Jim Cramer highlights Nvidia and Broadcom as key AI plays following the GPT-6 Astra launch, emphasizing their critical roles in model training and inference respectively.
Jim Cramer explicitly recommends Nvidia as the primary beneficiary of the GPT-6 Astra model, citing its critical role in training infrastructure. The company's Grace Blackwell NVLink72 systems are the backbone of OpenAI's model development, with 100,000 units already in use and 400,000 more being brought online.
He views Nvidia as the essential foundation for AI training, noting that OpenAI's success is directly tied to Nvidia's hardware ecosystem.
Broadcom is positioned as a key beneficiary through its custom 'Jalapeno' chip, which is optimized for inference rather than training. Cramer highlights its role in the Astra rollout, while the company benefits from massive infrastructure commitments from major clients like Anthropic and OpenAI.
Broadcom focuses on inference—the day-to-day work of running finished models—via custom silicon, whereas Nvidia dominates the training phase.
Intel is held in Jim Cramer's charitable trust as part of a diversified strategy to capture growth across the entire AI chip stack. Cramer maintains a neutral-to-positive stance, betting that the broader AI wave will benefit multiple players beyond just the primary training providers.
Cramer believes the AI wave is real and broad, preferring to spread his investment across the entire stack rather than picking a single winner.
Micron is included in Cramer's charitable trust as a component of his broad AI investment thesis. Like Intel, it serves as a hedge to ensure exposure to the overall expansion of AI infrastructure rather than relying on a single chip supplier.
The article does not explicitly link Micron to Astra, but Cramer holds it as part of a diversified bet on the overall AI infrastructure buildout.
Jim Cramer has identified Nvidia and Broadcom as the primary beneficiaries of OpenAI's new GPT-6 Astra model. While Nvidia dominates the training infrastructure with its Grace Blackwell systems, Broadcom is positioned for inference gains through its custom Jalapeno chip. The market reaction remains split, as investors weigh the distinct roles these chipmakers play in the evolving AI ecosystem.
Cramer views Nvidia as the essential foundation for training large-scale models like Astra, citing the massive scale of Grace Blackwell systems currently being deployed.
The Jalapeno chip is a custom silicon solution designed specifically for inference, or the day-to-day execution of finished AI models, offering improved power efficiency and lower latency.