Webcast
Hear the full discussion on egress fees, data mobility, and why storage design is being rewritten.
If you run a data-intensive business — AI, SaaS, media, analytics — you’ve likely felt it: egress fees quietly dictate how teams build, store, and ship data.
In our recent conversation with Backblaze VP of Sales Nilay Patel, he put it bluntly:
“Amazon charges these incredibly high fees to get your data out… they’re just crazy, egregious fees.”
The hyperscaler model hasn’t merely made storage expensive — it has made using your own data expensive. Which leads to a bigger issue: companies architect around avoiding egress, not around what their workloads actually need. Cold tiers, fractured pipelines, delayed migrations, single-cloud lock-in — most of these patterns stem directly from a fear of the “egress tax.”
Patel gave a concrete example that captures the absurdity:
“They charge you about $90 per terabyte to access your data once.”
Run that math on any sizable workflow — video delivery, AI training sets, nightly ETL jobs — and the consequences are obvious:
One of the stories Patel shared shows how extreme this gets:
“A media company needed to re-index assets after changing their media management system… it took so long and cost so much to pull the footage from Glacier that they said, ‘We are never doing that again.’”
That’s not an engineering decision. That’s a penalty.
Cold storage emerged for archival-era problems. Today’s reality is different:
Patel captured how quickly this has shifted:
“There’s all this data that maybe was thought to not be valuable anymore. But now suddenly it is — and you need access to it.”
AI and multimodal training pipelines flipped the script: archived data became active data overnight. And cold tiers, with their retrieval delays and egress penalties, are architecturally incompatible with that shift.
The podcast dives deep into the economics of why hyperscalers price egress so aggressively. Patel doesn’t hedge:
“On the egress side, they are just charging you a tax. It’s not because they have some innovative way to make it cheaper… They want to lock you into using only their services.”
That economic model creates three distortions:
“One company had to shut off their free tier and delay engineering hires because AWS storage and egress became too expensive.”
Egress fees don’t just distort cloud bills — they distort roadmaps, product strategy, and growth.
A major theme of the episode: modern infrastructure is moving toward choice, not consolidation.
AI startups are using GPU clouds like CoreWeave, Lambda, Voltage Park. Media companies are pushing video through CDNs like Cloudflare, Fastly, Akamai. SaaS platforms are running analytics in Snowflake, BigQuery, Databricks.
Your storage can no longer assume a single downstream compute or delivery path. Patel captured the reality:
“One quarter of all our outbound traffic now goes to neo clouds. These companies didn’t exist a few years ago.”
Infrastructure is fragmenting — fast. And egress-heavy storage is fundamentally incompatible with that trend.
To support data mobility and high-throughput workloads, storage needs to be:
The next decade belongs to cloud architectures that are flexible, modular, and multi-cloud by default — not locked into a single provider’s economics.
In the webcast, we explore:
