Source: Wall Street Journal
As AI models move towards a massive "Agent" era, the computing power bottleneck in data centers is gradually shifting towards "connectivity," triggering a fundamental infrastructure revolution from copper cables to fiber optics.
On the second day of the Computex conference in Taipei, Matt Murphy, Chairman and CEO of Marvell, a leader in AI custom chips, optical communication, and data center interconnect, delivered a keynote speech. NVIDIA CEO Jensen Huang surprised attendees as a special guest, resulting in both leaders—at the pinnacle of AI computing power and network interconnect—publicly showcasing the deep strategic bond between their two companies. This joint appearance quickly became the highlight of the conference so far.
(Marvell CEO Matt Murphy and Jensen Huang discuss at the Computex conference)
After taking the stage, Jensen Huang set the tone for the event with a statement: "Ladies and gentlemen, the next trillion-dollar company"—he referred to Marvell. The audience erupted in applause. According to an article from Wall Street Journal, behind this is NVIDIA's announcement months ago of a strategic investment of $2 billion in Marvell, emphasizing their collaborative focus in the AI data center infrastructure space.
With the release of the previous quarter's financial report, the market is highly focused on Marvell's benefits during the AI supercomputing cycle. In response, Murphy presented a noteworthy report to the market: ten years ago, Marvell's data center business revenue accounted for less than 10%, while last quarter, this proportion exceeded 75% and is accelerating by approximately 40% annually. Based on the latest financial guidance, Wall Street generally expects its revenue to reach an astonishing $16.4 billion next year.
Behind this explosive growth, Huang and Murphy revealed the core investment theme for AI infrastructure during their discussion—once the bottlenecks of computing power and memory are broken through, "connectivity" will define the ultimate performance of the system. The two CEOs found common ground in their core consensus:
The next decisive battleground for AI infrastructure is not computing power, nor memory, but connectivity. Marvell is at the center of this revolution.
Notably, Marvell's stock price surged over 16% in after-hours trading.
The end of computing power is connectivity: AI enters the "useful stage," igniting the demand for infrastructure interconnect
Why has connectivity become so important today?
Murphy explained in his speech with a clear logic chain why "connectivity" has become the most critical constraint at present:
The bottlenecks in AI infrastructure occur sequentially and are broken through one by one—computing power (led by NVIDIA, becoming the first company in the world to reach a market value of $5 trillion) → memory (recently three new trillion-dollar companies have emerged in the memory sector) → connectivity (which is currently happening).
"The world's top cloud service providers are re-planning their overall network architecture, realizing that scaling AI infrastructure has become the primary connectivity challenge," Murphy stated, adding, "This is not just my personal view, but feedback from our largest clients."
Huang offered the most straightforward business logic during the discussion:
"Useful AI has arrived; it is now profitable, and tokens can also be profitable. When token production is profitable, everyone wants to produce more tokens—that's why the demand for Marvell is so strong and why our demand is so high."
Huang pointed out that the current AI is moving towards an "Agent" model, which requires breaking tasks down and distributing them across massive computing clusters. "When you decompose a computing problem into multiple parts and distribute them throughout the data center, what becomes most indispensable is connectivity." Huang generously praised his partners, even stating on stage: "Ladies and gentlemen, (Marvell) this is the next trillion-dollar company."
Murphy indicated that relying on a single processor is no longer sufficient to meet AI workloads; in the future, millions of processors need to work together.
"The expansion of computing scale is essentially a connectivity challenge. The entire industry has solved the computing power bottleneck, is currently addressing the memory bottleneck, and the next bottleneck limiting infrastructure to its extremes is connectivity."
"Use copper wherever you can; use optics wherever you must."
One of the most market-relevant segments of the dialogue between Murphy and Huang was their assessment of the timeline for transitioning from copper cables to optical fibers.
The strategic framework provided by Huang was straightforward: "Use copper wherever you can, and only use optics where you must."
He explained that copper cables have physical limits on bandwidth and transmission distance; before breaking this boundary, copper is the straightforward, low-cost, practical choice; only after crossing the critical point can fiber optics take over the scaling needs between racks, between data centers, and across data centers.
His core conclusion is:
"In the next 5 to 10 years, we will continue to use a lot of copper, while also using a vast amount of optical devices. These data centers have now become part of the infrastructure."
This judgment of "dual use of copper and optics, while each respects its boundaries" means for the market: whether in the copper or optical fiber domain, Marvell is in a position to continuously benefit—Marvell is one of the few companies that can provide complete solutions in both directions.
The timeline for switching from copper to optics is driven by inescapable physical laws. Murphy explains: The distance over which copper cables can transmit is inversely proportional to bandwidth; every time bandwidth doubles, transmission distance is halved. Currently, the fastest production systems have a single-channel speed of 200 Gbps, corresponding to a copper cable length of approximately 2.5 meters, while the height of a rack is about 2 meters—considering the internal wiring, 2.5 meters is already at the limit. "When we upgrade to 400 Gbps, copper will no longer fully connect an entire rack. The Copper Wall is moving, and it has already started." Every time the Copper Wall moves to the right, the number of connections increases at least by an order of magnitude, which will directly ignite the demand for optical communication.
To address these physical limits, Marvell is heavily investing in CPO (Co-Packaged Optics) technology, solving density and power consumption issues by directly integrating optical fibers into packages, adjacent to computing chips. On the day of the conference, Marvell officially released the new 100T Ethernet switch designed for AI data centers, featuring the industry’s lowest power consumption, and showcased a 51.2T switch based on CPO, completely eliminating copper routing at the board level.
“This is not a concept for the future; it is already being realized,” Murphy stated. Once optical interconnects completely break the distance limits, future data centers will no longer have rigid physical boundaries for computing and memory; the infrastructure will be able to dynamically combine on a large scale according to the needs of AI models.
NV Link Fusion builds a heterogeneous ecosystem: Marvell aims to be the "Switzerland" of the AI era
To meet extremely complex network architecture requirements, NVIDIA has previously made a strategic investment of $2 billion in Marvell, expanding cooperation into multiple dimensions including optical communication, silicon photonics, and NV Link Fusion.
The emergence of NV Link Fusion aims to address the customization pain points for cloud service providers (CSPs). Huang explained that while cloud vendors are designing their own custom chips (ASICs), they still want to integrate with NVIDIA's system architecture.
"You don’t have to buy everything from us, just buy part of it. By merging NVIDIA's technology platform with Marvell's solutions, we essentially construct a decoupled, distributed, and heterogeneous data center."
In such an ecosystem, Marvell has found its irreplaceable role. Murphy emphasized Marvell's neutral and critical position:
"We work deeply with compute companies and we work deeply with storage companies. In many ways, we are like the ‘Switzerland’ of the industry, cooperating with all enterprises."
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
