The Efficiency Paradox

Start with seven observations and follow them to their conclusion.

Operating system code optimised for performance produces faster computation or greater throughput on a given set of hardware. BSD has the oldest continuous lineage in modern computing and is among the most portable operating systems ever written. Longevity permits code optimisations to accumulate over time. Portability necessitates efficiency and correctness — you cannot hide sloppy assumptions behind a known hardware target when the same codebase must run on ARM, x86, RISC-V, and powerpc simultaneously. Hardware vendors want to sell more hardware. BSD extracts more compute from the same hardware.

None of these observations is contested. Together they describe a market distortion: the better the operating system, the less commercial incentive exists to fund its ecosystem. The efficient option is starved of investment not because it is inferior but precisely because it is superior. Efficiency suppresses the upgrade cycle. The upgrade cycle is where the revenue lives.

Economists have a name for the inverse of this dynamic. William Stanley Jevons observed in 1865 that improvements in steam engine efficiency increased total coal consumption — cheaper operation expanded adoption faster than it reduced consumption per unit. BSD inverts the Jevons paradox. Efficiency improvements here do not expand the market. They shrink the invoice. The hardware vendors receive no upgrade call from the organisation running BSD on decade-old infrastructure, because that organisation does not need to call.

This is not a complaint. It is a diagnosis. The treatment requires regulation.

Primary sources, not advocacy.

Rumen Palov is the CTO of E-Card, a Bulgarian online gaming company operating since 2000. Writing in November 2024, he documented twenty-five years of continuous FreeBSD deployment across infrastructure handling 100,000 queries and 20,000 transactions per second during peak load. The hardware runs multi-terabyte ZFS storage pools across database clusters with Galera replication. The infrastructure has scaled from the company's inception to present without an OS-driven hardware refresh cycle.

E-Card are not exceptional. They are what happens when an organisation applies the oldest principle in SME financial management — sweat your assets until they physically cannot perform — to infrastructure decisions. BSD's answer to the question "does it still work?" is, consistently, yes.

The second exhibit sits on a desk in Islington. A MacBook Air from 2013/2014, Intel Core i5-4260U Haswell ULT, 4GB RAM. Apple classified this hardware as vintage and withdrew support. It runs FreeBSD 15.0-RELEASE, Hyprland compositing on Wayland, ZFS root filesystem, battery at 100%.

The third exhibit is an HP Compaq nc6400 from 2008. Intel Core 2 Duo. Intel GMA950 graphics — a chip so old it predates the assumption that integrated graphics warranted serious driver investment. It runs OpenBSD 7.8. It streams BBC iPlayer without dropped frames. It is seventeen years old.

The efficiency paradox would be merely interesting if it existed in isolation. It does not. It intersects with three distinct sovereignty failures that European organisations are now confronting simultaneously.

The e-waste failure. Windows 10 reaches end-of-life in October 2025. Tens of millions of affected devices across the United Kingdom alone, a substantial fraction running on hardware that Windows 11's TPM and CPU requirements exclude. That hardware runs FreeBSD and OpenBSD without complaint. The nc6400 is seventeen years old and streaming HD video. The externality is real and unpriced: when an OS vendor's EOL policy forces hardware retirement that the hardware does not require, the e-waste cost is socialised. The vendor bears none of it.

The legal sovereignty failure. The CLOUD Act authorises US government demands for data held by US companies regardless of physical server location. A European organisation's data on AWS Frankfurt is subject to US legal process. FreeBSD is developed under the auspices of the FreeBSD Foundation, a Colorado public benefit corporation. OpenBSD is maintained by the OpenBSD Foundation, an Alberta charitable foundation. Neither provides services. Neither holds customer data. Neither has a subpoena surface. The BSD licence is geopolitically neutral in a way that no hyperscaler's terms of service can be.

The supply chain failure. An organisation running BSD on existing hardware is not purchasing new silicon on a vendor-driven refresh cycle. E-Card's infrastructure has been scaling for twenty-five years without OS-mandated hardware replacement. In April 2026, with Hormuz closed and component supply chains disrupted, the distinction between organisations that sweat their assets and organisations that refresh on schedule has a monetary value that was previously theoretical.

The European Union does not need to construct new institutions to address this. The instruments exist. What is missing is the framing that connects them.

The Sovereign Tech Fund — a German federal government initiative — already funds BSD networking infrastructure. Its published investment rationale describes the BSD networking stack as critical digital infrastructure. JunOS runs on BSD. Netflix's content delivery network runs on FreeBSD.

The Alpha-Omega Project — funded by Microsoft, Google, and Amazon — commissioned Synacktiv to conduct a comprehensive audit of FreeBSD's bhyve hypervisor and Capsicum sandboxing framework. The FreeBSD Foundation published the full findings in November 2024. The audit is public. The methodology is documented. This is what software supply chain security looks like when the development culture treats transparency as an architectural property.

The efficiency paradox has a downstream consequence that none of the regulatory instruments yet addresses.

Hardware that is not manufactured does not require minerals. The minerals in computing hardware — tantalum, tungsten, tin, gold, cobalt — come predominantly from supply chains with documented connections to conflict-affected and high-risk areas. The Democratic Republic of Congo supplies a substantial fraction of the world's cobalt and coltan. The human rights record of those supply chains is not a matter of dispute. It is a matter of published documentation, ICC proceedings, and the IPC-1755 conflict minerals standard that the industry adopted precisely because the problem was undeniable — and then continued to operate within anyway.

The hardware refresh cycle that BSD interrupts is not merely an e-waste mechanism. It is a demand signal. Every three-year enterprise refresh cycle, every consumer device replaced because the OS vendor withdrew support from functional hardware — each of these generates a purchase order that travels upstream through a supply chain and eventually reaches a mine.

The organisation that sweated its assets did not fund the mine. E-Card's twenty-five years on the same OS architecture is twenty-five years of attenuated demand for new extraction. The nc6400, seventeen years old, streaming iPlayer on OpenBSD, represents a device that did not enter the replacement market and did not generate the upstream demand that replacement would have required.

See also: Clean Slate — existing hardware, new capability →