Piper Sandler analyst David O'Connor initiated coverage of Marvell Technology (MRVL) with a Buy rating and a $270 price target, citing the company's data center positioning and a major agreement with Google as the driving factors. MRVL stock was trading around $231.69 at the time of the note and has rallied 167% year-to-date.
O'Connor, who also issued Buy ratings on Nvidia, AMD, Broadcom, and Arm Holdings, and Hold ratings on Qualcomm and Intel, highlighted Marvell's roughly 10% share of the data center component market, dominant positions in digital signal processors and custom connectivity solutions, and its Celestial AI acquisition. The central piece of the bullish case is the $120 billion Google agreement, with TPU programs expected to begin in fiscal 2029 and projected to generate roughly $18 billion in annual revenue once fully ramped. Marvell's core XPU customers are believed to include Amazon's AWS Trainium and Microsoft's Maia accelerator chip programs.
Piper Sandler also named Nvidia and AMD as top sector picks. O'Connor set a $300 price target on Nvidia, implying about 34% upside, and called the company the outright AI compute market leader with an 80% market share. He views recent lagging performance as temporary mispricing and flagged gigawatt-scale enterprise compute deals as a key near-term catalyst. Nvidia recently guided for $108 billion in current-quarter sales with a 74% gross margin.
AMD received a $600 price target, the highest in Piper Sandler's semiconductor coverage group. The analyst sees agentic AI accelerating demand for CPU server chips, where AMD is taking share from Intel, while its Helios GPU line is logging inference share gains against Nvidia. AMD has anchor customers including OpenAI, Meta, and Anthropic, but O'Connor wants to see a broader customer base before the company lands its own gigawatt-scale enterprise deals. AMD guided for about $13 billion in current-quarter revenue, up 41% year-over-year.
The broader backdrop is a self-reinforcing AI compute cycle: smarter models require more training compute, rising adoption of agentic AI is driving inference demand, and compute capacity has become the primary bottleneck. Piper Sandler noted GPU prices have increased by at least 25% year-to-date.
