For early-stage AI founders, the Start Up Loans scheme is one of the few accessible, non-dilutive-of-equity finance routes — though it is still debt. Here’s how it works for AI startups in 2026, cited to the British Business Bank. (dgm implements osFoundry as an independent partner.)

What Start Up Loans are

Start Up Loans are government-backed personal loans for early-stage businesses, delivered via the British Business Bank. The headline terms:

  • £500–£25,000 per individual;
  • a fixed 7.5% per annum (confirm the current rate at application);
  • for businesses in roughly their first three years of trading; and
  • free mentoring included.

Crucially, use is unrestricted — so the funds can go toward tech, software, compute, or early AI development costs.

A useful quirk for founding teams

Because Start Up Loans are personal loans to individuals, each eligible co-founder can apply for up to £25,000. A founding team of two or three can therefore raise more in aggregate than a single applicant — each assessed on their own circumstances and the business plan. That makes the scheme more meaningful for a startup than the headline £25,000 first suggests.

It’s debt, not a grant

Be clear-eyed: a Start Up Loan is debt you repay (typically over one to five years, with no early-repayment fees), not a grant and not equity. The upside is no dilution; the responsibility is a personal repayment obligation. It suits early runway, not a large-scale build-out.

How much capital does an AI startup actually need?

Less than the old “raise a big seed round to buy infrastructure” model assumed. Usage-priced AI platforms and bring-your-own-key access mean you can build and test an MVP without large up-front licences or committed compute spend. So the realistic use of a Start Up Loan is runway — covering your time and early costs while you prove traction — rather than funding heavy infrastructure on day one.

Where osFoundry and dgm fit

dgm helps early-stage founders scope an MVP that’s cheap to test and quick to learn from. We build on osFoundry, which is usage-priced with no per-seat fees, supports bring-your-own-key (pay model providers directly), and even runs local inference on your own hardware via llama.cpp — all of which keeps early burn low. As you grow, it self-hosts in your own cloud or uses an EU region for UK data-sensitive work (osFoundry publishes US/EU/JP regions, not a UK one).

dgm is an independent integration partner with zero integrations so far, and is not a finance broker. For a Start Up Loan, apply via the official scheme. To scope a lean AI MVP, book a consultation with dgm. General information, not financial advice.