Live Nov 18 · Cohort 3Nov 18, 19:00 MSK — free hands-on AI agents live workshop · Live Cohort 3 starts Nov 23Join Nov 18 Live →Cohort 3 (Nov 23)
ReleaseAPI / Product · Actionable3DNews AI→ original

Russian Businesses Overpay for Foreign AI by Up to 9x: Where the Intermediary Markup Comes From

Russian companies pay up to nine times more for access to foreign AI models than the same API costs at the developers' list price. Intermediaries explain the difference by the handling of ruble payments, availability guarantees and legally clean access. In essence, businesses are buying the removal of payment and legal risks. But at large token volumes, such a markup quickly becomes the main expense item.

AI-processed from 3DNews AI; edited by Hamidun News
Russian Businesses Overpay for Foreign AI by Up to 9x: Where the Intermediary Markup Comes From
Source: 3DNews AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary source3DNews AI ↗
Editor-in-Chief
Key metric #1Compare the intermediary's price with the provider's official price list
Key metric #2Calculate the price per million tokens, not just the subscription
Key metric #3Keep a backup provider in case access fails

Imagine this: a team in Moscow connects a language model to customer support. In the developer's price list the cost looks modest, but on the invoice from a Russian intermediary it turns out to be many times higher. According to 3DNews, the gap reaches nine times. This is not an invoice error, but the price of an entire chain of services that businesses do not always notice when choosing a supplier.

What the Markup Consists Of

The intermediaries themselves name three reasons.

First: transferring ruble payments. It is difficult for a Russian company to pay for a foreign API directly, so the money is routed through third-party schemes. These cost commissions and carry risks that the supplier builds into the price.

Second: availability guarantees. The service takes on dealing with regional restrictions, accounts and backup channels so that the client's requests do not fail at the most inconvenient moment.

Third: legally clean access. The client receives a contract, closing documents and a scheme that accounting and lawyers understand.

An important caveat: this is the suppliers' own version. The material contains no independent breakdown of the markup by item. It is unknown what part of the ninefold difference is real costs and what part is margin.

What It Looks Like in Numbers

Take a hypothetical example, not from the research. A product spends $1,000 per month on a model at the official price list. In a pilot, a 9x markup makes it $9,000. That is unpleasant but tolerable for an experiment.

Now the load grows 20 times. At the developer's list price it comes to $20,000, through an intermediary already $180,000. The difference of $160,000 per month turns into a separate budget that could support an entire team.

That is why the markup should be counted not as a one-time commission, but as a multiplier on all future load. The more successful the product, the more this multiplier hurts.

What Developers and Product Managers Should Do

First: demand a transparent formula. A fixed access fee and a multiplier on each token produce completely different economics as volumes grow.

Second: calculate the cost per unit of value, not per token. If a task can be solved by a small, cheap model, there is no point paying a premium for a flagship. The markup hits expensive models hardest.

Third: build an abstraction over the provider. If model calls go through a single layer, switching suppliers or routing part of the traffic to another source takes days, not months. This protects against both price increases and sudden access failures.

Fourth: split the workload. Mass routine tasks such as classification, field extraction and draft replies can often be moved to open models on your own servers. Leave complex cases to foreign models. The limitation is obvious: you need GPUs and people to maintain them, so the benefit appears only at sufficiently large volumes.

Fifth: do not skimp on the legal side. A low price from an opaque reseller can turn into problems with documents, terms of use or a sudden shutdown. Sometimes the markup really does pay for peace of mind.

Conclusion

A ninefold overpayment looks frightening, but it is more a diagnosis of the market than a scandal. While direct access is difficult, intermediaries sell not tokens but the removal of risks. The team's task is to understand exactly what it is paying for, and to notice in time the moment when it becomes more profitable to build its own architecture with several providers and open models.

Story

Key Questions Answered

What the Markup Consists Of?

The intermediaries themselves name three reasons. First: transferring ruble payments. It is difficult for a Russian company to pay for a foreign API directly, so the money is routed through third-party schemes. These cost commissions and carry risks that the supplier builds into the price. Second: availability guarantees. The service takes on dealing with regional restrictions, accounts and backup channels so that the client's requests do not fail at the most inconvenient moment. Third: legally clean access.

What It Looks Like in Numbers?

Take a hypothetical example, not from the research. A product spends $1,000 per month on a model at the official price list. In a pilot, a 9x markup makes it $9,000. That is unpleasant but tolerable for an experiment. Now the load grows 20 times. At the developer's list price it comes to $20,000, through an intermediary already $180,000. The difference of $160,000 per month turns into a separate budget that could support an entire team.

What Developers and Product Managers Should Do?

First: demand a transparent formula. A fixed access fee and a multiplier on each token produce completely different economics as volumes grow. Second: calculate the cost per unit of value, not per token. If a task can be solved by a small, cheap model, there is no point paying a premium for a flagship. The markup hits expensive models hardest. Third: build an abstraction over the provider.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…