Marvell’s AI Edge: 30% Growth at 26x Earnings
Investors are obsessed with the AI compute war, focusing on who wins market share. They focus on Nvidia (NVDA), AMD, and the hyperscalers who are building their custom chips.
But they are overlooking a key factor. As AI chips get faster and models get larger, the “wires,” or the interconnects between them, need to keep up.
By 2027, electricity will be too inefficient and too hot to move AI data. The focus could shift to a metric no one is talking about: the interconnect tax.
That’s what Marvell (MRVL) can extract. It already has a deep foothold in data center networking and interconnect silicon, is taking steps to position itself as the de facto architect of next-generation data center interconnects, enabling it to capture value from nearly every AI transaction, regardless of which GPU or TPU is doing the work.
Why Are Interconnects So Important?
AI performance depends on thousands of chips working together across servers. In that setup, data movement becomes critical else expensive GPUs sit idle.
This is where physics becomes the constraint. Electrical signals work well over very short distances, but at AI speeds they consume too much power, generate heat, and degrade over distance. The industry is steadily shifting to optical connections because “Light” is far more efficient for moving large amounts of data.
What’s driving this data explosion? AI inference. Unlike AI model training, which runs on localized clusters, the “AI inference” (think how you use Gemini, ChatGPT, Claude etc) requires a global, low-latency network spanning distributed data centers. As models evolve into always-on utilities, the system must handle massive, real-time traffic, pushing architectures to scale out.
Besides interconnects, CPUs are also likely to see a revival as AI inference workloads soar, and Intel is likely to be a beneficiary.
What Advancements Is Marvell Making?
Marvell is a key supplier behind the shift to optical connectivity in AI data centers. It provides optical DSPs used in pluggable modules that move data between servers and holds a dominant share in high-speed segments like 800G modules (often estimated in the 60–80% range).
But this is only the first phase. Today’s optical systems still rely heavily on signal processing to compensate for losses in electrical and optical channels. The next step is to reduce how much of that correction is needed by moving optics closer to compute. This is where Celestial AI, a company that Marvell acquired last year, fits in.
Over time, this also enables faster access to memory and tighter system-level integration. For instance, this could mean that hyperscalers would need up to 30% less HBM4 memory – which has soared in price and remains in short supply.
What About the Competition?
Sure, Marvell has rivals in both the interconnect and custom ASIC space, the most notable being Broadcom (AVGO). However, Marvell could have an edge. The aggressive price hikes by Broadcom following its VMware acquisition serve as a quiet reminder to hyperscalers of vendor concentration risk. That dynamic creates an opening for Marvell to position itself as a more neutral, strategic counterweight to Broadcom, not just on performance but on long-term partnership credibility.
How The Stock Can Be Rerated Higher
For FY’26, 74.4% of revenue came from Marvell’s data center segment, which generated $6.1 billion, highlighting how central AI infrastructure already is to the business. Within this, Marvell’s interconnect portfolio is a key driver, expected to grow more than 50% in fiscal 2027. The company has also raised its data center switch target to surpass $600 million, roughly double FY ’26.
Yet the stock trades at just 26x consensus FY’27 earnings and a mere 18x FY’28 consensus earnings, despite expected revenue growth of more than 30% annually over the next two years.
The valuation gap is even more striking in absolute terms. Marvell’s market cap is about $85 billion versus Broadcom at roughly $1.4 trillion, leaving substantial room for rerating. If Marvell delivers on interconnect and compute wins, its exposure to nearly every AI workload makes the current valuation look increasingly conservative.
While Marvell’s opportunity looks compelling, if you want to deploy high-conviction, data-backed strategies across your entire portfolio without managing the day-to-day execution yourself, we can help. Our Trefis HQ strategy has outperformed its market benchmark (a combination of the S&P 500, S&P mid-cap, and Russell 2000) to produce over 105% returns since inception. Readers can see the full HQ performance metrics and 5 reasons why HQ outperformed.