Prediction: This Artificial Intelligence (AI) Chip Stock Will Become the Next Nvidia by 2030

Marvell’s ASIC business has scaled to $1.5 billion in annual revenue, with 18 hyperscaler design wins and reported Google co-development talks for two custom AI chips sending the stock up more than 13% in one session. Management is targeting $15 billion of revenue in fiscal 2028, after fiscal 2026 sales rose 42% to $8.2 billion. The recently announced $3.25 billion Celestial AI acquisition strengthens Marvell’s photonic interconnect capabilities and supports a longer-term AI infrastructure growth thesis.

Analysis

The market is starting to price Marvell less as a component supplier and more as the toll booth for hyperscaler silicon complexity. The second-order winner is not just MRVL itself, but the ecosystem that solves integration bottlenecks: advanced packaging, EDA, foundry capacity, and high-speed memory/interconnect vendors should see a longer demand runway as custom AI moves from a chip story to a systems story. That shift is important because once a hyperscaler commits to a co-designed architecture, switching costs become operational rather than just economic.

The key competitive implication is that Nvidia's moat is being attacked from the edges, not head-on. If hyperscalers keep diverting inference workloads into proprietary ASICs, Nvidia's training dominance can remain intact while its addressable share of AI capex mix slowly dilutes over 12-24 months. That is bullish for cost-down narratives at GOOGL, AMZN, and MSFT, but it also creates a hidden dependency: the more they optimize around their own chips, the more they need elite networking and photonics to keep performance scaling, which expands MRVL's role if execution holds.

The main risk is that the current enthusiasm gets ahead of revenue conversion. Design wins and co-development talks are useful leading indicators, but the stock can re-rate sharply if the market realizes the monetization curve is back-end loaded and subject to slips in tape-out, yield, or customer qualification. Over the next 1-3 quarters, any disappointment on hyperscaler capex pacing or a delay in photonics adoption could compress multiples quickly, especially after the sharp recent run.

The contrarian view is that this may be less a ‘next Nvidia’ moment than a very profitable, but more cyclical, infrastructure enabling business. Consensus may be underestimating the margin leverage of being embedded in multiple custom programs, but overestimating the durability of narrative-driven upside if the next few quarters do not show accelerating backlog conversion. In other words, the fundamental setup is good, but the risk/reward is likely better on pullbacks or via structured upside exposure than chasing strength outright.

Key Decisions for Investors

• Buy MRVL on a 5%-8% pullback; target 9-12 month upside if custom-silicon revenue inflects as expected, with downside capped by taking profits into any post-announcement gap-up.
• Pair long MRVL / short NVDA for a 6-12 month relative-value trade if you expect hyperscaler ASIC adoption to keep expanding; thesis is mix shift, not GPU collapse.
• Overweight GOOGL and AMZN on the view that internal chip programs reduce unit economics of AI inference over 12-18 months; best expressed as a basket versus QQQ to isolate idiosyncratic capex leverage.
• Use call spreads in MRVL rather than outright common stock to capture upside from design-win re-rating while limiting valuation compression risk if execution slips over the next 1-2 quarters.
• Watch for weakness in high-speed networking and photonics peers as a secondary beneficiary basket; if MRVL confirms photonic adoption, the next leg should spill into the broader interconnect supply chain.