Alternatives
NS&I publishes a prize fund rate and odds. Both are accurate and neither describes what a typical holder gets. Here is the distribution, derived from the published September 2026 prize table.
Premium Bonds are the largest and longest-running prize-linked savings scheme in the world. NS&I publishes everything you need to evaluate them, which is unusual and welcome. What it does not publish is the one number most holders think they are being quoted.
All figures below were checked on the NS&I website on 29 August 2026 and apply to the September 2026 prize draw.
From NS&I's Premium Bonds page, current terms as at 29 August 2026:
NS&I's press release of 18 August 2026 announced the increase from 3.80% and the shortening of odds from 22,000 to 1, adding over 308,000 further prizes and around £63 million to the fund.
The September 2026 draw:
| Prize | Number of prizes | Value |
|---|---|---|
| £1,000,000 | 2 | £2,000,000 |
| £100,000 | 95 | £9,500,000 |
| £50,000 | 192 | £9,600,000 |
| £25,000 | 382 | £9,550,000 |
| £10,000 | 954 | £9,540,000 |
| £5,000 | 1,909 | £9,545,000 |
| £1,000 | 19,892 | £19,892,000 |
| £500 | 59,676 | £29,838,000 |
| £100 | 2,366,135 | £236,613,500 |
| £50 | 2,366,135 | £118,306,750 |
| £25 | 1,717,659 | £42,941,475 |
| Total | 6,533,031 | £497,326,725 |
Two numbers fall straight out.
Mean prize = £497,326,725 ÷ 6,533,031 = £76.13
Implied stock of eligible bonds. If the fund is 4.35% a year paid monthly, the fund is (0.0435 ÷ 12) of the eligible bonds:
£497,326,725 × 12 ÷ 0.0435 = £137,193,579,310
Cross-check that against the odds. At 21,000 to 1 per bond per draw, the number of prizes should be:
£137,193,579,310 ÷ 21,000 = 6,533,028
against 6,533,031 published. The two figures NS&I quotes are internally consistent to five parts in ten million. That is a good sign about the numbers and it is also the point: the rate and the odds are two views of the same fund, not two separate promises.
The 4.35% is the mean return across the whole £137 billion. Your individual return is the sum of a random number of random prizes, and that sum has a distribution that is nothing like its own average.
Start with the simplest exact result. If you hold H pounds of bonds, you have H entries per draw and 12H entries a year, each winning with probability 1 ÷ 21,000. The expected number of prizes is:
λ = 12H ÷ 21,000
and the number of prizes you win is very well modelled as Poisson with that mean, so the probability of winning nothing at all in a year is e^(−λ).
| Holding | λ (expected prizes per year) | P(no prize all year) |
|---|---|---|
| £100 | 0.057 | 94.4% |
| £1,000 | 0.571 | 56.5% |
| £5,000 | 2.857 | 5.7% |
| £10,000 | 5.714 | 0.3% |
| £50,000 | 28.571 | ~0% |
Solve e^(−λ) = 0.5 for the break-point: λ = ln 2 = 0.6931, so H = 0.6931 × 21,000 ÷ 12 = £1,213.
Any holder with roughly £1,200 or less is more likely than not to win nothing in a whole year. Their advertised return is 4.35%. Their most likely return is zero. Both statements are true.
For larger holdings the question is not whether you win but how much. Because 98.73% of prizes are £25, £50 or £100 — carrying 80.0% of the fund between them — while 1.27% of prizes carry the other 20%, most holders collect a run of small prizes and never touch the top of the table.
Drawing prizes at the published frequencies, with the prize count Poisson at λ = 12H ÷ 21,000, gives:
| Holding | Mean return | Median return | Share of holders below the 4.35% mean |
|---|---|---|---|
| £1,000 | £44 (4.35%) | £0 (0.0%) | 65% |
| £5,000 | £217 (4.35%) | £150 (3.00%) | 65% |
| £10,000 | £433 (4.35%) | £350 (3.50%) | 68% |
| £20,000 | £868 (4.35%) | £725 (3.62%) | 67% |
| £50,000 | £2,181 (4.35%) | £1,900 (3.80%) | 69% |
(Simulated over 400,000 trials per holding at the September 2026 prize table; the mean column recovers 4.35% to two decimal places, which is the check that the simulation is faithful.)
Read the last column. About two thirds of holders, at every holding size, earn less than the headline rate. The mean is dragged up by a small number of very large prizes that almost nobody wins. That is not a flaw in NS&I's disclosure — the prize table is published in full — but it is the reason the headline rate is a poor guide to your own outcome.
A £50,000 holder has 600,000 entries a year. Against the implied £137.19bn bond stock:
Those are the prizes doing the work in the 4.35%. Holding the legal maximum, you would expect to wait roughly 114,000 years for the million. It is a real prize and it is genuinely won every month; it is simply not part of any sensible personal forecast.
The case is not the return. It is the combination of three properties:
Tax-free prizes. Premium Bond prizes are exempt from UK Income Tax and Capital Gains Tax. That matters most once the Personal Savings Allowance is exhausted — £1,000 of savings interest tax-free for basic-rate taxpayers, £500 for higher-rate, and £0 for additional-rate. For an additional-rate taxpayer with substantial cash, 4.35% tax-free is equivalent to about 7.9% gross, which reframes the comparison entirely.
Capital security. Holdings are backed by HM Treasury, not by a deposit-insurance cap, so the £50,000 maximum is fully protected rather than protected up to a limit.
Liquidity with a lottery attached. You can cash out; you have not bought a bet.
The case against is equally plain: the median holder underperforms the mean, small holders often earn nothing, the rate is variable and has been cut before, and none of it is index-linked, so in real terms a period of high inflation is a loss. If you want the certain-return comparison instead, that is invest instead of tickets; if you want to know how Premium Bonds compare with other countries' schemes, that is prize-linked savings by country and the PLS comparator.
Last verified: 2026-08-29