Piper Sandler Picks Nvidia and AMD on Rising AI Compute Demand

In a recent research report, Piper Sandler analysts led by David O’Connor said smarter AI models and growing use of agentic AI are increasing both training and inference compute needs, making compute capacity the main bottleneck for the industry.

The report states that demand pressures are visible in market signals. GPU prices have risen by at least 25% year-to-date, and the firm highlighted large-scale, gigawatt enterprise compute deals as a near-term catalyst for semiconductor and compute stocks.

Piper Sandler set a $300 price target for Nvidia, implying about a 34% upside from current levels, and assigned the company roughly an 80% share of the AI compute market in its model. The firm noted Nvidia’s product cadence as a differentiator. The report put Nvidia’s current valuation at about 14 times forward earnings, near the low end of the 14x to 22x range the firm applies to the broader compute group. Nvidia’s management recently guided to roughly $108 billion in revenue for the current quarter and a gross margin near 74%, figures the analysts used in their growth assessment.

The firm assigned a $600 price objective to Advanced Micro Devices, the highest target in its semiconductor coverage, implying roughly a 15% upside. Piper Sandler framed AMD’s case around the increasing role of general-purpose processors as AI systems perform multi-step planning and orchestration tasks that add overhead on CPUs. The report said AMD is taking server CPU share from Intel and gaining inference share with its Helios GPU line.

Piper Sandler noted that AMD has secured anchor customers for inference chips, including OpenAI, Meta and Anthropic. The analysts also said they want to see AMD broaden its core customer base before it can realistically compete for gigawatt-scale enterprise compute contracts. AMD has guided to about $13 billion in revenue for the current quarter, which the report said would represent roughly 41% growth year over year.

The report linked larger models and agentic AI to higher training and inference demand and to increased spending on GPUs and CPUs. It identified rising GPU prices, enterprise-scale compute deployments and customer diversification as factors to monitor for the sector.