Marvell’s Unlikely Alliance: How Nvidia’s $2 Billion Bet Reshapes the Custom Chip Race

Nvidia's $2B investment in Marvell signals a strategic shift in AI infrastructure, with Marvell targeting $16.5B revenue by 2028 and a booming optical networking business.

When the industry’s dominant force writes a $2 billion check to a rival whose entire business model revolves around loosening that very dominance, the market takes notice. That is precisely the scenario unfolding at Marvell Technology, where Nvidia’s surprise investment has cast a new light on the shifting dynamics of AI infrastructure.

The partnership, centered on Nvidia’s NVLink Fusion technology and joint development of silicon photonics, marks a strategic pivot for both companies. Rather than resisting the hyperscaler push toward custom silicon—where cloud giants like Amazon, Google, and Microsoft design their own chips to reduce reliance on off-the-shelf GPUs—Nvidia has opted to co-opt the trend. Marvell will supply custom processors and networking gear for scaling, while Nvidia contributes its Vera CPU, ConnectX network cards, Bluefield DPUs, and Spectrum-X switches.

A Billion-Dollar Networking Ambition

Beyond the headline-grabbing investment, Marvell is quietly building a formidable business in optical networking—a segment the company believes will hit a $1 billion annualized revenue run rate by fiscal 2028. That would represent a doubling from the roughly $500 million generated in fiscal 2026. The driver: hyperscalers racing to connect massive AI clusters with high-bandwidth, low-latency links.

The company recently unveiled what it calls the industry’s first secure 1.6-terabit ZR and ZR+ modules for data center interconnection, powered by a new coherent 2-nanometer DSP platform now entering the sampling phase. These products sit at the heart of Marvell’s broader interconnect business, which management expects to grow more than 70% in the current fiscal year.

Acquisitions are fueling the push. The purchases of Celestial AI and XConn Technologies have added silicon photonics and CXL switching technology to the portfolio—both considered critical for next-generation data center architectures.

Raising the Bar on Revenue

Marvell’s financial outlook has become markedly more ambitious. The company now projects fiscal 2028 revenue of approximately $16.5 billion, up from a prior forecast of $15 billion. Even more striking is the medium-term target: the custom chip business alone is expected to generate more than $10 billion in revenue by fiscal 2029, driven by accelerated hyperscaler investment in proprietary chip architectures.

The momentum is already visible. In the most recent quarter, revenue climbed 28% to $2.42 billion, beating analyst expectations. For the current quarter, Marvell forecasts around $2.7 billion, again above the consensus estimate of $2.6 billion. CEO Matt Murphy attributed the strength to sustained AI demand accelerating growth across both optical products and custom chips.

The Volatility Trade-Off

For shareholders, the growth story comes with a stomach-churning caveat. The stock currently trades around €186.66, up 2.16% on the day and 3.56% over the past week, but still 35.71% below its 52-week high of €290.35 reached in June. The annualized 30-day volatility stands at a staggering 94.16%—a figure that separates conviction investors from the faint-hearted.

The recent seven-day rally of 12.32% has helped lift the stock out of oversold territory, with the RSI at 43.3. Analyst consensus points to a 12-month price target of roughly $240.48, based on 29 analysts’ estimates, with individual targets ranging as high as $400. The implied upside from current levels is around 19%, though the wide dispersion of forecasts underscores the uncertainty baked into the stock.

The Double-Edged Sword of Dependence

Marvell’s alliance with Nvidia and its upgraded guidance tell the same story from different angles: hyperscalers are pouring enormous sums into differentiated chips, and Marvell has positioned itself as an indispensable partner. Nvidia’s own capital stake reinforces that status.

Yet this very dependence cuts both ways. Marvell’s fate increasingly hinges on the investment decisions of a handful of hyperscale customers. Any pause in AI infrastructure spending, any shift toward in-house chip design by a major cloud player—and the growth narrative would come under immediate pressure.

The market now treats Marvell less as a cyclical semiconductor stock and more as a structural beneficiary of the AI buildout. That bet assumes custom chips will capture a growing share of hyperscaler infrastructure budgets, rather than Nvidia’s standard GPUs. Whether the wager pays off may ultimately depend less on quarterly earnings and more on whether the Nvidia partnership becomes a genuine force that reshapes the competitive landscape of AI chip design.

The next checkpoint arrives with Marvell’s quarterly report in late August, when investors will see if the company can sustain its pace in custom AI chips and 800G/1.6T networking technology. Until then, the stock’s 94% volatility serves as a daily reminder that this growth story comes with no guarantees—only the promise that every trading day will bring fresh conviction from those who believe in the custom silicon revolution, and fresh skepticism from those who don’t.