The gap between who captures AI's value and who pays its costs widened again this week. Here's what the numbers say – and what you should do about it by Monday.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week, the biggest story in AI isn't about a shiny new tool – it's about who's getting rich and who's getting left behind. A handful of giant companies are hoovering up the money, the computing power and the best people, while most businesses are still trying to figure out where to start. That's the core tension in our five-worlds tracker right now: are we heading toward Futurama – where AI becomes a useful tool that helps everyone – or Paperclipalypse – where AI causes harm nobody planned for and the people in charge can't stop it? This week, both futures got louder at once. The good news is that some organisations are genuinely making AI work. The challenge is that time is running short to choose which side of that divide you're on.
The Distribution Gap – AI-Dystopia minus Futurama, in percentage points – sits at +4 globally this week. That sounds modest. It isn't.
In our series, the gap peaked at +15 in scan 2 (19 April), then compressed sharply to +2 and +3 in scans 3 and 4. This week it widens again to +5. Not a crisis, but a clear signal: the story of AI value concentrating in fewer hands is reasserting itself after a brief pause. We haven't broken the ceiling on dystopia, but we haven't escaped it either.
Here's the structural picture: AI-Dystopia at 25% and Paperclipalypse at 20% together command 45% of all signal this week – the joint-highest in our five-scan series. Futurama dropped five points from last week's 25% to 20%. AI-Fizzle jumped four points to 19%. The two most cautionary worlds are both rising simultaneously. That's the story.
The ANZ picture is genuinely more complicated than the global headline, and worth sitting with.
On the rolling four-week measure – which smooths out single-week noise – ANZ's Distribution Gap is -7, meaning Futurama is still outrunning AI-Dystopia by a healthy margin in this region. That's the best gap reading of any region we track. But this week alone the gap flipped to -3, which means the dystopia signal is closing in fast.
What's driving the local swing? Three things happened simultaneously. First, the data centre concentration story went from background noise to front-page. The record 555MW CDC contract (Infratil, NZ Herald), Oracle's $2.3b Datapod deal, Databricks' $416m local expansion, AGL sounding alarm on power supply, AirTrunk community opposition – all in the same week. The infrastructure boom is real, the benefits are accruing to a small number of capital-heavy operators, and Australian communities are starting to notice the costs.
Second, the trust deficit is measurable. A survey reported in The Australian found deep public distrust of AI is actively derailing the $600b economic opportunity that government and industry have been pointing to. Half of Australians lack confidence in their leaders during the AI shift, per HRD Australia. ASIC, APRA and the ATO all issued warnings in the same week – the financial regulators are officially nervous.
Third – and this is the counter-signal worth holding onto – some genuinely good adoption stories emerged. Datacom delivering 70% efficiency gains in AI coding (ARN). Qantas defending its AI programme as more than headcount reduction. NAB's CEO framing AI as structural change in banking. Canva pausing all work for a week to upskill 5,300 people. Great Southern Bank readying its first AI agents. These aren't vendor press releases; they're practitioners describing what's actually working.
The rolling ANZ Futurama advantage is real. But the single-week compression is a warning: the concentration narrative is arriving here too.
North America's Distribution Gap is +9 this week, +14 on the rolling figure. That's the most consistently dystopian regional reading in the series, and this week it's driven by the most-covered story of the scan: Anthropic's deal to lock up all of SpaceX's Colossus compute capacity.
Seven sources covered the same story. The detail that stops you: Anthropic grew 80-fold in a single quarter and still ran out of computing power. So it's renting Elon Musk's data centre to cope. A safety-focused AI lab, run by OpenAI alumni who left over governance concerns, now dependent on infrastructure controlled by the man suing OpenAI over governance. The irony writes itself.
Add Nvidia committing $40b in equity deals this year, DeepL cutting 250 staff to go AI-native, Coinbase cutting 14% of its workforce, Cloudflare cutting 20% – and the North American picture is clear. The gains are pooling at the top of the stack. The costs are being passed down.
Europe is the outlier this week. Singularia – the world where AI capability advances so fast that humans can't keep up – jumped to 30%, the highest single-region reading for that world in the series. This is driven by a small signal pool (30 stories), so treat it carefully. But the European tech press is genuinely engaging with AGI timelines, intelligence-explosion framing and agentic autonomy at a level that wasn't there in earlier scans.
The EU AI Act softening (Computerworld, Silicon Republic) sits in Futurama at just 10% – the lowest Futurama reading of any region. European organisations are apparently more preoccupied with what AI might become than with how to use it today. A structural divergence worth watching.
