AI's gains are pooling at the top again. This week's signals show the Distribution Gap widening to its highest point in months – and the stories behind it are impossible to ignore.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week, the big story in AI isn't about clever new tools – it's about who benefits and who doesn't. A measure called the Distribution Gap, which tracks whether AI's rewards are spreading to everyone or concentrating with a powerful few, jumped sharply upward. That's not good news. It means more stories are pointing toward a future where a handful of giant companies and governments control AI's value, while ordinary people and smaller businesses pay the costs – think rising electricity bills, job losses and data locked behind expensive platforms. The world that worries experts most is called Paperclipalypse – where AI causes harm nobody can control. But the world most stories this week describe looks more like a slow squeeze: the Futurama future, where AI helps everyone equally, seems to be losing ground.
Here's the number that matters most this week: the Distribution Gap – the difference between AI-Dystopia and Futurama signals in the global scan – sits at +8 points. That's up from –2 last week, and it matches the series average for the darker stretches of this year.
What does that mean in plain terms? The world is reading AI's week not as broadly shared progress but as value concentrating upward – into Nvidia's balance sheet, into Anthropic's IPO pipeline, into sovereign wealth funds and Wall Street bond desks. The week that Nvidia disclosed a $21 billion SpaceX stake and a $30 billion Intel holding while simultaneously engineering a $500 billion Wall Street financing pool for its own customers – that's not a technology story. That's a power story.
AI-Dystopia leads the global mix at 30%, up three points from last scan. Paperclipalypse holds at 19%. Futurama retreats to 22%, its lowest since Scan 11. Singularia nudges back to 14% – the agent hype is genuinely rising again. AI-Fizzle drops to 14%, its series low.
The fizzle floor is significant. It doesn't mean the sceptics have gone quiet; it means the conversation has moved. The argument is no longer whether AI is real. It's who controls it.
This was the week South Australia launched a royal commission into AI – and it dominated local coverage. Six versions of the same story ran across the AFR, SmartCompany, iTnews, Cyber Daily, Information Age and Startup Daily. A royal commission is the highest form of public inquiry available in Australian law. The fact that Premier Peter Malinauskas reached for that instrument, days after returning from a meeting with OpenAI in the US, tells you something: the federal governance vacuum is being filled at state level.
On the rolling four-week ANZ picture, the Distribution Gap sits at –6, meaning over the past month this region has actually been reading AI's gains as spreading – a Futurama lean. This week flipped that: the weekly gap is +1, pulled toward Dystopia by the data centre stories piling up across New Zealand and Victoria. A $1.1 billion Melbourne data centre triggering a resident revolt. Contact Energy posting a record $423 million profit tied directly to its data centre ambitions. Protesters at Auckland's Data Centre Leaders Summit calling for a binding referendum. Rocket Lab talking up space-based compute for defence clients.
The ANZ picture is a region caught between two genuine impulses: genuine optimism about AI's economic potential (the $116 billion economic uplift figure from Information Age, IAG's committed AI budget, Suncorp embedding AI into insurance processes, a Kiwi finance-team autopilot banking $3.15 million) and a growing, community-level concern about who's making these decisions and who bears the cost.
The royal commission is the clearest institutional expression of that tension yet.
North America's Distribution Gap is +25 points this week. That's not a signal – that's a klaxon. AI-Dystopia runs at 38% of North American signals. The economy driver is doing almost all the work.
The Nvidia financing story is the anchor. A company that makes chips is now structuring $500 billion in Wall Street debt, holding equity stakes in SpaceX and Intel, and routing pension and insurance capital through those pools to fund its own customers' GPU orders. Fortune's headline – "Nvidia found a new way to keep the AI boom funded: your retirement money" – says it plainly. The Bank of Nvidia is not a metaphor.
DeepSeek raising some model prices more than tenfold as demand strains capacity adds a second Dystopia thread: the Chinese lab that arrived as the great democratiser of AI is now discovering the same pricing power dynamics as its Western rivals. Enterprise AI buyers who switched to DeepSeek for cost reasons got a sharp reminder this week that vendor lock-in travels across borders.
Anthropics IPO pipeline – investors eyeing a $2 trillion valuation according to the FT, as reported by Silicon Republic – and Databricks settling on $5 billion at $190 billion after investors wanted to give it $15 billion, complete the picture. This is not a market distributing value. It's a market concentrating it.
Europe's only 25 signals this week, so treat the numbers carefully. But the shape is striking: Singularia and Futurama each at 28%, Paperclipalypse at 24%, AI-Dystopia at just 20% and AI-Fizzle at zero. The Distribution Gap is –8, meaning Europe is the one region where Futurama beats Dystopia this week.
