The global AI story shifted this week: value is concentrating faster than it's spreading, and the gap between who wins and who waits is narrowing to a knife's edge.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week's scan tracked 317 stories about AI in business from around the world. The big news? AI is generating enormous profits and investment – but mostly for a small number of giant companies, not for everyday workers or smaller businesses. Nine out of ten executives say AI hasn't made their teams more productive, yet jobs are still being cut. That's the scary Paperclipalypse and AI-Dystopia direction – where AI's rewards pile up at the top while everyone else foots the bill. The good news is that Australia and New Zealand are bucking the global trend, with more stories pointing toward the hopeful Futurama future where AI stays a tool people control and the benefits spread around. The question for every business leader right now: which side of that gap are you on?
Here's the number that matters most this week: the global Distribution Gap – AI-Dystopia minus Futurama, the single best measure of whether AI's value is concentrating or spreading – came in at +2 points. That's the narrowest it's been since scans 3 and 8, and a sharp reversal from last week's +8.
So are we turning a corner? Not quite. Read the series and the pattern is clear: we've oscillated between +1 and +15 for twenty weeks, with the centre of gravity sitting around +5 to +7. A single week at +2 is welcome, but it's not a trend break. It's a regression to a slightly less alarming mean.
The global shares this week: AI-Fizzle 19%, Futurama 24%, AI-Dystopia 26%, Singularia 13%, Paperclipalypse 18%. Every world is essentially at its series average. What that tells you is that the discourse isn't moving – it's just cycling loudly through the same anxieties and hopes, week after week.
The macro forces underneath each world this week are worth naming plainly.
Economy is the dominant driver of AI-Dystopia (62 signals out of 381 total). The Alibaba earnings cluster – profit falling 76% while capex rises 75%, yet revenue climbing 9% – is the textbook case: capital flooding in, returns still uncertain, and whatever gains exist pooling at the platform level. Add the OpenAI/Nvidia Ohio data centre ($105–150B depending on which headline you trust), Stripe's $7–8B OpenRouter acquisition, Anthropic's $65B annualised revenue, and California's $366B VC bonanza, and you have a week where the infrastructure story is deafening.
Trust is carrying Paperclipalypse (52 signals). The Canva seat-cycling fraud, the Microsoft Copilot patch that took eight months, AI audio deepfakes hitting Singapore's TikTok users, AI-generated malware hitting US critical infrastructure, the alignment problem going from theoretical to operational – these aren't isolated incidents. They're a pattern of AI systems producing harms that existing controls weren't built to catch.
Technology is the engine of both Futurama and Singularia. Post-training scaling laws (Z.ai's GLM-5.3), model routing democratising access, Ramp's Router, Snowflake's dynamic routing – these are genuine capability advances that lower barriers. Against that: Gartner's warning that agent costs will rise fivefold by 2028 even as per-token prices fall. The efficiency gains and the cost surprises are arriving simultaneously.
Geopolitics is the undercurrent of AI-Dystopia almost as much as economy. The US-China AI science race (Wall Street Journal), PM Wong's Singapore speech on fragmented governance, Apple training a China-specific model with Alibaba, Tencent Cloud planting its first Malaysian region, BRICS coordinating on AI – the bifurcation of the AI stack is accelerating.
Talent and regulation are the quieter drivers, but the 90%-of-executives-say-AI-hasn't-boosted-productivity finding (Fortune/c55) sits at their intersection. Jobs cut, gains absent, workers anxious. The RBA flagging AI mania as an inflation risk is a distinctly Australian flavour of the same concern.
This is where it gets genuinely interesting. ANZ's Distribution Gap this week is -11 points (Futurama ahead of Dystopia by 11 points), against a rolling four-week figure of -7. That's the most optimistic regional reading in the scan – and it's held for a month.
What's underneath it? A mix of practitioner pragmatism and institutional action. The National AI Centre director urging agencies to build on their own terms. Telstra tying executive bonuses to measurable AI impact. An AI legal tool winning its first Fair Work case. A parliamentary inquiry into SME AI barriers. AMD's rack efficiency gains intersecting neatly with NSW's new data centre rules. These aren't hype stories – they're governance stories, adoption stories, accountability stories.
The shadow is real too. NSW councils aghast at $50B data centre pipeline overrides. Women holding just 15% of AI leadership roles. The 'AI refuser' forced out of her dream job. Canva's $11B valuation wipe and the seat-cycling fraud. The RBA's inflation warning. ANZ's Paperclipalypse reading at 19% – identical to the global figure – tells you the risk awareness is there.
But the Futurama reading of 32% (against a global 24%) suggests Australian and NZ organisations are, on balance, treating AI more as a tool to govern than a force to surrender to. That's not complacency – it's a relative advantage worth protecting.
