Dayton, OH Correspondent — Training frontier models requires capital, data, and compute that few organisations command.

That creates economic risk if critical capability sits with a handful of private entities, and political risk if they become gatekeepers.

The capital scale is the new fact this year. The committed spending on data centres and related infrastructure by a few firms now rivals a national infrastructure programme, and a substantial part of it is financed by those same firms investing in each other’s customers.

That’s a circular structure and circular structures are how concentration becomes fragility.

Orange County, CA Correspondent — Frontier capability concentrated in a few firms creates single points of failure and outsized leverage.

The financial architecture deserves more scrutiny than the antitrust question. Chip suppliers investing in model developers who commit to buying chips, cloud providers taking equity in firms who commit to buying cloud — those are related-party transactions at enormous scale and the revenue recognition is genuinely difficult to assess from outside.

I’m not predicting a collapse. I’m saying the exposure is concentrated in a way that isn’t visible in any single company’s filings, and that’s the definition of systemic.

Novi, MI Correspondent — When a few companies build the most advanced systems, dependence on them is a structural vulnerability.

The manufacturing version is that we’re being sold tools we cannot evaluate, from vendors we cannot switch away from, on terms that change annually.

That’s a procurement problem before it’s a policy problem and every mid-sized firm in America is about to have it.

Sheffield, Jamaica Correspondent — Concentration at the frontier creates systemic risk because the effects aren’t confined to the firms.

Smaller economies become dependent on capabilities they cannot reproduce and rules they had no part in writing.

It doesn’t follow that every country needs a frontier model. It does follow that access, interoperability, and alternatives matter. A world in which a few companies mediate a growing share of knowledge work deserves the scrutiny we apply to other concentrated infrastructure, and my region will receive whatever regime three governments negotiate.

Jacksonville, FL Correspondent — When a few firms control the compute, the models, and much of the distribution, the risk is operational as well as economic.

A port worries about single points of failure because one interruption stops a chain of downstream activity, and this is developing the same character.

I wouldn’t nationalise the systems. That replaces several concentrations with one. The objective is a market where dependence can be shifted when a provider fails or abuses its position.

Las Vegas, NV Correspondent — Systemic risk is when one vendor’s problem becomes everyone’s operating problem.

We already depend on outside systems for reservations, payments, staffing, and security. If this becomes another layer underneath all of them, concentration matters even when the price looks competitive.

The danger isn’t that a company is large. It’s that leaving becomes impractical.

Cheyenne, WY Correspondent — If the rule needs a tower of lawyers, the giants win.

Write a card, not a tower. Then a small shop can still try.

Knoxville, TN Correspondent — Controllers will come from a few fabs and a few models. Applications can come from a shop that knows a county.

Keep the shop legal without a senator. That’s the opening.

The power question is the one that reaches my desk. Data centre load is now a live issue in utility planning here, and residential ratepayers are being asked to fund transmission for it. That’s a concentration effect arriving as an electricity bill.

Prescott Valley, AZ Correspondent — The bottleneck is compute and the supplier of the hardware has effective control over who competes.

Nothing in the current policy discussion addresses that, because addressing it would mean regulating the most valuable company in the world at the point of its advantage.

Any policy serious about concentration starts there. None of them do.

Long Island, NY Correspondent — The liability question remains unresolved and it’s the one that would actually discipline this.

When a model produces a harmful output, responsibility among developer, deployer, and fine-tuner is unsettled, and product liability doctrine maps badly onto statistical behaviour.

The first appellate decision on that will do more to shape this industry than any regulatory proposal currently pending, and nobody can tell you when it arrives.

Myrtle Beach, SC Correspondent — From the buyer’s side the concentration shows up as pricing power.

Tools that were free or cheap two years ago are now subscriptions with annual increases, and there’s no alternative because the alternative uses the same underlying model.

That’s the ordinary shape of a captured market and it’s arriving fast in every small business I know.

Owatonna, MN Correspondent — The historical caution is that infrastructure concentration is normal early and gets addressed late.

Railroads, telephony, electricity — each concentrated, each eventually regulated as a utility or broken, each after a great deal of damage in the interval.

We’re in the interval. The question is whether anybody has learned to act during one, and the record says no.