ChatPPT, an AI tool that generates presentation slides in seconds, has cut its cloud computing costs by more than 50% by shifting some of its processing to run directly on users' personal computers instead of relying entirely on remote servers. The collaboration between ChatPPT and Intel demonstrates a growing shift in how companies are rethinking where AI actually runs, balancing the power of cloud computing with the privacy and cost benefits of local processing.

Why Companies Are Moving AI Off the Cloud?

ChatPPT's original approach processed 100% of its work in the cloud, which created three significant problems. First, cloud-based processing generates high energy costs and token expenses, which add up quickly when serving thousands of users. Second, sending sensitive corporate data to cloud servers raises security risks. Third, enterprise customers increasingly worry about proprietary information being shared with external cloud-based AI models.

These concerns aren't unique to ChatPPT. Across the industry, companies are recognizing that not every AI task needs the full power of a distant data center. Some workloads, like simple formatting changes or minor edits, can run efficiently on a user's own device without sacrificing quality or speed.

How Does Hybrid AI Split the Work Between Cloud and Device?

Intel's solution, called AI Super Builder V2.8, enables what's known as hybrid AI, a system that intelligently divides tasks between local and cloud processing. ChatPPT's new Intel AI PC Edition uses this approach to handle different types of work in different places.

• Complex Tasks on Cloud: Large workloads like generating 50-page presentation decks still run on cloud servers, where massive computing power is available and cost-effective for heavy lifting.
• Simple Tasks on Device: Smaller operations like changing font size, adjusting colors, or reformatting slides now run directly on the user's computer, eliminating unnecessary cloud calls.
• Data Privacy Locally: Sensitive corporate information stays on the user's device and never travels to external servers, addressing enterprise security concerns.

To make this work, Intel's engineering team spent months collaborating with ChatPPT to solve two critical technical challenges. First, they needed to deploy complex workflows that could run entirely on a local PC. Second, they had to figure out how to run multiple AI models simultaneously, like planning, generation, and fact-checking models, without overwhelming the computer's resources.

"Intel's AI Super Builder provided a complete on-device inference framework, allowing us to quickly deploy both the Logiliner traceability model and our document generation agent locally, achieving a complete on-device loop from content planning to final output," explained Jack Zhou, CEO of ChatPPT.

Jack Zhou, CEO, ChatPPT

The teams used a technology called OpenVINO to compress the AI models, making them smaller and faster without losing accuracy. This compression allowed multiple models to run smoothly on a standard PC without slowing down the system.

What Are the Real-World Results?

The numbers tell a compelling story. ChatPPT's Intel AI PC Edition reduced overall cloud compute token costs by more than 50% compared to the cloud-only version. At the same time, users could run the tool for over 32% longer before needing to switch back to cloud processing. For a company paying for cloud AI services, cutting costs in half while improving user experience represents a significant win.

These improvements matter because cloud AI services charge based on tokens, a unit of text that the AI processes. A single request might consume hundreds or thousands of tokens, and costs add up quickly across an organization. By handling routine tasks locally, ChatPPT dramatically reduced the number of tokens flowing to cloud servers.

What's Next for Local AI?

ChatPPT and Intel aren't stopping here. The companies are working toward an ambitious next milestone: moving document rendering, the final step that converts content into formatted slides, entirely to the user's device. This would create a truly 100%-local workflow for document creation, eliminating cloud dependency altogether for that function.

Beyond cost and performance improvements, the partnership is enabling new capabilities. ChatPPT plans to leverage Intel's multimodal support framework to expand into image-text composition and intelligent chart generation. The company also plans to create specialized versions tailored to specific industries, including academia, business analysis, education, and financial services.

"Intel is not just a technology partner, but a key enabler in our journey to deliver secure, efficient, and cost-effective on-device intelligent creation," stated Jack Zhou.

Jack Zhou, CEO, ChatPPT

The ChatPPT and Intel partnership reflects a broader industry trend. As AI becomes more embedded in everyday tools, companies are discovering that the cloud-only model isn't always the best approach. Local processing offers faster response times, better privacy, lower costs, and reduced dependence on internet connectivity. The hybrid approach, which uses both local and cloud resources strategically, appears to be the emerging standard for how enterprise AI will operate.

This shift has implications beyond presentation software. The same principles apply to image editing, document analysis, coding assistance, and countless other AI-powered tasks. As devices become more capable and AI models become more efficient, expect to see more applications adopting this hybrid model, keeping sensitive work local while leveraging the cloud for tasks that truly benefit from its scale.