Keir Starmer recorded the lowest satisfaction level of any prime minister in British history. Amazingly, things have only gotten worse since then – particularly following the disastrous local elections which took place on 7 May.
As a result of this chaos, Starmer has found himself fighting to remain as PM. As such, it’s surprising to see him launching a £500m initiative which involves Tony Blair and AI – two things which have proven immensely unpopular with the British electorate.
— Mat (@fragmentshore) May 12, 2026
Starmer: AI carumba
As reported by the Daily Mail:
The Government’s Sovereign AI fund, launched in April, is a £500million state-backed investment initiative which aims to grow the UK’s AI startups in order to reduce dependence on foreign tech giants.
When it comes to AI, there are three types of company which we could be funding, including those which:
- Sell the hardware needed to run AI models (e.g. Nvidia).
- Create the AI models themselves (e.g. ChatGPT, Anthropic, Google, Facebook, and Microsoft).
- Use AI models like ChatGPT to create niche software services.
There’s no way the UK could create a chip company that could rival Nvidia – certainly not for £500m. Even China is struggling to catch up to Nvidia, although the same was once true for renewable technology and electric cars, and they now dominate both sectors.
It would equally be impossible for the UK to start training its own AI models, as it would cost well over £500m to train each model – at least if they were to be competitive anyway:
Interesting look at the exponential cost of training OpenAI ChatGPT models from generation to generation.
Based upon these estimates the cost of GPT-5 could be as high as $2.5 Billion representing about 17x+ the cost to train GPT-4.
Great for nvidia and overall capex. Quite… https://t.co/sAB3yfy5nk
— Daniel Newman (@danielnewmanUV) May 28, 2024
You might ask: ‘could we not just spend £2bn training a model and then claw that back with the profits we make?‘
The problem is that AI companies are burning more money than they’re making. This software is incredibly expensive to run, and it’s not useful enough for people or businesses to be willing to pay what it actually costs:
Companies like OpenAI are kept afloat with investor handouts.
Oh, and speaking of handouts:
Who is pumping all these billions into Tony Blair’s son’s AI company after its seventh consecutive year of losses? Why? Oh well probably nothing pic.twitter.com/Mv6btRu6WV
— Eyup Lovely (@eyuplovely) August 17, 2024
Getting to the third type of company, it’s likely this Sovereign AI Fund will be used to fund companies which rent ChatGPT or Claude in order to create niche software services. This would be fine, except these companies are reliant on AI models like ChatGPT and Claude continuing to exist. This is far from guaranteed, because – as noted – THESE COMPANIES AREN’T PROFITABLE.
Your buzzwords won’t save you
We should clarify that the government hasn’t put Tony Blair’s daughter-in-law in charge as some sort of wedding gift. As the Daily Mailreported on Suzanne Ashman:
The 37-year-old venture capitalist previously worked as a general partner at two London-based venture capital firms, LocalGlobe and Latitude. She will lead investments with the goal of boosting ‘homegrown’ AI companies in the new role.
Ashman herself said:
Excited to share that I’ll be joining the new Sovereign AI Fund as Managing Partner.
Britain already produces world-class AI founders and researchers. The challenge is helping more of them build enduring global companies from the UK.
I’m excited to help build a fund that backs those founders building strategically important parts of the AI stack – providing capital, access to compute, procurement pathways and long-term institutional support.
And one that aims to deliver meaningful returns for the British taxpayer alongside broader national capability. I don’t see those goals as being in tension.
Buzz words, synergy, waffle – none of which will mean a thing if the underlying AI companies collapse in on themselves.
It gets worse
When it comes to the problems with AI, we really are just scratching the surface. Of course, if AI models magically become profitable for some unforeseen reason, we’re going to look pretty silly.
We wouldn’t bet £500m on it, though.
Featured image via The Canary
By Willem Moore
From Canary via This RSS Feed.