Last week, David Heinemeier Hansson explained why his company, Basecamp, is leaving the cloud. They will manage their own servers to reduce the $3m annual AWS bill by 60%.

Setting aside the technical questions around such a migration for a paragraph or two, this move will improve Basecamp’s efficiency materially.

For a hypothetical startup, a 60% reduction in infrastructure costs boosts sales efficiency by 11% & net income margin by at least 12%. That means reducing the payback period from 21 to 18 months. These simple calculations ignore the additional margin improvements arising from depreciating the servers which should add a few more percentage points to the bottom line.

Basecamp’s 60% reduction in cloud COGS isn’t anomalous. A recent survey showed the typical company would halve its costs by managing its own servers.

We are in an era where profitability (net income margin) & sales efficiency constitute two of the top three correlates to valuation.

Reverting back to self-managed infrastructure might hold some appeal for companies focused on financial optimization rather than growth. I wouldn’t be surprised if private equity held software companies adopt this practice to generate extra income in the next few years to offset some of this era’s greater interest expenses on debt.

That optimization comes at some cost: hiring a team to migrate the software & to manage the infrastructure. In addition, the cloud offers burst-ability, the potential to serve a significant surge in traffic. Downtime implies a loss of revenue, which isn’t factored into the math above.

For many high-growth startups, the juicy 11% improvement in sales efficiency likely isn’t worth the squeeze. Inspecting & controlling cloud costs may be the more appealing path. But for slower growth businesses or those looking to optimize their efficiency, moving to self-hosted infrastructure may be worth considering.