AI vendors are abandoning traditional user-seat fees in favor of token-based and outcome-linked models, alongside a surge in AI-capable notebooks, reshaping enterprise technology economics and workflows.
AI vendors are rapidly moving away from the old software playbook. Instead of charging by user seat, many are now leaning on token-based billing, usage meters or outcome-linked contracts, a shift that reflects a simple reality: generative and agentic tools are no longer experimental add-ons, but products that vendors need to monetize more aggressively. That is likely to make enterprise AI significantly more expensive, even as it gives customers a clearer way to pay only when work is actually completed.
According to TechRadar Pro, Shashi Upadhyay, Zendesk’s president for products, engineering and AI, argued that software pricing should track customer success rather than headcount. The challenge, he said, is that outcome-based pricing depends on both sides agreeing on the result that triggers payment. That approach is already gaining traction across the industry as vendors look for ways to replace lossmaking adoption pricing with models that better reflect value delivered.
One response is the AI PC. TechInsights expects AI-capable notebook shipments to surge and says such systems could make up 95% of the notebook market by 2029, while Gartner and Omdia both see the category becoming mainstream as chipmakers add increasingly powerful neural processing units. TechRadar Pro noted that Microsoft’s Copilot+ class set a 40 TOPS benchmark, and newer designs from Intel, AMD and Qualcomm have pushed beyond 50 TOPS, with even higher ceilings now appearing. In practical terms, that means more summarization, transcription, image cleanup and other routine AI work can be handled locally instead of in the cloud.
That matters because cloud inference is becoming harder to predict and harder to budget for. Omdia research director Ishan Dutt told TechRadar Pro that the definition of an AI PC is still moving, and that raw TOPS alone no longer tells the whole story. Memory and graphics capacity now matter too, especially for creative and agentic workloads that run continuously in the background. In that sense, the AI PC is less a fixed product category than a shifting hardware threshold for which tasks stay on device and which are pushed into centralized compute.
Even so, the surge in AI-capable systems does not necessarily mean companies are abandoning cloud infrastructure. Omdia expects AI PC share of the overall PC market to climb sharply by 2030, but it also says the Windows 10 replacement cycle has inflated recent shipment figures. Gartner’s Ranjit Atwal has suggested that buyers will focus on machines that support specific business outcomes, while Dutt said clearer links between hardware upgrades and future on-device AI features could reshape refresh cycles.
The bigger constraint may be cost elsewhere in the supply chain. TechRadar Pro reported that memory makers are redirecting capacity toward high-bandwidth memory and high-capacity DDR5, creating a second chip shortage that is different from the pandemic-era one. Omdia says many channel partners now expect delays or cancellations in refresh plans, and Gartner sees weaker sales before a possible rebound later in the decade. Apple’s recent price increases on many Mac models underline how quickly hardware economics can shift.
The result is likely to be a hybrid future rather than a clean break from cloud computing. Local processing can absorb predictable, high-volume tasks and help firms regain control over spending, but cloud systems will still be needed for large models, shared data, centralized workflows and deep integrations. As Upadhyay told TechRadar Pro, the cloud still matters for multi-channel customer data, routing, reporting and API connections across hundreds of tools. The pricing war may be changing, but the cloud is not going away.
Disclaimer: This article may have been created with AI assistance and reviewed by our editorial team. It is provided for general informational purposes only. Readers should verify information independently before relying on this content.
