CSPs). Jensen Huang explained that while cloud providers design their own custom chips, they still seek to integrate NVIDIA's system architecture. By integrating NVIDIA's technology platform with Marvell's technology solutions, a decoupled, distributed, and heterogeneous data center can be built without requiring customers to purchase everything from a single vendor. In this ecosystem, Marvell has found an irreplaceable position. Murphy emphasized Marvell's neutral and critical role, stating that they collaborate deeply with compute and storage companies, acting like the industry's 'Switzerland' by maintaining partnerships with all major players regardless of their specific compute or storage stack.