A disconnect between technical capability and end-user adoption is driving current decision-making in AI and infrastructure among managed service providers and IT leaders. While companies like Broadcom, Perplexity, Apple, and NVIDIA are introducing hardware and platforms for on-premises and local AI processing, the primary lever for value remains organizational adoption rather than available technology. Costs and infrastructure investments are being shaped by how—and whether—users engage with these tools.
The most significant evidence is from Voya Financial, where a five-hour AI training for 11,000 employees led to 98% adoption and frequent use, despite no change in underlying technology. Industry data cited shows that while generative AI now reaches 80% of occupations, only 2.8% of tasks have more than 50% usage, and none reach 70%. This indicates broad awareness but little real transfer of work to AI, making current spending primarily reflective of shallow adoption rather than deep operational change.
Additional developments include a clear industry shift toward privately owned AI infrastructure, as shown by recent product launches and by NVIDIA’s ongoing focus on cloud-rented compute. Gartner projects that consumption-based and inference-specific storage will expand rapidly by 2029, but these estimates are based on limited current adoption, underlining a mismatch between infrastructure readiness and immediate need.
For MSPs and IT providers, the risk is overcommitting to AI infrastructure before clients are ready. Real opportunity lies in structured adoption programs for existing AI tools—identifying staff with teaching skills, piloting training internally, and focusing on increasing client usage. Deferring on-premises AI infrastructure projects until client adoption and economic signals align is positioned as a safer, more pragmatic approach.
24:54
13:42
40:34
13:20
14:04