The lede: Washington's cloud antitrust push has a narrative problem it hasn't noticed yet. The FTC's 2023 cloud RFI (docket FTC-2023-0028) and the parallel UK CMA cloud services market investigation — which issued its final decision on July 31, 2025, finding that egress fees and technical switching barriers "harm competition" — are widely read as an AWS-and-Azure story. But the mechanism regulators are actually chasing is enterprise lock-in, and the purest remaining version of that lock-in isn't egress fees. It's software licensing. Oracle already conceded the egress-fee fight: in February 2026 it eliminated outbound data-transfer charges globally across all 48 OCI regions, marketing itself as the anti-lock-in cloud. That leaves Oracle's actual moat — Oracle Database licensing terms that make running on OCI (or Oracle Cloud@Customer) dramatically cheaper than running Oracle software anywhere else — as the next-most-obvious target if the FTC/DOJ or a UK-style "Strategic Market Status" regime pivots from egress fees to licensing-driven interoperability mandates. A rule requiring vendor-neutral licensing terms across clouds would hit Oracle's core database franchise harder than it hits AWS or Azure's infrastructure margins.
Who cashes in: DELL and SMCI benefit from any policy current that pushes workloads back toward on-prem or neutral-hardware infrastructure rather than proprietary-licensed cloud stacks — both sell the boxes regardless of which cloud (or no cloud) runs on top. ANET gains as enterprises hedge multi-cloud networking to avoid re-concentrating in any single stack, since Arista's switching fabric is cloud-agnostic by design and multi-cloud architectures require more, not less, networking gear. AVGO is largely insulated and modestly favored — its VMware and custom-silicon franchises profit from enterprises diversifying workloads across clouds rather than committing to one vendor's licensing terms.