Cloud Egress Costs vs. Serverless Efficiency: A Trade-Off Dilemma?

I've been thinking a lot about the push-pull between the cost savings often touted with serverless architectures and the sometimes overlooked, but significant, impact of cloud egress charges. While serverless can dramatically reduce operational overhead and scale almost infinitely, large data transfers, especially between regions or out to the internet, can quickly erode those savings. Are we seeing a point where egress costs become the primary inhibitor for certain serverless-heavy designs, particularly for data-intensive applications? How do you factor these competing forces into your early architectural decisions? What strategies have you found effective for minimizing egress while still benefiting from serverless agility?