The cost of deploying AI is rising. Specifically, growing agent usage may significantly increase token consumption and cloud-processing costs. Goldman Sachs expects a 24-fold increase in token consumption by 2030 thanks to expanded use of AI agents. Increasing costs mean rising concerns. An EY US AI Pulse Survey found that 82% of senior leaders are apprehensive about AI token usage and related costs, with 98% saying they are reevaluating their approach.
“Rising cloud costs associated with agentic AI, along with latency, privacy, and security considerations, are driving demand for AI workloads at the edge,” says Bruce Broussard, HP’s interim CEO, on its second quarter earnings call. “As a result, customers are building AI at the edge using smaller, open source, and proprietary models with more capable hardware and secure software layers.”
AI PCs and workstations, purpose-built for AI processing, offer a path to more budget-friendly AI, allowing smaller models and workloads to be handled locally with more predictability than consumption-based pricing. Here are three ways AI PCs and workstations can help counter rising cloud costs.
- Balance edge and cloud infrastructure. AI PCs and workstations shouldn’t be viewed as a wholesale replacement for cloud, but rather as an alternative option for running certain inferencing AI workloads at the edge. After the initial investment, local processing can reduce reliance on certain recurring, query-based cloud charges. Combined with cloud resources in a well-architected AI infrastructure, AI PCs and workstations can help balance overall expenses.
- Run appropriate AI workloads locally. AI PCs and workstations are equipped with dedicated neural processing units (NPUs), AI accelerators, and GPUs. This makes them well-suited for running small models locally, often providing better latency and performance, depending on the hardware configuration and the workload. AI PCs and workstations like those from HP can deliver up to 85 tera operations per second (TOPS), making them ideal for certain AI and data analysis workflows. Processing advancements also optimize power consumption, providing all-day battery life for certain workloads. And built-in security featuresand hardware-enforced virtualization provide advanced endpoint protection.
- Align compute choices with business outcomes. Every AI strategy should take workload placement into account. IT needs to evaluate requirements such as model size, latency, privacy, security, connectivity, and cost. Those considerations will determine whether edge, workstation, or cloud processing is the best fit.
“End-user compute is a strategic pillar that supports where the organization is going on its AI journey over the long term,” says Shaun Copper, Field Application Engineer at AMD. “CIOs must determine the most cost-effective and efficient avenue to run AI workloads, whether that’s on AI PCs and workstations or in the cloud.”
HP’s portfolio of next-generation AI PCs, laptops, and workstations – advanced by AMD – features dedicated NPUs designed to accelerate complex AI tasks locally and create a more responsive experience. Those capabilities, bolstered by multi-layered security protections, position the edge as a viable and cost-effective cloud alternative for running high-performance AI workloads.
Click here for more information on HP’s AI PC and workstations, advanced by AMD.