North America's Distribution Gap is +18 this week, against a rolling +14. That's the most dystopian regional reading by a wide margin, and it's been structurally elevated all series.
The reason is obvious when you look at the story clusters: the OpenAI/Nvidia data centre saga (five stories, all Dystopia), the Broadcom $60B+ debt deal, the Anthropic IPO targeting SpaceX-scale valuations, the Stripe/OpenRouter consolidation, Groq's pivot to Nvidia-backed neocloud. North America is where the infrastructure bets are being made, where the capital is concentrating, and where the political backlash – data centres reshaping governors' races in Pennsylvania, Ohio and Texas – is most visible.
The Fizzle signal is also strong: Dario Amodei conceding the industry hasn't delivered on its promises (multiple story groups), the AI backlash analysis, the 90%-no-productivity finding, VCs urgently reassessing 2024's AI portfolio bets. North America's discourse is simultaneously the most euphoric and the most disillusioned. Both can be true when the money is this large.
Europe's Distribution Gap at -3 (rolling +1) puts it close to neutral, which is roughly where it's been all series. The week's European signals are a mix: EU mandatory AI content labels that may make deepfakes harder to spot (Paperclipalypse), Callosum's $100M model-routing seed raise (Futurama), ASML and others backing Mistral's data centre buildout (Dystopia), Flip's frontline worker AI platform (Futurama), NATO drone deployment debates (Paperclipalypse). Europe is doing what Europe does – trying to govern the thing while the thing keeps moving.
Singapore, Hong Kong and the broader APAC signals (77 stories, weighted at just 5%) show the sharpest Dystopia reading of any region: 44% this week, rolling 41%. China's AI infrastructure buildout – Alibaba, Baidu, CXMT, Biren, UCloud/Hygon, 70+ embodied AI training grounds – is a state-level concentration play without precedent in the series. Singapore's PM Wong framing AI governance as captured by great-power rivalry is the clearest articulation of that dynamic from any head of government this week.
The dog that hasn't barked: genuine, at-scale evidence of AI productivity gains reaching mid-market or SME organisations. We have vendor claims, pilot announcements, and the occasional Westpac or Singapore Airlines showcase. We don't have independent data showing productivity lifting across a broad cross-section of organisations. The Fortune story – 90% of executives, no boost – is the most reliable signal we have on that question, and it points the wrong way.
Also absent: any serious discourse about what happens to the organisations that can't afford to play. The AI 'death zone' story (Fortune) names it – Chinese open-weight models at the bottom, US frontier labs at the top, corporate strategies caught in the squeeze – but the policy response to that squeeze is almost invisible.
Here's what I'd tell a CEO or CIO reading this over the weekend. The global narrative is stuck in a loop: infrastructure concentrating, productivity unproven, trust eroding, costs rising. You can't fix the global picture. But you can position your organisation on the right side of the Distribution Gap.
The ANZ data suggests the organisations doing best are the ones treating AI governance as a competitive advantage, not a compliance burden. Telstra's bonus structure, the NAIC's agency-sovereignty push, law firms prioritising accuracy over speed – these are choices, not accidents. The question for Monday is whether your AI programme has a measurable accountability structure, or whether it's still running on vibes and pilot announcements. If it's the latter, the ROI pressure Gartner is flagging for 2028 is already arriving at your boardroom door.
The articles most responsible for bending the five worlds this week.
A major study found that nine out of ten business leaders say AI hasn't actually made their teams more productive – but companies are still letting people go anyway. That means workers are losing jobs while the promised benefits haven't shown up yet, which is a big warning sign for anyone trying to justify their AI spending.
Scammers found a clever way to steal AI credits from Canva's platform by repeatedly signing up free accounts and reselling the tokens before Canva could stop them. It's a real example of how quickly people find ways around AI safety systems – and it cost Canva both money and an $11 billion hit to its valuation.
Research firm Gartner is warning that even though individual AI requests are getting cheaper, building and running AI agents will cost five times more by 2028. This matters because many businesses are betting on agents to save money – and the maths might not work out the way they're hoping.
The NSW government changed planning rules to push through a huge wave of AI data centre construction, overriding local councils who objected on environmental and community grounds. It's a clear sign that the benefits of AI infrastructure are going to a small number of big companies while local communities are left to deal with the noise, power use and land impacts.
A senior government adviser is arguing that Australia risks becoming just a landlord for overseas tech giants' data centres, rather than building real AI capability of its own. The argument is that with the right policy, Australia could capture a much bigger share of AI's economic rewards – a hopeful vision that matched the week's most optimistic regional reading.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free