A new generation of AI-capable laptops, desktops, workstations, and servers will hit the market this year, promising to change what we know about computing devices. IT asset disposition (ITAD) professionals will also need to prepare for a wave of hardware unlike anything they’ve processed before. We looked at a number of new devices announced in March 2025, like the NVIDIA DGX Spark, Dell Pro Max with GB300, HP ZBook Fury G1i, and Lenovo’s AI-enhanced ThinkBook series, and we saw a significant leap in performance, architectural complexity and data-bearing potential. Within four to five years, these systems will begin entering decommissioning pipelines. And when they do, ITAD companies will need to rethink everything from teardown to resale strategy. A large part of the future problems that ITAD will experience will clearly come from lack of OEM and manufacturer attention, as evidence in points 1 through 3 below.

If you are a recycler or an ITAD processor, here are ten things we began to identify as game changers:

1. Embedded AI Components Will Complicate Teardowns
   Devices will contain neural processing units (NPUs) and AI accelerators deeply integrated into CPUs and motherboards, making disassembly and value extraction more difficult.
2. Modular Parts Will Be Rare
   Expect more soldered-in components—RAM, SSDs, co-processors—reducing the ease of part harvesting and increasing labor time for disassembly.
3. AI Modules May Contain Residual Data
   Sanitization will need to go beyond wiping drives. ITADs must develop methods to purge AI model caches, learned states, and user-bound profiles from embedded silicon.
4. Firmware and Identity-Tied AI Software Will Limit Resale Value
   Many AI features are tied to enterprise licenses. Without them, devices lose functionality, pushing ITADs to reimage systems with clean, license-free environments.
5. Disassembly Time Will Increase Without OEM Guidance
   OEMs aren’t prioritizing reusability. Without teardown guides, ITAD teams will need more time and expertise to safely dismantle AI devices.
6. New Regulatory Pressures Are Likely
   Expect future compliance frameworks requiring proof of AI state data sanitization—including documentation for firmware resets and AI module wipes.
7. Energy Use Will Affect Resale Markets
   AI-capable systems may have higher idle power draw, which could deter resale in energy-conscious markets unless devices are optimized or AI features are disabled.
8. Testing AI Capabilities Will Be Crucial for Remarketing
   Devices like DGX Spark and Pro Max GB300 will hold resale value only if their AI performance can be tested, benchmarked, and certified.
9. Value Will Shift from Software to Raw Hardware
   As AI tools become subscription-based, the resale proposition will depend more on compute capability than on bundled AI experiences.
10. Opportunities Will Mirror the GPU Mining Boom
    Certain AI-enabled devices will attract niche markets—startups, research labs, universities—offering high-margin resale opportunities for those prepared to identify and serve them.

Additional Thoughts:

The machines we reviewed in the month of March 2025 and the thousands more to come are not simply faster or thinner versions of their predecessors. They are functionally different, largely because of the inclusion of dedicated neural processing units (NPUs), AI accelerators, and firmware-integrated software agents optimized for generative inference, video processing, or predictive workloads. These components will be very different than the ones that ITADs have always known. The fact that they are intricately embedded into the CPU or motherboard, will undoubtedly present operational and practical challenges for recyclers and refurbishers attempting to extract value—or even to perform secure data sanitization. This situation could lead to friction with OEMs, and may even prompt legislators to add new rules in their produce responsibility laws.

One of the biggest problems I see is the vanishing of easily harvestable modular components. What used to be straight forward and well-documented processes of harvesting part, the situation has shifted in favor of components like RAM, AI co-processors, and even SSDs being soldered in. Devices like the Lenovo ThinkBook Flip and Dell Rugged 14 may offer hot-swappable drives or batteries, but they may be the exception as others—especially workstations and high-performance laptops—lean toward monolithic boards as single blocks designed for efficiency, not for reuse. The results of this trend I anticipate is that part harvesting will become more labor-intensive, and device disassembly times will increase unless OEMs begin to offer model-level teardown guides. The cost of operating an ITAD service will increase as a result, and tensions with OEMs could escalate.

This ITAD’s anticipated problem is not the only pain I anticipate. AI-enhanced silicon will create a more uncertain data security environment as AI will store context, learn locally, and will interact with identity-bound software environments. Even if the NPU doesn’t permanently store data in a traditional sense, future regulatory frameworks may put ITAD providers in a difficult position, with laws likely requiring them to account for temporary AI state data, model cache, or user-specific AI profiles. Sanitization will no longer be a matter of wiping the SSD. It may require firmware resets, AI module purging, and a documented process to prove that no learned data remains resident in the system.

ITAD companies and their OEM partners will have to figure out how to protect the secondary market from the anticipated development in the second. One of the issues we should anticipate in software licensing, as AI tools are increasingly tied to enterprise identity or cloud subscriptions—such as Microsoft Copilot, Adobe Firefly, or proprietary AI assistants. We anticipate that without the enterprise cloud license, much of the “AI” value vanishes, making resale a difficult goal to achieve, without requiring refurbishers to reimage machines with neutral, license-free operating environments.

Besides these challenges, many of the AI systems will provide high-margin opportunities. Two of the nine systems we checked in March, the NVIDIA DGX Spark and Dell Pro Max GB300, can run large models locally, and may command a strong secondary market interest vertical sector in need of high-performance systems. These machines, just like in happened during the GPU mining boom, could remain valuable for years—if ITAD companies can test and certify their AI inference performance.