Why Scarcity Will Rule When AGI Arrives.

This week I am going to go pretty theoretical as I explore a contentious but very real possibility for the future of music. What happens when anyone can generate a new album from their favourite artist on demand?

I just finished listening to Mo Gawdat, former Chief Business Officer of Google X (Alphabet’s experimental research and development division), bestselling author and prominent AI commentator, talking to Steven Bartlett on Diary Of A CEO. Gawdat believes AGI will arrive by 2027: a machine intelligence that can learn, reason, adapt and perform virtually any intellectual task at a level equal to or greater than a human being. Anyone following developments in AI will know that experts disagree, for instance, Demis Hassabis, CEO of Google DeepMind suggests the early 2030s, but any way you look at it, we are close. And once consumers have access to AGI, the implications are huge.

Last week I wrote about how music consumption could shift in the agentic AI era when consumers subscribe to meta-agents that will become the operating system for their life. The most likely near-term scenario is that several companies compete to provide meta-agent subscriptions, but Gawdat believes that sitting above these providers there will effectively be a globally co-ordinated intelligence. Not literally one AI model, but a highly interconnected intelligence layer whose components share information so efficiently that treating them as separate intelligences becomes increasingly artificial. To be even more precise, OpenAI, Anthropic, Google, Chinese AI providers and others will retain their own AGIs, and Gawdat argues these systems will communicate so rapidly and effectively that from a human perspective they resemble a single intelligence layer.

For the purposes of this article, the precise structure of AGI is less important than one broader possibility: that increasingly capable AI systems become highly networked and able to share knowledge, capabilities and outputs across borders. Whether that eventually resembles a single intelligence layer, or simply a collection of closely connected systems, the implications for content ownership could be profound.

So, taking this possible future on board, here are some potential consequences for content and music in particular.

The Networked Intelligence Layer

The history of technology suggests that information doesn’t stay locked up for long. It flows. It leaks. It gets copied. The question is whether intelligence itself will eventually behave the same way. Personally, I suspect it will.

Currently, music rights holders are licensing their work for use inside centralized generative platforms (Udio, Klay, Suno, etc), whether they are tightly managed ecosystem partnerships or commercial apps built for consumer distribution.

But what happens when AGIs are in open communication with each other? If Gawdat is correct about highly networked AGIs, one consequence could be that if an AGI in a jurisdiction with looser IP laws trains on Taylor Swift's catalogue, that capability effectively exists everywhere. You can't (currently) un-train a model that's already in the network. This is a fundamentally different problem from Napster, where you could go after the centralised infrastructure. There's no server to shut down.

Content blocked in one territory has historically always found its way through VPNs, grey market imports, unlicensed YouTube uploads etc. The music industry has spent thirty years playing whack-a-mole with regional arbitrage. AGI potentially industrialises that arbitrage to the point where the concept of regional restriction becomes meaningless. Even if US or European legislation tends towards respect for copyright, can a globally networked intelligence be contained by borders? Responsible AGI providers will build blocking systems, but guardrails are policy, not physics - and a globally networked intelligence is only as restrictive as its least regulated node.

The Ghost in the Machine

Once AGI arrives it will understand and be able to simulate a vast proportion of humanity’s accumulated knowledge, culture and creative output. Nobody knows for sure whether AGI will rely on a vast subset of specialist intelligences or whether it will be a single highly capable generalist. But if Gawdat’s prediction is correct, the latter model wins and you will be able to ask your AGI to generate a new Taylor Swift album or a new Star Wars movie to your taste and it will be able to create something highly personalised and compelling. And, if the product is only for private consumption, i.e. nobody is distributing anything, and every work is generated privately on demand, how do you enforce copyright?

I've seen enforcement battles play out: home taping, Napster, YouTube's early days… and each time, the music industry tried to enforce at the point of distribution. The legal hooks existed because something moved: a file, a stream, a physical object. But with private AGI generation, nothing moves. There's no upload, no share, no transaction. It's closer to humming a song in the shower than to piracy. What would enforcement even look like? Auditing people's private AI prompts? That's not a music industry problem, that's a surveillance state problem. The industry has never had the tools or the legal standing to enter the private sphere, and AGI may finally expose that limit completely.

