AMD secured a 6 gigawatt GPU partnership with Meta, marking one of the largest AI infrastructure deals disclosed to date. The chipmaker simultaneously invested in Nutanix to strengthen hybrid cloud AI capabilities.
Red Hat and NVIDIA co-launched Red Hat AI Factory with NVIDIA, an enterprise-grade agentic AI platform. Supermicro certified its accelerated computing systems for the platform, targeting mission-critical AI workloads across hybrid cloud environments.
"Supermicro has an extensive portfolio of Red Hat-certified systems and is dedicated to delivering the most advanced accelerated computing infrastructure for AI factories," said Vik Malyala, Supermicro executive. The validated solutions aim to simplify deployment and scaling of enterprise AI workloads.
Cryptocurrency mining companies are converting facilities to AI data centers. DMG Blockchain Solutions adjusted equipment operations to prioritize profitability over hashrate generation, receiving a $1.5 million energy efficiency incentive. Mawson Infrastructure Group reported a preliminary net loss of $23.8 million for 2025 while adopting a stockholder rights agreement that permits board consideration of strategic transactions.
The infrastructure race extends to storage. Pure Storage rebranded as Everpure and announced an acquisition expected to close in Q2 FY27, positioning for AI-driven data growth.
AMD's dual strategy of securing massive GPU contracts with hyperscalers while investing in hybrid cloud infrastructure companies reflects the fragmented nature of enterprise AI deployment. Organizations are splitting between public cloud GPU rentals and on-premises or hybrid solutions.
The Red Hat-NVIDIA collaboration targets enterprises hesitant to commit fully to public cloud AI, offering Red Hat's open-source enterprise platform combined with NVIDIA's accelerated computing stack. This approach competes directly with hyperscaler-exclusive AI services.
Cryptocurrency mining facilities offer ready-made power infrastructure and cooling systems that AI data centers require. The pivot by DMG and Mawson indicates that repurposing existing high-power facilities is faster than building new AI data centers from scratch.
The 6 gigawatt Meta-AMD deal dwarfs typical data center power allocations, suggesting Meta is preparing for AI model training at unprecedented scale. For context, a single gigawatt can power roughly 750,000 homes, making this a 4.5 million home-equivalent power commitment to GPU infrastructure.