Jim Cramer points to a second AI wave for AMD as the industry shifts from training to inference

Jim Cramer used Advanced Micro Devices’ AI push to make a broader point about where corporate spending may go next in artificial intelligence: from model training to inference, and then toward more “agentic” applications that can take actions. In remarks tied to AMD’s artificial intelligence opportunity, Cramer framed the company’s position as opening up a new growth channel rather than fading after the initial wave of spending to train large models.

Cramer’s core argument, as described by Yahoo Finance, is that the AI industry is moving through stages. After the phase when organizations race to train ever-larger models, companies increasingly need computing capacity to run those models at scale. That shift, in turn, changes what buyers prioritize, including latency, throughput, and power efficiency for inference workloads.

The segment also ties AMD’s opportunity to agentic AI, a term commonly used for systems that do more than generate text or images. In practice, agentic AI refers to software and model stacks that can plan steps, call tools, and execute tasks toward a goal. If enterprises move from passive responses to systems that act, the economics of running models continuously and coordinating decisions become more important.

For AMD, that framing matters because demand for AI hardware is not only about getting models trained. It also depends on what customers buy to deploy them across data centers and cloud environments, and how those deployments scale. While the company’s specific product strategy was not detailed in the Yahoo Finance excerpt, Cramer’s comments suggest the market is watching whether vendors can capture spending tied to ongoing inference, not just one-time training buildouts.

Cramer’s remarks, as characterized in the report, were positioned as an additional “growth opportunity” for AMD as industry priorities shift. That is consistent with how the AI supply chain has been evolving in recent years, with chips and systems increasingly judged on real-world deployment performance and costs for sustained workloads.

At the same time, the description of the Yahoo Finance piece does not provide new financial guidance, named AMD products, or disclosed performance benchmarks. It also does not spell out whether AMD’s near-term results are being driven by inference demand specifically, or how quickly agentic AI deployments are translating into incremental orders.

For readers tracking the semiconductor angle of the AI buildout, the next question is how AMD’s public disclosures, including earnings materials and investor presentations, link product shipments and backlog to inference and enterprise deployments. If AMD highlights traction in inference-optimized systems or customer wins tied to agentic workloads, that would help confirm the thesis implied by Cramer’s comments.

What remains unclear from the information available here is the timing and magnitude of any “second wave” for AMD. Without additional detail on the companies and workloads Cramer referenced, the impact may be more directional than immediate, and investors will likely look for corroboration in AMD’s reported segment trends and guidance.