For the first time in eight weeks, the concentration of AI gains is barely ahead of the spread. One number explains why this week felt different – and why it might not last.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week, a big question got a little less scary: is AI making a tiny group of people very rich while everyone else pays the price? The honest answer is – still yes, but only just. Our scan of nearly 300 stories from around the world showed that for the first time, almost as many stories described AI spreading its benefits widely as described those benefits being grabbed by a few powerful companies and governments. That gap – what we call the Distribution Gap – shrank to almost nothing. But here's the catch: in North America, where most of the big AI companies live, the gap is still huge. And the Paperclipalypse worry – AI acting in ways nobody planned and nobody can stop – is getting louder, not quieter. The Futurama future, where AI stays a helpful tool and ordinary people share the upside, is still possible. We're just not there yet.
Here's the thing. For seven consecutive scans, AI-Dystopia has outscored Futurama globally. Sometimes by a little (Scan 3, gap of 2 points). Sometimes by a lot (Scan 2, gap of 15). This week, the Distribution Gap – AI-Dystopia minus Futurama – closed to just 2 points: 28% versus 26%. That's the tightest reading in the series. It is not a reversal. It is not a cause for uncomplicated celebration. But it is a signal worth taking seriously, and it's not an accident.
What drove Futurama higher this week wasn't a feel-good narrative. It was concrete: CBA testing an AI companion for business loans. Bunnings and banks scaling expertise across channels. NZ's early adopters reporting real returns. BlueAPACHE winning mid-market deals. Malta distributing AI to every citizen. The Lenovo CIO Playbook showing 95% of ANZ enterprises planning higher AI investment with an execution, not experimentation, mindset. Futurama at 26% – its series high – reflects organisations that are actually doing the work.
Singularia dropped sharply: from a rolling 15-17% range to 10% this week. That matters. The frontier capability story – agentic AI, coding agents, protein models, world models – is still loud (Anthropic's near-$1 trillion valuation, Claude Opus 4.8, the Cognition $1B raise, DeepMind CEO predicting AGI in three years). But this week's signal pool weighted enterprise adoption and workforce economics heavily, which pulled Singularia back toward its proper share. It's a reminder: the labs' ambitions and what enterprises actually do with AI are two very different stories.
ANZ is the outlier this week – in a good way. Futurama at 35% is the highest single-week regional reading for any major region in the series. The rolling four-week figure of 29% confirms this isn't a one-week spike. The ANZ Distribution Gap is –9: Futurama is running nine points ahead of AI-Dystopia. That's a genuinely different posture from the global picture.
What's behind it? A cluster of adoption stories with real names attached. CBA's Matt Comyn seeking a 'social licence' for AI – naming the trust problem rather than papering over it – while simultaneously deploying AI companions for loan verification. Kmart rolling out Google Gemini for virtual try-on. Realestate.co.nz using Snowflake for image search. AI agents working for charcoal chicken shops and home gym businesses. These are not press releases. They're operational deployments.
The counterweight is real, though. Paperclipalypse at 21% (rolling 20%) is the highest of any region. Australia's AI war machine being built with no public accountability. AI-generated deepfake memes of the Prime Minister creating legal exposure. Qilin ransomware targeting Australian organisations. Courts cracking down on AI hallucinations. NZ experts warning that David Seymour's light-touch regulation plan is dangerous. The OAIC finding near-zero public trust in AI companies. ANZ leaders are more optimistic about what AI can do for them – and more worried about what it might do to them – than any other region in the scan.
The Firmus IPO delay is the single most interesting ANZ story this week. An AI company repeatedly failing to list despite the boom signals that Australia's capital markets are more discriminating than the US headlines suggest. That's probably healthy.
North America tells a different story. AI-Dystopia at 33% (rolling 32%) and a Distribution Gap of 13 points – the widest of any region. This is where capital and power are pooling fastest. Anthropic's valuation race with OpenAI. The Snowflake-AWS $6 billion chip deal. SK Hynix, Micron and Nvidia in the trillion-dollar club. Wix laying off 20% citing AI. Uber's COO unable to justify the AI budget. Corporate America starting to ration AI as costs skyrocket. The Pope's encyclical generating seven separate story treatments.
The AI-Fizzle signal is also distinctly North American this week: Uber burning through its 2026 AI budget in four months. Starbucks quietly retiring its inventory agent after hallucinations. GitHub Copilot's token-billing backlash. 'Nobody knows anything', says Wharton's Ethan Mollick. These aren't marginal voices. These are major enterprises and credible analysts saying the ROI story is unproven. Fizzle at 20% in NA (rolling 17%) is meaningful.
Separating industry self-promotion from genuine adoption evidence: 23 of 332 stories this week were vendor-sourced. Dell, Proofpoint, Claroty, Snowflake and others talking about their own stacks. Treat those accordingly. The honest adoption signal in NA is mixed: real gains at Vanguard, Glean crossing $300M ARR by cutting AI costs for enterprises, Walmart's AI-driven delivery. But also Starbucks' failure, Uber's budget crisis, corporate AI rationing. The gap between the industry's narrative and enterprise reality is still wide.
