“Our goal is to give PC partners the flexibility to differentiate their systems while giving customers greater choice in how they adopt and use AI PCs,” says Vinay Awasthi, Senior Vice President, Asia Pacific Japan & India, AMD.
When it comes to AI PCs, PC manufacturers are focusing on how they can bring more capabilities to users at the best cost, while also ensuring that performance is not compromised.
According to statistics from IDC , worldwide PC shipments fell over 20.1% year over year in Q3 2026 to 62.7 million units. The drop not only represented the second consecutive decline, but it was also a deeper one than the 3.8% drop in Q2. The reason for this is undoubtedly due to the continued supply constraints and the inventory pull-in earlier in the year.
For organizations though, there is still a strong need for AI PCs, especially with more AI workloads being used for daily tasks. PC manufacturers are also ensuring that they are capable of keeping costs manageable in order to ensure customers are able to update their devices to support their AI productivity goals.
On the components side, AMD has been focused on redefining the modern workplace by ensuring its chips are able to support increasingly demanding tasks and compute needed to use AI.
Vinay Awasthi, Senior Vice President, Asia Pacific Japan & India, AMD shares how with CRN Asia.
What are the trends you are seeing when it comes to AI PCs in APAC? Is there really a strong demand and a shortage or are organizations still making do with their existing devices?
AI PCs will evolve beyond productivity devices for employees to become what we describe as Agentic Computers. These are systems designed not only to help users complete tasks, but to continuously run AI agents locally – essentially a computer for your agents to run.
This will fundamentally redefine the modern workplace, shifting employees from performing every task manually to delegating, supervising and validating work performed by AI agents, while focusing their own time on critical thinking, creativity and higher-value decision-making. Over the next few years, employees will increasingly work alongside AI agents that run locally and continuously on their PCs or dedicated computing systems, automating suitable tasks.
Organizations are taking a pragmatic approach to PC refreshes. Existing devices will continue to play a role, but as AI becomes more integrated into everyday workflows, businesses are increasingly considering how their next generation of PCs can support those new experiences.
The decision goes beyond AI performance alone. IT teams need to balance security, manageability, compatibility and lifecycle stability, while on-device AI can offer additional benefits such as responsiveness, greater data control and less reliance on cloud infrastructure. Ultimately, organizations will adopt AI PCs at the pace and in the areas that make sense for their workloads and business needs.
Why enterprises need to budget for AI agents the way they budget for headcount?
As organizations move from AI assistants to agentic AI, AI becomes more than an infrastructure investment or subscription cost, but an ongoing operational consideration because agents continuously reason, retrieve information, call tools and coordinate tasks.
This means a single employee can now trigger multiple concurrent agents and sub-agents running in parallel across different workflows, so compute consumption scales with agent activity, not with the number of people at their desks. As a result, businesses will need to plan not only around headcount, application seats or developers, but around the AI capacity required to support these workflows as employees increasingly work alongside multiple AI agents and sub-agents across different parts of their day.
The goal isn’t simply to spend more on AI, it’s to spend smarter. The productivity multiplier from AI Agents comes with a compute multiplier. More agents mean more reasoning, more context, more models and more actions. All of that requires more compute and more tokens. The more useful AI becomes, the more compute it demands behind the scenes. That's why agentic AI is accelerating token consumption at an astonishing pace. We're moving from short interactions to AI systems that can think, act and keep working.
The organizations that will succeed are the ones that route each workload to the right compute resource and environment, using the right-sized model for the task instead of treating every workload the same or relying exclusively on the largest models running in the cloud. That means building a balanced, hybrid AI strategy that combines local and cloud AI to optimize cost, performance, security, and scalability.
The greatest risk is falling behind in organizational capability. As competitors embed AI agents into everyday operations, they could gain advantages in how work is performed, how decisions are made and how employees collaborate with AI. Organizations that delay adoption risk losing valuable time to develop the skills, governance frameworks and operational experience needed to deploy AI effectively at scale.
That doesn't mean organizations should deploy AI agents without a clear strategy. Successful adoption requires modern infrastructure, trusted data, strong governance and an open, flexible AI platform that can evolve alongside changing business needs. Enterprises that begin building these foundations today will be better positioned to scale AI confidently and capture long-term business value.
Why are AI PCs increasingly viewed as AI infrastructure investments rather than just end-user devices?
AI PCs are becoming an increasingly important layer of enterprise AI infrastructure as we approach the era of personal AI – where intelligence moves closer to the user, understands their context and expands what they can imagine and achieve.
As smaller, increasingly capable AI models continue to improve and become practical to run locally, work that once required a cloud subscription can increasingly happen directly on an AI PC. Tasks such as document analysis, coding assistance, knowledge retrieval and workflow automation can now run locally, giving enterprises the ability to own and use the compute rather than paying a recurring per-token cloud charge for every interaction. This can help reduce cloud inference costs while providing lower latency and greater control over where sensitive data is processed.
