HPE and AMD Advocate Return to On-Prem Data Centers to Cut AI Costs
Photo: ITmedia
Quick answer
HPE and AMD urge businesses to move AI computations back to on-premises data centers to reduce token costs and dependency on external models.
As AI agents proliferate, companies grapple with soaring token processing costs. HPE CTO Antonio Neri, speaking at the company’s annual event, warned that dependence on external generative AI models leads to unpredictable hidden expenses.
HPE and AMD advocate a strategy to return AI workloads to corporate data centers. This shift enables companies to slash token costs while regaining full control over infrastructure, avoiding reliance on external providers. AMD’s example demonstrates how modern hardware solutions can make this strategy viable in practice.
Moving computations to in-house data centers also unlocks performance and security optimizations. Companies can tailor infrastructure to their needs, mitigating data leak risks and boosting AI model efficiency.
Common questions
- Why are companies facing rising AI token costs?
- Costs are escalating due to increased token consumption when using external models for generative AI. The more AI agents deployed, the higher the data processing expenses.
- What solution do HPE and AMD propose to lower costs?
- They recommend relocating AI computations to corporate data centers, allowing better cost control and reduced reliance on third-party providers.
- What benefits do on-prem data centers offer for AI?
- In-house data centers provide greater flexibility, infrastructure control, and optimized costs for computing and data storage.
Dzen feed: /feed/dzen.xml · RSS: /feed.xml