The commodity trading industry is adding AI fast. Every serious vendor now has an AI roadmap. Most will claim automation, intelligent document processing, and operational efficiency. Few will tell you where the AI actually runs.

That is the question that matters.

AI in trading requires access to sensitive data

AI is only useful in commodity trading if it can work with the data that drives decisions - contracts, invoices, positions, P&L, counterparties, and internal workflows. That is not generic business information. It is commercial intelligence built over years of trading relationships and market knowledge.

When that data is sent to a third-party AI outside the firm’s control, the firm inherits questions that compliance checklists do not answer: who controls the infrastructure, where does the model run, and which jurisdiction applies to the data?

What Euclid deploys, in plain language

Euclid deploys on client-owned infrastructure or inside Euclid's private Swiss datacenter.

This is not a Swiss cloud framing. It is private infrastructure - either in the client’s environment or in Euclid’s private Swiss datacenter - designed so sensitive trading data stays inside a controlled perimeter.

What sovereignty actually requires

A US hyperscaler with a Swiss data centre is not sovereignty. Under the CLOUD Act, jurisdiction can still apply regardless of where servers physically sit.

Sovereignty is architectural control: where the system runs, who operates it, and which legal regime can reach it.

What Euclid AI does today (and what it enables next)

Today, Euclid AI ships two capabilities inside the CTRM environment:

• Document data integration / extraction across 50+ formats (contracts, invoices, confirmations, reports), including handwritten contracts - with 98% less time typing and 99.3% OCR accuracy.

• Natural-language querying of the CTRM database.

Those capabilities are the first visible proof point of private AI inside the operational core. They are also the foundation for extending the same private AI layer further into workflow support over time - without changing the deployment model or pushing sensitive data outside the perimeter.

The cost structure

Legacy cloud-dependent CTRM typically costs $45k to $200k per year. Euclid runs from $5k to $50k annually. The cost is infrastructure and electricity in a controlled environment, not unbounded per-token billing tied to a third-party API.

Deployment takes less than one week on existing hardware, versus 6 to 18 months for a comparable legacy implementation.

The conclusion

The CTRM market will keep adding AI. Where the model runs, who controls the infrastructure, and which jurisdiction applies to the data will become harder to ignore.

Euclid is built around a clear answer to those questions.

See how private AI works inside your CTRM.