Marvell has significantly increased the company’s short-term and long-term guidance during its investor day in NYC. CEO Matt Murphy, followed by the rest of the executive team, were excited to share how Marvell’s TAM would reach $400B by 2030, supported by nearly $3T of CAPEX.
The four main pillars of growth were Interconnect (fastest growing), Switching and Storage, Custom (both XPU and XPU attach), along with traditional Communications. The overall market remains constrained by bandwidth, latency, power, and reach/distance. Each of these has a copper wall that will be broken with optics (NPO, CPO, photonics).
Networking plays a key and enabling role in AI factories, and Marvell has a robust and growing portfolio.
Scale-In is going to transform how we build and package GPU/XPUs as we run out of traditional beachhead. Today, high-speed SERDES of 112G and 224G dominate the market, but 448G is right around the corner. Strong supply chain and packaging will drive this market.
Scale-Up is still in its infancy and will rapidly expand from single-rack to multi-rack. CPO is a key enabling technology to get to multi-rack in volume, with some customers using NPO as an interim step. Scale-up networking will exceed $100B by the end of the decade and will become the largest of the DC networking markets.
Scale-Out is well known, with Ethernet already surpassing InfiniBand last year. Scale-Out benefits from the highest speed of switching and optics (800G->1.6T). Marvell highlights its silicon becoming a multi-billion-dollar opportunity here. Adding in optics, this is currently the largest market in AI Networking.
Scale-Across is where coherent-based technologies shine. DCI has been around for nearly a decade, and Inphi was a pioneer in this space. The need to connect AI data halls together and to allow agents to exit the data center drives a significant increase in the market. Filling in the distance gap takes this market from millions to tens of millions of ZR/ZR+/Coherent Lite optics by the end of the decade.
When we add all those together, the systems opportunity exceeds $300B by the end of the decade, and with the incorporation of CPO, we see the market for AI networking exceeding $1T in the early part of the 2030s.
Marvell also highlighted its XPU and XPU attach market with strong shipments into all four hyperscalers. The company also highlighted its work in memory, which we view as going hand in hand with the XPU in order to reduce the cost of tokens and provide more robust context windows.
Overall, the Marvell analyst day reinforced our view of the overall strength in Cloud and DC Capex as well as the sustainability of AI investment. We look forward to seeing Marvell grow into this multi-trillion-dollar opportunity.
