Analysts are intrested in these 5 stocks: ( (NVDA) ), ( (MRVL) ), ( (AMD) ), ( (AVGO) ) and ( (QCOM) ). Here is a breakdown of their recent ratings and the rationale behind them.
Nvidia is once again in the spotlight as analysts highlight its dominance in AI compute, where it holds roughly 80% market share and half of unit sales. David O’Connor at Piper Sandler has initiated coverage with an Overweight rating and a $300 price target, arguing that Nvidia remains one of the cheapest names in the AI space despite rapid revenue and earnings growth.
The broker’s thesis leans heavily on a fast-expanding AI market, including Agentic AI workloads that are pushing GPU demand far beyond current supply. O’Connor models Nvidia’s revenues growing around 47% annually, with earnings projected to reach about $30 per share by FY30, supported by an estimated $2T total addressable AI compute market by 2030 and a long period of constrained supply.
Marvell Technology is emerging as a quiet winner in data center plumbing, as O’Connor initiates the stock at Overweight with a $270 price target. The analyst notes Marvell’s strong position in optical DSP chips and custom connectivity, backed by deep ties to hyperscalers and a strategy that looks well aligned with the next wave in co-packaged optics networking.
A centerpiece of the bullish case is Marvell’s $120B Google agreement, which the report views as transformative by ramping “attach” programs around cloud TPUs from FY29 onward. Earnings are projected to grow at roughly 45% annually to about $19 per share by CY30, while an upcoming analyst day is flagged as a potential catalyst for investors watching how management frames its long-term model.
Advanced Micro Devices is positioned as an Agentic AI sweet spot, with O’Connor starting coverage at Overweight and setting a punchy $600 price target. The call hinges on AMD’s rising share in CPU server chips as AI workloads demand more orchestration, driving a shift toward a one-to-one CPU-to-GPU ratio in data centers.
On the GPU side, AMD’s Helios and Instinct accelerators are ramping with anchor clients like OpenAI, Meta and Anthropic, who together have committed to about 14GW of AI capacity over the next few years. The report sees AMD’s revenues growing near 50% annually and earnings at a 65% clip, with potential earnings power of $53 by 2030 if the company can secure enough supply to meet this hunger for compute.
Broadcom is cast as the “ASIC compute king” for AI inference, and the stock is initiated at Overweight with a $460 target based on a 14x multiple of FY28 earnings. The firm controls about 75% of the AI ASIC market, supplying custom chips to major hyperscalers for workloads where tailored silicon delivers better total cost of ownership than general-purpose GPUs.
O’Connor estimates Broadcom will double ASIC unit shipments from 5M in 2026 to 10M in 2027, reaching 24M by 2030 as demand for Agentic AI inference accelerates. With networking products such as Tomahawk and Jericho often sold alongside ASICs, the report sees combined content of $20–30B per gigawatt of AI capacity and an earnings growth trajectory near 51% a year, even as competition and supply constraints remain key watchpoints.
Qualcomm rounds out the coverage with a more cautious tone, as O’Connor initiates at Neutral and a $190 price target. The company is described as late to the AI data center party but still arriving at a time when the market is compute-starved and architectures are not yet locked in, giving it room to carve out a role.
Management has announced a suite of custom, accelerator and CPU server design wins that could lift data center revenues to about $15B and roughly a quarter of total sales by FY29. Yet with much of this story already priced in and execution risks looming, the report values Qualcomm at 14x FY29 earnings of around $15, while noting that smartphone exposure should shrink to about 40% of sales and 50% of profits as AI-related businesses grow.