Europe is statistically the most balanced region this week: AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse all within a few points of each other, with AI-Dystopia and Futurama tied at 22%. The rolling Distribution Gap of 5 points is narrower than North America's 14 but wider than ANZ's –6. Mistral chasing superintelligence to counter US dominance. SoftBank plowing $52 billion into French data centres. The ECB warning banks on AI risks. Malta distributing AI to citizens. Moneybox seeking regulatory approval for AI advice. All major AI models violating EU regulations in up to 93% of tested cases – that last one belongs in the Paperclipalypse category and it's a big deal that it's coming from a study rather than a vendor.
Five percent of the global weighting, but 64 signals. APAC's AI-Dystopia at 41% (rolling 39%) is the highest concentration reading of any region. China restricting travel of elite AI engineers. Huawei's chip milestone altering the US-China race. ByteDance building custom CPUs. Alibaba and Tencent pivoting to embodied AI and robotics. Singapore's VC boom concentrated in AI and deep tech despite fewer overall deals. The OpenAI–Singapore $300M MOU deepening dependence on a single hyperscaler while claiming sovereignty. Futurama holds at 25% – Grab and RSM reaching 12,000 SMEs, Alibaba Cloud giving away free AI tools to students and businesses, Singapore factory output up 17.6% on AI-related electronics demand. The APAC story is one of governments and large enterprises actively managing a power-concentration risk they can clearly see.
Technology (102 signals) is the dominant force, as always, but this week it's splitting cleanly. On the Singularia side: frontier model releases, agentic infrastructure, embodied AI. On the Futurama side: practical deployments, SMB tooling, national AI strategies bearing fruit. Economy (80 signals) is pulling hard toward AI-Dystopia: infrastructure capex concentrating in hyperscalers, chip valuations, AI token costs outrunning budgets. Trust (64 signals) is almost entirely Paperclipalypse fuel: AI-driven cyber exploitation overtaking phishing as the primary attack vector, shadow AI, deepfakes, hallucinations, the ECB and OAIC both signalling that AI is outpacing oversight. Talent (37 signals) is split between Dystopia (layoffs, junior pathway destruction, China's travel bans) and Futurama (reskilling, augmentation, AI broadening access). Geopolitics (19 signals) is almost exclusively AI-Dystopia: the US-China chip race, Jensen Huang joining Tsinghua's board, Mistral's sovereignty play.
What's structurally absent from the discourse this week: mid-market and small-business ROI evidence from outside ANZ. Government AI adoption that's working at scale (not pilots, not strategies – actual operations). Workforce transition programmes with measurable outcomes. Climate and environmental accounting for AI infrastructure costs beyond headline warnings. And honest post-mortems from organisations that tried and failed – beyond Starbucks, which was reported reluctantly.
The Distribution Gap closing to 2 points doesn't mean the concentration problem is solved. It means some organisations are capturing value broadly enough to shift the signal. If you're a CEO or CIO reading this: the question isn't whether to invest in AI. Ninety-five percent of ANZ enterprises already plan to. The question is whether you're building the operating model – data governance, security controls, trust frameworks, human oversight – that determines whether AI value stays with the organisation or leaks out in cost blowouts, security incidents and failed deployments. Uber burned its annual AI budget in four months. Starbucks retired its agent after hallucinations embarrassed the baristas. The organisations showing up in Futurama this week – CBA, Vanguard, Grab, Kmart – are the ones that treated governance as the product, not the afterthought.
The Distribution Gap is narrowing. Don't let that make you comfortable. Let it make you precise.
The articles most responsible for bending the five worlds this week.
Uber's COO spent the whole year's AI budget by April and couldn't point to customer value gained. When one of the world's biggest tech companies publicly says it can't justify what it's spending on AI, that's a warning sign for every business planning to pour money into the same tools. It was the clearest 'pump the brakes' moment of the week.
The new Pope's first major statement wasn't about faith – it was about who controls AI and who pays the price. He called out a small group of powerful companies setting the rules for everyone else, and that message landed in seven separate news stories across our scan. When a moral authority with a billion followers weighs in on AI power concentration, it shifts the conversation in a way no analyst report can.
New Zealand infrastructure company Infratil is making serious money from AI data centres and energy assets – and openly saying so. The gains are real. But they're going to infrastructure investors, not to workers or communities. This story appeared six times across different publications, making it the most-covered business story in ANZ this week.
The AFR captured something important: the early excitement about AI is wearing off. Students are angry about job prospects, companies are seeing costs blow out, and the promised returns haven't arrived. It's a useful gut-check for any business leader who assumed the hard part was deciding to invest in AI rather than making it actually work.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free