When Less Power Means More Possibilities
The AI Platform Alliance and its members have challenged conventional wisdom about scaling out AI inference infrastructure. Whether running inference workloads directly on energy-efficient Ampere® CPUs or various Alliance partners’ domain-specific accelerators, the Alliance intends to offer viable alternatives to traditional, power-hungry multi-GPU deployments for certain deployment sizes.
We’re excited to showcase a breakthrough solution developed through the collaboration of Alliance partners Kamiwaza and Ampere Computing at Computex 2025. This solution demonstrates that sophisticated generative AI workloads can run efficiently on our AmpereOne® M processors, delivering performance previously achieved through dedicated hardware accelerators such as GPUs.
The Power Efficiency Imperative
As AI adoption accelerates due to smaller, more tailored models gaining traction, the issue of right-sizing compute hardware becomes ever-more important. Hardware acquisition and operational costs need to be weighed, and multi-GPU systems that might require specialized cooling no longer are the de facto standard.
This joint effort between Kamiwaza and Ampere addresses these concerns around scalability and efficiency of large-scale AI inference deployments. By optimizing AI workloads for cloud-native CPU architecture, this solution enables enterprises to deploy sophisticated AI capabilities with:
- Substantially lower power consumption compared to hardware
- Significantly reduced cooling requirements
- Standard rack infrastructure without specialized modifications
- Simplified deployment and management
Seeing is Believing at Computex
Attendees at Computex 2025 can witness Kamiwaza’s advanced AI orchestration system running on our AmpereOne M processors. The demonstration includes:
- A 7 billion parameter visual language model processing complex handwritten notes in seconds
- Extraction of structured data with human-level accuracy
- Concurrent handling of multiple document workflows, including file creation, translation and currency conversion
- Workflow blueprints for sales data analysis, cross-functional product improvement planning, and web research report creation
- Real-time visualization of AI agents collaborating on complex enterprise-scale tasks
This isn’t just a technical demonstration—it’s a proof point that energy-efficient AI is ready for enterprise deployment today.
Sustainable AI for Data-Sensitive Environments
Beyond power efficiency, our collaboration with Kamiwaza enables enterprises to deploy AI capabilities in environments where data sovereignty and security are paramount. By running sophisticated AI workloads on standard CPU infrastructure, organizations can:
- Process sensitive data on-premises without compromise
- Maintain complete control over information flow
- Implement AI in regulated environments
- Leverage existing infrastructure investments
For industries like healthcare, financial services, and government where data must remain secure, this approach removes a significant barrier to AI adoption.
From Demonstration to Deployment
This collaboration paves a practical path to AI deployment that aligns with enterprise infrastructure strategies. Kamiwaza and Ampere work with leading OEMs via the AI Platform Alliance to deliver validated configurations that organizations can deploy quickly and confidently.
Early implementations have demonstrated that this approach reduces deployment timelines from months to days, without requiring specialized AI infrastructure expertise. For enterprises that have hesitated to adopt AI due to infrastructure concerns, our collaboration offers a compelling alternative to traditional GPU deployments.
Redefining Enterprise AI Infrastructure
Attendees at Computex 2025 can see this solution in action. We’ll be sharing more details in the coming weeks, including customer testimonials, benchmark results, and implementation guidance.
As an industry, we have a responsibility to make AI more accessible, sustainable, and practical for organizations of all types. The collaboration between Kamiwaza and Ampere through the AI Platform Alliance demonstrates that energy-efficient AI isn’t just possible—it’s available today.
Learn more and access the solution brief here.