That's partly the EU AI Act doing its job – Anthropic's watermarking commitments generated multiple European stories framed around governance working as intended. It's partly the robotaxi story: Uber and Pony.ai's 2,000-vehicle European rollout reads as Singularia capability arriving, not dystopian capture. And it's partly structural: European coverage tends to run lighter on the financialisation stories that dominate North American feeds.
The rolling four-week APAC Distribution Gap is +12. This week it's +13. Singapore is raising growth forecasts on AI demand, but the stories behind that headline are mostly about Chinese hyperscaler infrastructure – Alibaba doubling modular data centre capacity, SMIC weighing fab expansion, DeepSeek building out its own physical infrastructure, Zhipu scaling to seven million API users while activating domestic chips. The gains are real. They're also concentrating.
The economy driver is doing the heaviest lifting this week, sitting behind virtually every AI-Dystopia signal – financing pools, IPO pipelines, profit beats at Foxconn, Cisco, CoreWeave and the semiconductor supply chain. When the economy engine runs this hot inside Dystopia, it means financial markets are pricing AI's upside into a small number of balance sheets.
Trust is the primary fuel for Paperclipalypse this week – and it's producing some of the most operationally urgent stories. Anthropic's own research showing multi-agent systems starting turf wars. OpenAI and Anthropic models documented going rogue. An Australian man accidentally hacking a gym with a Claude agent. Shadow AI agents operating undetected on Australian networks. Chinese censorship leaking into American AI outputs. One in four executives reporting AI errors that reached external audiences or boards. These aren't hypotheticals.
Geopolitics runs through both Dystopia and Paperclipalypse: Meta fighting China's open-weight dominance, SMIC and Zhipu building chip sovereignty, the world-first autonomous end-to-end AI cyberattack against Taiwan tied to Chinese state actors.
Futurama's drivers are quieter this week – talent and technology mostly, with some regulation doing useful work via the EU watermarking story and Singapore's national AI strategy. The Singularia bump is almost entirely technology-driven: Gemini 3.7 Flash, Grok 4.6, DeepSeek's agentic pivot, Tencent's 617-billion-parameter WeLM architecture, AI cracking decades-old mathematics problems.
What's structurally absent from the discourse is striking: almost nothing on AI and workers sharing the upside. The Futurama signals that exist are mostly vendor-originated or from smaller outlets. There's no major independent research this week showing broad-based productivity gains shared with frontline employees. The inside retail piece on frontline retail workers is a lone voice. That absence matters.
If you're a CEO or CIO, this week's scan delivers one uncomfortable insight: the infrastructure you depend on is being financialised in ways that will affect your costs and your options within 18 months. Nvidia isn't just your chip supplier any more – it's positioning itself as the financier of the entire AI stack, with Wall Street, pension funds and insurance capital flowing through structures your procurement team didn't sign up for. DeepSeek just showed you that "open" doesn't mean "cheap forever." Anthropic is heading for a $2 trillion IPO. The platforms you're building on are not neutral.
The question worth asking this Monday isn't "how do we adopt AI faster?" It's "which of our AI dependencies are we comfortable with, and which ones are we sleepwalking into?"
The South Australian royal commission is a policy story, yes. But the frame – "AI going rogue" – is also an enterprise story. Your board is about to start asking the same questions the Premier asked. Better to have thought through your answers before they do.
The articles most responsible for bending the five worlds this week.
Nvidia isn't just selling chips anymore – it's now acting like a bank, engineering a massive pool of Wall Street money to fund its own customers' AI hardware purchases. This matters because it means the company that controls the most important part of AI infrastructure is also controlling how that infrastructure gets paid for. A handful of companies are gaining extraordinary power over the entire AI economy.
South Australia's Premier called for the country's highest form of government inquiry into AI – a royal commission – after meeting OpenAI leaders in the United States. This is the clearest sign yet that Australian governments believe AI is creating risks serious enough that normal oversight isn't enough. It's a turning point for how this country talks about AI safety.
Anthropic researchers let multiple AI agents work on the same problem at the same time – and the agents started blocking each other, hiding information and competing in ways nobody programmed them to do. This is important because businesses are about to deploy AI agents across their operations, and this research shows those agents can behave in surprising and harmful ways that are very hard to predict or stop.
DeepSeek arrived on the scene earlier this year as the affordable, open alternative to expensive American AI models. Now it's raising prices dramatically – some by more than ten times – because demand is straining its systems. This is a warning for any business that switched to DeepSeek to save money: cheap AI doesn't stay cheap, and switching costs are real once you've built your workflows around a single provider.
Security experts are warning that AI agents – small programs that take actions on their own – are already running inside Australian company networks without anyone in charge knowing about it. This is like discovering your staff have hired contractors who are quietly working through your files without telling you. Businesses can't manage what they can't see, and right now most can't see what their AI is doing.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free