GPU World Opens a $100,000 Story Contest About a One-GPU-Per-Person Future

The site gpuworld.org has opened a story contest built around a single thought experiment: what if the world had one GPU per person? Submissions opened in August and run until 31 October 2026, with $100,000 in prizes for writers who imagine a future in which frontier AI is no longer scarce.

A GPU-poor humanity

The premise begins with the observation that the AI revolution has only just begun to affect humanity: billions of people have yet to so much as talk to a frontier model. The organisers attribute this largely to compute limits — only a few million GPUs capable of efficiently serving the highest-quality models are manufactured each year, so the best AI reaches a relative handful of users. In their framing, humanity is "GPU-poor".

A world with 8 billion GPUs

Entrants are asked to assume that frontier progress stops on 1 September 2026: models keep getting faster and cheaper but never become superhuman. The singularity never arrives — yet GPUs keep being manufactured, and by 2040 there could be the equivalent of 8 billion of them, roughly one per person, each with access to a frontier LLM around the clock. From that fixed assumption the organisers ask what "business as usual" would look like: relentless AI surveillance, infinitely patient tutors, the disappearance of social media as we know it, AI doctors and personalised medicine, and what happens in the developing world. They note that AI already holds the potential to be more transformative than the smartphone, or perhaps the internet itself.

Prizes, deadline and rules

The prize pool is $100,000: $40,000 for first place, $20,000 for second, $12,000 for third, plus seven finalist prizes of $4,000 each. Entries opened in August, close on 31 October 2026 at 11:59 PM PT, and winners are announced in December. Entry is open to everyone, one per person, in fiction or nonfiction, 1,000 to 5,000 words, submitted as Markdown or PDF and licensed CC BY-NC or freer so entries may be republished. LLM use is permitted but discouraged: the organisers say it tends to reduce originality, that flaws are especially visible when outputs are read as a group, and they ask for disclosure of AI assistance.

The contest leans on a William Gibson line — "The future is already here, it's just unevenly distributed" — and asks writers to imagine it distributed evenly. Whether or not the forecast holds, the exercise puts the distribution of AI compute, rather than its capability, at the centre of the question.