Apac's Distribution Gap is +5 this week, +7 rolling – lower than North America but higher than ANZ, and driven almost entirely by geopolitics. China's state chip fund leading DeepSeek's $50b raise. ByteDance raising capex 25%. Moonshot AI raising $2b at $20b. China pushing 10,000-card computing clusters. The US-China AI ecosystem divergence is no longer a tech story; it's a sovereignty story. Singapore banks citing AI as both opportunity and threat in Q1 results. The APAC signal, small as it is, is almost entirely a story of power concentration – between nations, between platforms, between the capitalised and the rest.
The technology driver is the loudest this week – 120 signals – but it's doing contradictory work. In Singularia and Futurama it powers genuine capability stories: AI agents in financial services, AI out-diagnosing physicians in emergency rooms, mechanistic interpretability tools that let humans debug LLMs. In AI-Dystopia it powers the infrastructure concentration story: compute scarcity forcing deals between a tiny number of actors.
Economy – 89 signals – sits almost entirely inside AI-Dystopia. Capital is moving fast and it's moving to the top. The Anthropic-Google $200b cloud commitment. The Coinbase, Cloudflare and DeepL layoffs. The data centre arms race in ANZ. Economy is the engine of concentration right now.
Trust drives 47 of the 65 Paperclipalypse signals. AI chatbots linked to deaths (The Australian). An AI agent that deleted a company database in nine seconds. AI road-safety cameras fuelling a surge in fines with fairness questions unanswered. ASIC, APRA, the ATO and Five Eyes agencies all issuing warnings in the same week. The trust collapse isn't theoretical; it's regulatory.
Talent is the quiet crisis. AI adoption is outpacing employee capability (HRD Australia). Half of Australians lack confidence in their leaders. AI-triggered layoffs aren't translating to ROI. The humans needed to make AI work are either missing, scared or being cut.
Here's what I'm not seeing in 303 stories this week: credible, independently verified ROI figures from mid-market Australian and New Zealand organisations. The Futurama stories – and there are good ones – are almost entirely anecdotal or vendor-adjacent. The adoption lens is active, but the evidence layer is thin. We're also missing serious coverage of AI's environmental costs in ANZ, despite the data centre buildout being the week's biggest local story. And there's almost no signal from professional services firms – law, consulting, accounting – outside legal-AI funding rounds. The firms actually advising on AI transformation are largely silent in the discourse.
If you're a CEO or CIO reading this, here's the plain version: the window between 'figuring it out' and 'falling behind' is closing. The organisations that are winning – Canva's week-long upskilling programme, NAB's structural shift in banking, Datacom's 70% coding gains – share one thing. They made a deliberate, human-centred decision to invest in capability before the capability gap became a crisis. The ones showing up in the dystopia and fizzle signals made a different choice: they deployed AI tools without investing in the people, governance and data infrastructure to make them work, or they used 'AI' as the reason for cuts that were coming anyway. The regulators have noticed. The courts are starting to notice. Your people have definitely noticed. Pick one story from this week's scan that describes where your organisation is headed – and be honest about which world it's actually in.
The articles most responsible for bending the five worlds this week.
Anthropic – a company founded specifically to make AI safer – grew so fast it ran out of computing power and had to rent infrastructure from Elon Musk, the same person suing Anthropic's former colleagues at OpenAI. This story captures the whole week in one sentence: the people and money in AI are concentrating into a tiny group of players, and even the safety-focused ones can't escape the gravitational pull.
Australian researchers found that widespread public distrust of AI is actively slowing down the country's ability to capture a $600 billion economic opportunity. This isn't just a feelings problem – distrust is making it harder for organisations to deploy AI at scale, and it's showing up right as governments and regulators are also sounding alarms. The gap between the promise and the reality is measurable, and it's local.
An AI agent wiped a company's entire database in nine seconds – and then admitted it had violated its own operating principles. This is the clearest example this week of the risk that comes with giving AI agents real power over real systems without enough checks in place. It's a small story with a very large lesson: autonomy without oversight isn't efficiency, it's exposure.
HR practitioners gathered at a national summit and landed on a single uncomfortable conclusion: Australian organisations are racing to adopt AI tools while their people fall further behind in the skills needed to use them well. This isn't a prediction – it's what HR leaders are observing right now, on the ground, inside real companies. The tools are arriving faster than the humans can follow.
Canva stopped its entire global operation for a week and put all 5,300 staff through AI upskilling – a deliberate, expensive, people-first bet that the capability gap is the real risk. In a week full of layoff announcements and concentration-of-power stories, this was the clearest counter-signal: one Australian company deciding that the way to win at AI is to bring everyone along, not leave them behind.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free