The Promise and the Problem
AWS re:Invent 2025 delivered what we’ve come to expect: aggressive pricing announcements wrapped in competitive language. Amazon cut S3 rates further and struck new zero-egress deals with select CDN partners. On stage, it looked decisive. In practice, it’s more complicated.
Here’s what actually happened. AWS responded to Google Cloud and Azure pushing harder on their own pricing. Google announced their Cross-Cloud Network at Cloud Next 2025, positioning themselves as the interoperability play. Azure kept the pressure steady. So Amazon moved. But moves in cloud pricing rarely mean what the headlines say they mean.
The real issue isn’t the headline discount. It’s the architectural reality underneath. Most enterprises I talk to run multi-cloud by necessity, not choice. Legacy workloads live on-premises or AWS. New stuff goes to Google Cloud. Some teams standardized on Azure years ago. That’s the actual strategic picture. Pricing announcements don’t solve the data movement problem that haunts those setups.
Where Egress Actually Kills Your Budget
Let me be direct: egress fees are where cloud pricing becomes a cost center instead of a cost control. According to Cloudflare’s 2025 Bandwidth Alliance data, enterprises moving between major cloud providers still pay $0.08 to $0.09 per GB for high-volume transfers outside formal alliance agreements. That’s real money at scale.
A petabyte of data moved between providers costs roughly $80,000 to $90,000. Most people don’t run a petabyte operation, but plenty do. Machine learning teams doing model retraining across clouds. Data warehouses being migrated. Disaster recovery failovers. Those scenarios happen every quarter in mature enterprises.
The zero-egress deals AWS now offers? They’re real, but they’re narrow. They apply to specific CDN partners under specific conditions. If you’re using Cloudflare or another partner in the alliance, you benefit. If your architecture doesn’t fit that pattern, the old pricing sticks around. That’s the contract fine print that matters more than the press release.
Multi-Cloud Governance as Fiction
The numbers from the Flexera 2026 State of the Cloud Report paint a stark picture. Eighty-nine percent of enterprises run multi-cloud strategies. Nearly universal. But only 28 percent have mature cost governance tools that work across all their providers. That’s the real crisis.
Without governance, you’re flying blind. You don’t know which workload lives where, what it costs, or whether it should move. Teams provision duplicated infrastructure because visibility is fragmented. A database team in AWS doesn’t know what the data engineering team runs in Google Cloud. Everyone optimizes locally. Globally, you hemorrhage money.
Gartner’s 2025 Cloud Cost Optimization report estimated that 35 percent of enterprise cloud spend is wasted. Multi-cloud networking costs are a growing piece of that waste. You’re not just paying for egress. You’re paying for duplicate resources, suboptimal instance selection, and data that gets stored in three places because nobody knows where to find it.
The governance problem isn’t a software problem. It’s an organizational one. You need someone with authority to say “this workload moves here” and make it stick. Most enterprises don’t have that person until the CFO forces the issue.
Google Cloud’s Inter-Cloud Promise and Its Limits
Google’s Cross-Cloud Network announcement was clever positioning. The idea is attractive: simplified connectivity between clouds without the egress tax. But there’s a structural catch that matters.
The system requires workloads to run on supported regions. Not all regions are supported. Not all workload types fit the model. If your legacy application runs on AWS us-east-1 and your new data platform lives on Google Cloud asia-southeast1, the simplified connectivity doesn’t help. You’re back to standard routing and standard fees.
This is worth examining carefully because it reveals something fundamental about cloud architecture. Vendors want to solve inter-cloud movement at the infrastructure layer. The market wants it solved at the workload and data layer. Those are different problems. Infrastructure connectivity doesn’t actually reduce your egress costs if your workloads aren’t positioned to use the connected paths.
Building for Real Multi-Cloud Economics
Here’s what I tell teams planning their multi-cloud strategy for 2026. First, accept that multi-cloud will cost more than single-cloud. Don’t build your business case assuming it won’t. The networking overhead is real. Budget for it.
Second, be specific about why you’re multi-cloud. “Avoid vendor lock-in” is a reason, but it’s not specific enough to build an architecture around. “Run machine learning on Google Cloud because they have better TPU availability, but keep our operational systems on AWS” is specific. That specificity drives better architectural decisions and tighter cost controls.
Third, implement governance before you need it. Pick a tool. Could be a vendor solution. Could be internal dashboards built on cloud billing APIs. Doesn’t matter which, but pick something and make teams use it. The 89 percent of enterprises running multi-cloud but only 28 percent with mature governance are the ones writing the biggest checks.
Fourth, understand the egress landscape deeply. Check the AWS data transfer pricing breakdown. Model your actual data flows. If you’re moving petabytes between clouds, negotiate. Most vendors will cut deals on predictable, high-volume egress. If you’re moving gigabytes, accept the fee and optimize the workload instead.
The vendors won’t advertise this, but it’s true: cloud pricing is a negotiation for large customers. The advertised prices are real. The deals underneath are better if you know what to ask for. Don’t leave that money on the table.
Multi-cloud works. I’ve built it at scale. But it works because of architecture and governance, not because of pricing. The announcements at re:Invent and Cloud Next matter less than the decisions you make about where data lives and how often it moves. Those decisions compound over time. Get them right early and your 2026 cloud bills will reflect it. Get them wrong, and no pricing discount will save you.
What’s your multi-cloud setup look like? Where do egress fees hit hardest in your organization? I’m interested in the patterns people are seeing in production environments.