AI PCs also play a role in agentic AI. Local systems with sufficient memory and compute can run models that support tool calling, context-rich workflows and multi-agent applications, helping enterprises reduce the amount of AI inference they need to send to cloud services for appropriate workloads. This creates an opportunity to think about AI infrastructure in terms of token economics as well as compute performance: workloads that are frequent, latency-sensitive or economically attractive to run locally may benefit from local execution, while larger or highly elastic workloads can continue to use cloud infrastructure.
Why are skyrocketing costs causing AI compute to be increasingly distributed across cloud, on-premises infrastructure, edge, and AI PCs?
Cost is an important consideration, but AI compute is becoming more distributed because no single environment is optimal for every workload. Enterprises should assess latency, data sensitivity and residency, compute scale, elasticity, connectivity, recurring inference costs, governance and operational control before deciding where a workload belongs.
Cloud infrastructure remains well suited to frontier models, large-scale training and highly elastic demand. On-premises systems can offer greater control over enterprise data, while edge systems and AI PCs are often better suited to real-time, always-on or privacy-sensitive processing. These decisions should be considered as models, economics and business requirements change.
This makes the choice of technology partner important. Enterprises need an AI solutions provider with a broad portfolio and consistent software foundation that can support workloads across cloud, data centre, edge and client environments without forcing every application into a single architecture or deployment model.
AMD delivers an open, full-stack AI platform that brings together leadership compute, open software and deep ecosystem partnerships, giving customers the flexibility and choice to deploy the right compute for every workload as performance, cost, security and governance requirements evolve.
How is AMD working with its partners like Dell, HP and Lenovo, especially when it comes to innovation in AI PCs?
Our partnerships with leading PC manufacturers are critical to bringing AI PCs to a broad range of users. We work closely across the ecosystem to combine AMD compute and AI capabilities with our partners’ expertise in system design, software and user experiences.
With HP and Lenovo, for example, we’re bringing the latest AMD Ryzen™ AI Max PRO 400 Series processors to high-performance commercial workstations and small form-factor systems, including the HP ZBook Ultra G3a and Lenovo ThinkCentre X Ultra. With Dell, our collaboration now spans a growing portfolio of commercial PCs and mobile workstations powered by AMD Ryzen AI PRO processors, including the Dell Pro Max 14 and 16, giving businesses more choice for AI-enabled professional workloads.
However, innovation in AI PCs is about more than the processor. It requires hardware, software and applications to work together so users can take advantage of AI locally. AMD is investing across that ecosystem expanding the capabilities of ROCm™ and Ryzen AI Software, Lemonade, while AMD AI Playbooks help developers build, optimize and deploy local AI applications fast.
We are also working with a growing number of software partners such as Perplexity to expand the experiences that can take advantage of local AI compute. Perplexity Portable Computer, for example, is now available on supported Windows PC powered by AMD Ryzen AI Max Series processors, including the Ryzen AI Halo developer platform, bringing agentic AI experiences closer to the user. HP is also preloading Perplexity Computer on the ZBook Ultra G3a, showing how our OEM and software partners can come together to deliver new agentic AI experiences on AMD platforms.
Our goal is to give PC partners the flexibility to differentiate their systems while giving customers greater choice in how they adopt and use AI PCs.
How is AMD looking to respond to NVIDIA RTX Spark in the PC market?
AMD has been advancing this category for several years. We were the first to bring a dedicated AI engine to x86 Windows PCs in 2023, and in 2025 we introduced Agentic PCs with the AMD Ryzen AI Max 300 Series with systems capable of running AI agents locally. Today, with more than 50 Agentic PC designs in market, we continue to expand platform availability and optionality.
We are also taking performance even further with the latest Ryzen AI Max PRO 400 Series which supports up to 192GB of system memory, and is the world's first x86 client processor capable of running models with more than 300B parameters at 4-bit quantization, with no need for cloud offload. That additional capacity becomes increasingly important for agentic workloads, where multiple models and tools may need to run concurrently alongside everyday applications.
We believe Agentic PC platforms should be evaluated on more than peak AI performance. Agentic computing is a system workload, so what matters is how the CPU, GPU, memory and software work together to complete real-world tasks, while still providing the application compatibility, productivity and responsiveness users expect from a PC.
Software is equally important to that experience. AMD ROCm provides an open software foundation for AI development, with support for widely used frameworks and tools such as PyTorch, ComfyUI and LM Studio. Together with AMD AI Playbooks and a growing ecosystem that includes Liquid AI and Perplexity, we’re focused on making it easier for developers to move from selecting a model to building, optimizing and deploying useful AI applications.
That is also why we don’t think the software conversation should be reduced to a comparison between individual programming platforms. What matters is the complete developer journey - how easily developers can use familiar models and frameworks, get productive, optimize their workloads and ultimately deploy applications across the environments that make sense for them. AMD’s open approach is designed to give developers that flexibility and choice.







