Marvell stock price target raised to $200 by TD Cowen on networking

TD Cowen raised its price target on Marvell Technology to $200 from $180 while keeping a Hold rating, citing stronger networking and connectivity prospects through fiscal 2028. The firm said Marvell’s networking franchise should be the largest contributor to dollar growth, with momentum in interconnect offsetting ongoing visibility challenges in custom XPU. Shares were already up 208% over the past year, and the stock traded around $200.10 with a $176.43 billion market cap.

Analysis

The key signal is that the market is migrating from a pure AI-beta trade in semis toward a more selective infrastructure trade where networking, interconnect, and optics are the scarce bottlenecks. That matters because the next leg of earnings upside in data center supply chains is likely to accrue to companies that sit between accelerators and the fabric, not necessarily to the accelerator vendors themselves. In other words, the market is beginning to price in that bandwidth and latency constraints can force a larger dollar pool into switches, SerDes, and optical content even if custom compute visibility remains fuzzy.

The second-order effect is that multiple expansion may now outrun fundamental clarity. A stock that has already rerated sharply can keep grinding higher on estimate revisions, but the probability distribution widens materially once the consensus moves from "beat-and-raise" to "prove it in bookings." That creates a classic setup where the equity can stay elevated for months on incremental guidance updates, yet becomes fragile to any pause in order acceleration or a change in hyperscaler capex phasing.

For competitors and supply chain peers, the read-through is asymmetric: vendors with exposure to network silicon and optical interconnect should see the strongest durability of demand, while names more dependent on custom XPU win rates may face a lagging confirmation cycle. The real watch item is whether improved networking demand is simply pulling forward spend from adjacent categories; if so, the durability of the thesis into the back half of next year will depend on whether AI clusters continue to densify fast enough to offset pricing pressure and customer concentration.

The contrarian view is that the market may be underestimating how much of the optimism is already embedded after a large run and a wave of target hikes. If the business mix shift toward networking is structurally good but not enough to justify continued multiple expansion, then the stock can become range-bound despite positive fundamentals. That makes this less attractive as an outright momentum chase and more interesting as a relative-value expression versus weaker infrastructure names or as a hedged long against broader semiconductor beta.

Key Decisions for Investors

• Long MRVL vs. SOXX for 1-3 months: own the name with the clearest networking/interconnect leverage while neutralizing broad semiconductor beta; target outperformance if the market continues rewarding AI infrastructure bottlenecks over generic chip exposure.
• If already long MRVL, sell covered calls 10-15% out of the money with 6-8 week tenor to monetize elevated implied volatility; the setup favors upside grind more than explosive rerating after the recent move.
• Pair long MRVL / short a custom-XPU-dependent semiconductor peer basket for 2-4 quarters: thesis is that networking content has better visibility than accelerator win rates, so relative estimates should converge in MRVL’s favor if AI capex remains lumpy.
• Do not add aggressively on strength above current levels without a pullback or booking confirmation; use a 5-8% retracement or next quarterly print as the entry window, since the risk/reward deteriorates once revisions are fully priced.
• Set a catalyst watch for the next guidance update and any commentary on interconnect order velocity; if networking growth decelerates while target prices keep rising, reduce exposure quickly because the stock’s downside will be driven by multiple compression rather than earnings misses.