Nvidia targets CPU market as SpaceX builds AI stack
Nvidia is expanding beyond GPUs with its new Vera CPU to counter custom silicon threats, while SpaceX pursues vertical integration backed by $81 billion in AI compute contracts.
Nvidia has revealed details of its next-generation Vera CPU, marking a strategic expansion beyond its dominant GPU business. The chipmaker will now compete directly with Intel and AMD in the central processing unit market.
This pivot responds to a growing shift among technology companies toward using CPUs for AI models and agents. Tech giants like Alphabet are designing their own custom CPUs for AI data centers, threatening Nvidia's 86% share of the GPU data center market. By offering its own CPU architecture, Nvidia aims to defend its position as AI workloads diversify.
While Nvidia adapts to shifting hardware demands, SpaceX is building a fully integrated AI infrastructure. Following its purchase of xAI, the aerospace company has expanded into a neocloud provider, selling data center compute power to external tech firms.
SpaceX has secured $81 billion in computing contracts with clients including Anthropic and Alphabet. Analysts view SpaceX as a "sovereign AI" company because it controls its entire technology stack. The company owns its Terafab semiconductor manufacturing facilities and operates the Grok AI model.
This vertical integration comes with significant capital demands. SpaceX spent nearly $21 billion on capital expenditures last year, and has already spent $10 billion in the first quarter of 2026. The massive cash burn presents a considerable risk for investors, who must bet that these heavy infrastructure investments will generate sufficient returns as demand for compute power scales.
The contrasting approaches highlight a maturing AI sector. Nvidia is broadening its product lines to remain the dominant vendor for third-party data centers, while SpaceX is attempting to internalize the entire AI supply chain.
