Switzerland ships 25 Gbit fiber for $65/mo. The US market failed on purpose.

What happened

Stefan Schueller's post *The Free Market Lie* hit 117 on Hacker News with a claim that is hard to argue with once you see the invoice: Init7, a mid-sized Swiss ISP, delivers symmetric 25 Gbit/s fiber to the home for CHF 64/month — about $71. That's not a promo rate, not a business tier, not an enterprise circuit with a three-year contract. It's the residential product. Upload equals download. No data cap.

Meanwhile the median US household in 2026 pays roughly $80/month for asymmetric 300/20 Mbit cable from a regional monopoly, and considers themselves lucky if the actual throughput clears half of that during evening hours. The gap between Swiss and US residential internet is now roughly two orders of magnitude on upload, at the same price point.

Schueller's argument, delivered with the tone of somebody who has explained this at one too many dinner parties, is that the standard American framing — *Switzerland is small, dense, socialist; the US is big, sprawling, capitalist; that's why we have worse internet* — gets the causation exactly backwards. Switzerland isn't uniformly dense (try laying fiber to a chalet at 1,800 meters). And the US wasn't outcompeted on infrastructure. It was actively prevented from building it.

Why it matters

The mechanism is well-documented and boring, which is why nobody talks about it. In 2005, the FCC reclassified broadband from a Title II common carrier to a Title I information service. The practical effect: incumbent cable and telco companies were no longer required to offer wholesale access to their lines. If Comcast owned the last mile, no competitor could rent it. If you wanted to sell internet to a house on that street, you had to trench your own fiber past every other operator's already-trenched fiber. Nobody does that. It's not a market. It's a set of overlapping monopolies with a shared interest in not building anything.

Switzerland took the opposite path. Swisscom, the incumbent, is required by the regulator (ComCom) to lease dark fiber to any operator that asks, at a regulated price. Init7 doesn't own the fiber running to Schueller's house — it rents it, lights it with its own optics, and competes on service. That single policy choice is what lets a ~150-employee ISP undercut the incumbent's own gigabit plan by ten to one on speed and still make money.

The HN thread predictably splits into three camps. The libertarian read is that Switzerland's model is a subsidy in disguise — Swisscom built the fiber under a state-adjacent monopoly, then was forced to share it, so of course prices are low. The counter, offered by several commenters with actual infra experience, is that the US had exactly the same setup until 2005 and threw it away. Verizon FiOS was built during the brief window when unbundling was still on the table; the moment that window closed, FiOS expansion also closed. The third camp, mostly former ILEC employees, points out that the last-mile CapEx has already been recovered several times over from the Universal Service Fund and the various rural broadband subsidies of the last twenty years. American households have already paid for fiber to the home. They just haven't received it.

What's changed in 2026 is that the excuse has run out. Starlink demonstrated that even rural coverage — the traditional trump card for US telecom — is a solved problem at $120/month and 200 Mbit down. The remaining defense of the current market structure is essentially aesthetic: we don't want the regulator involved. Fine. The regulator was involved in the outcome anyway; it just chose the incumbents' side.

What this means for your stack

If you build anything that involves uploading — CI artifacts, container images, video, ML training data, backups, self-hosted services — your ceiling in the US is not set by your ISP's technical capability. It's set by a policy from the second Bush administration that nobody has re-litigated. A Swiss dev pushing a 40 GB container image does it in 13 seconds on the residential plan; the US equivalent takes about 45 minutes on a 'business' fiber circuit that costs six times as much.

The practical implications: if your architecture assumes that heavy upload happens in a datacenter because home connections can't handle it, that assumption is a US-specific artifact, not a law of physics. Teams distributed across Europe and Asia routinely run substantial infra from residential connections — not because they're clever, but because their upload is fast enough to make the tradeoff irrelevant. If you're evaluating self-hosted AI inference, edge deployments, or moving CI closer to developers, the upload ceiling is the load-bearing assumption. Check it before you design around it.

Second, if you're thinking about relocating — and post-remote a lot of senior engineers are — the residential internet delta is a real quality-of-life variable, not a hobbyist concern. The gap between 25 Gbit symmetric and 300/20 asymmetric is the difference between 'my home is a viable compute node' and 'my home is a thin client for AWS'.

Looking ahead

The honest read is that US residential internet is not going to improve through market forces, because the market was structurally prevented from existing. It will improve when — if — the FCC reverses the 2005 reclassification and reinstates line-sharing, or when a state does it unilaterally and survives the inevitable lawsuit. Neither is on the 2026 policy calendar. In the meantime, the Swiss invoice is a useful artifact to keep in your back pocket the next time somebody at your company insists that fiber to the home is 'just too expensive to build here'. It isn't. It's been built. Somebody just isn't lighting it.