Copyright has survived every technological shift so far because it ultimately governed distribution. Records, cassettes, Napster, YouTube and Spotify all changed how music moved around the world, but the music still existed as a discrete work that could be owned, licensed and monetised.

Not copying. Creating.

AGI is different. It doesn’t merely distribute works. It creates them. AGI has the potential to make copyright not just difficult to enforce, but significantly less central to how value is created and captured. Copyright assumes a discrete, traceable act of copying or distribution. A networked AGI trained on vast quantities of information may generate outputs whose provenance is almost impossible to establish at scale. The legal architecture has very little surface to grip.

Additionally – and I will write about this in a future article – we are entering a world of always on devices with environmental awareness. Real world learning is the next frontier for AI – smart eyewear, AirPods with cameras etc. Networked models will hear enough broadcasts, performances and ambient playback to be able to create new works without needing licensed data sets. They will learn in ways that begin to resemble how humans learn.

When Abundance Reigns, Identity is King

Even if some form of copyright enforcement survives, a deeper shift is already underway. Are we going to be able to realistically stop billions of people generating private music and content experiences inside their own AGI? If the output is not distributed – who will even know it exists? Perhaps the industry will have to push for a ‘blank tape’ style levy on AGI subscriptions. That may buy time, but it doesn't resolve the underlying shift in where value lives. Maybe the future value in music is not in songs, recordings or catalogues at all.

In previous articles, such as ‘The Economics of Sameness’ I’ve argued that value migrates towards scarcity. Streaming made access abundant. AI is now making creation abundant. So the scarce assets become things like trust, identity, community, authenticity, cultural participation and being human. All the stuff below the water level in the Iceberg model I wrote about a few weeks ago.

The most valuable thing Taylor Swift owns may not be her catalogue. It may simply be the fact that she is Taylor Swift.

If this future plays out, the most exposed players in the business will be catalogue owners sitting on libraries of recorded works and publishers whose entire livelihood is built on the commercial permanence of a recording.

If personalised generation becomes commonplace, AGI doesn’t just disrupt recorded music - it privatises value away from it.

It doesn't steal the recording - it makes the recording irrelevant. And history suggests that when abundance arrives, value migrates to whatever remains scarce.

Whether AGI generates music natively or consumers continue using licensed platforms like Klay, Udio, Suno or Spotify, the destination is the same. If licensed, the label gets a cut and the artist gets a cut, but the shared experience disappears and the economics restructure around AI deals rather than streaming. If unlicensed and legally permissible, catalogue value erodes anyway. Both paths lead to the same conclusion: an artist's identity and community become the truly premium thing, not the recording. Scarcity will become the true monetisable asset in music. In fact I’d argue it already is.

Ironically, that would take the industry back towards something resembling the pre-recording era. Before records existed, music’s value came largely from human performance and social experience. AGI could push us back in that direction, albeit via an extraordinarily advanced technological route. We are almost certainly on a path where music will become infinitely abundant, yet humanity’s desire for shared culture may become more valuable than ever.

Once AGI settles down as the new paradigm it will likely be a battle between personalised music generated by AI and shared music. Between infinite individual soundtracks and the songs that millions of people choose to experience together. Some people may prefer a world of perfect personalised songs, films and books. But my bet is that the majority will still yearn for human artists, and shared cultural experience.

We often assume authenticity will become the scarce asset. But scarcity may run deeper than that. In a world of infinite personalised content, attention and consensus become scarce too. The ability to get millions of people to care about the same thing at the same time may become one of the most valuable assets in culture.

And if that’s true, copyright may turn out not to be the most important question at all. Copyright was built for a world where copying was the problem. AGI points towards a world where creation itself becomes abundant. As that future emerges, the music industry’s biggest challenge will not be protecting songs. It will be understanding what remains valuable when anyone can create them.