Original author: Dong Jing, Wall Street Watch
As AI models move into the vast "Agent" era, the computational power bottleneck of data centers is gradually shifting towards "connectivity," igniting a fundamental infrastructure revolution that is transitioning from copper cables to optical fiber.
On the second day of the Computex conference in Taipei, Taiwan, Matt Murphy, Chairman and CEO of Marvell, a leader in AI custom chips, optical communication, and data center interconnection, delivered a keynote speech.
NVIDIA CEO Jensen Huang unexpectedly made a guest appearance, and the two leaders, standing at the forefront of AI computing power and network connectivity, brought the deep strategic relationship between their companies to the forefront. This joint appearance quickly became the most brilliant moment of the conference so far.
(Marvell CEO Matt Murphy and Jensen Huang discussing at the Computex conference)
After standing firm, Huang set the tone for the audience with a single statement: "Ladies and gentlemen, the next trillion dollar company." — he was referring to Marvell.
Thunderous applause erupted. According to an article from Wall Street Watch, behind this is NVIDIA's announcement of a strategic investment of 2 billion dollars in Marvell months ago, marking the latest footnote of cooperation between the two companies in the field of AI data center infrastructure.
Accompanying the release of the previous quarter's earnings report, the market is highly focused on Marvell's level of benefit from the AI supercomputing cycle.
In response, Murphy presented an eye-catching answer: ten years ago, Marvell's revenue from data center business accounted for less than 10%, while last quarter, this proportion had exceeded 75% and is accelerating at an annual rate of about 40%.
Based on the latest earnings report guidance, Wall Street generally expects its revenue to reach an astonishing 16.4 billion dollars next year.
Behind this performance surge, Huang and Murphy revealed the core investment line in 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 core consensus between the two CEOs is:
The next decisive battlefield for AI infrastructure is not computing power, not memory, but connectivity. Marvell is at the core of this revolution.
It is noteworthy that 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 interconnection
Why has connectivity become so important today?
In his speech, Murphy used a clear logical chain to explain why "connectivity" has become the most crucial constraint at present:
The bottlenecks in AI infrastructure appear sequentially and are broken through in order — computing power (led by NVIDIA, becoming the world's first company to reach a market value of 5 trillion dollars) → memory (recently three new trillion dollar companies have emerged in the memory field) → connectivity (which is happening).
"The world's top hyperscale cloud service providers are redesigning their entire network architecture, realizing that the expansion of AI infrastructure has become a primary connectivity challenge," Murphy said, "This is not my personal opinion; it is feedback we received from our biggest clients."
Jensen Huang provided 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 strong."
Huang pointed out that current AI is moving towards the "Agent" model, which requires tasks to be broken down and distributed across massive computing clusters. "When you break a computing problem into multiple parts and distribute it across the entire data center, the most indispensable factor is connectivity."
Huang did not hesitate to praise his partner and even stated on stage: "Ladies and gentlemen, (Marvell) this is the next trillion dollar company."
Murphy expressed that relying solely on one processor can no longer satisfy AI workloads; in the future, millions of processors will need to work together.
"Scaling of computation is essentially a connectivity challenge. The entire industry has solved the computing power bottleneck and is addressing the memory bottleneck, while the next bottleneck that limits infrastructure's limits is connectivity."
"Use copper wherever you can, use optics wherever you must"
In the dialog between Murphy and Huang, the segment with the most market reference value was their assessment of the timeline for transitioning from copper cables to optical fibers.
The strategic framework put forth by Huang was clear and direct: "Use copper cables wherever you can, and use optical devices only where necessary."
He explained that copper cables have physical limits on bandwidth and transmission distance. Before breaking this boundary, copper cables are a simple, low-cost, practical choice; once crossing the critical point, optical fibers take over to meet the expansion needs between racks, between data centers, and across data centers.
His core conclusion is:
"In the next 5 to 10 years, we will still use a lot of copper cables, while also using a massive amount of optical devices. These data centers are now part of the infrastructure."
This judgment of "copper and optical together, each guarding its own boundary" means for the market: whether in the copper or optical field, Marvell is in a position to continuously benefit — and Marvell is one of the few companies in the industry that can provide complete solutions in both directions.
The timeline for the copper-optical switch is backed by unavoidable physical laws. Murphy explained: the transmission distance of copper cables is inversely proportional to bandwidth; as bandwidth doubles, the transmission distance is halved.
The current fastest mass-produced system single-channel rate reaches 200 Gbps, corresponding to a copper cable length of about 2.5 meters, while the rack height is about 2 meters — considering internal wiring, 2.5 meters is already at the limit.
"When we upgrade to 400 Gbps, copper cables will not be able to fully connect the entire rack. The Copper Wall is moving, and it has already begun." Every time the Copper Wall moves a step to the right, the number of connections increases by at least an order of magnitude, which will directly ignite the demand for optical communication.
To address this physical limit, Marvell is heavily investing in CPO (Co-Packaged Optics) technology to solve the density and power consumption issues by connecting optical fibers directly into the package, adjacent to the computing chip.
On the day of the conference, Marvell officially launched a new 100T Ethernet switch specifically designed for AI data centers, boasting the industry's lowest power consumption, and showcased the 51.2T switch based on CPO, completely eliminating copper wiring at the board level.
"This is not a concept for the future; it is already being implemented now," Murphy stated. Once optical interconnections completely break the distance limits, future data centers will no longer have rigid physical boundaries of computation and memory, and 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 the extremely complex network architecture requirements, NVIDIA previously made a strategic investment of 2 billion dollars in Marvell, and the cooperation between the two is expanding into multiple dimensions such as optical communication, silicon photonics, and NV Link Fusion.
The emergence of NV Link Fusion aims to solve the customization pain points of cloud service providers (CSP). Huang explained that while designing their own custom chips (ASICs), cloud vendors still wish to integrate NVIDIA's system architecture.
"You don't have to buy everything from us; just buy a part. By merging NVIDIA's technology platform with Marvell's technology solutions, we can essentially build a decoupled, distributed, and heterogeneous data center."
In such an ecosystem, Marvell has found its irreplaceable ecological niche.
Murphy emphasized Marvell's neutral and critical position:
"We work closely with computing companies and also collaborate deeply with storage companies. In many ways, we are like 'Switzerland' in the industry, collaborating 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。
