Someone said it on stage at the India DPI Summit, almost in passing: "The future is already here — it's just not evenly distributed." It is William Gibson's line. The room nodded the way rooms do at a good quote, and moved on. I couldn't move on — because most people hear that sentence as an observation, a neat way to say the world is uneven. I hear it as a business model. The uneven part isn't the footnote. It is the entire opportunity. And there was something fitting about hearing it in a room convened to talk about digital public infrastructure — the one thing India has built specifically to make the future land evenly.
The Capability Already Exists. The Access Doesn't.
Look at AI right now and Gibson's line stops being poetic and starts being literal. The capability is here. Models that write production code, run analysis, draft contracts, and answer in any language — that future arrived. It is not coming; it landed. But walk into a twelve-person manufacturing unit in a tier-3 Indian city and ask how much of that capability has actually reached them. The answer is roughly none. Same planet, same year, completely different decade — depending on which side of the distribution gap you are standing on. That gap is not a temporary glitch on the way to everyone catching up. The gap is where value lives.
When Intelligence Is Free, Distribution Is Everything
Here is the shift I keep coming back to: intelligence is becoming free. Not cheap — free, directionally. The raw ability to generate, reason, and produce is collapsing toward zero cost. And the moment a thing becomes abundant, the scarce thing moves somewhere else. It moves to whoever controls how that abundance reaches people: whoever owns the trust layer, the identity layer, the attestation that something is real, the infrastructure the capability depends on to actually get into someone's hands and do something useful. That is the part that doesn't distribute itself. Capability spreads fast. The plumbing that lets capability land — safely, accountably, in someone's own language, inside their actual workflow — spreads slowly, and is built deliberately, by people who decided to build it. This is the dependency layer, and it is where value accrues when intelligence becomes free.
India Already Wrote This Playbook Once
What made the DPI room the right place to hear that quote is this: India is the one country that has already built a distribution layer for capability and shipped it to more than a billion people. UPI didn't invent payments. Aadhaar didn't invent identity. DigiLocker didn't invent documents. What they built was the plumbing that made the capability land — in any language, in any village, inside someone's own phone. That is the template. The DPI pattern maps directly onto the agent economy: the AI distribution layer — trust, identity, attestation, language, workflow — is the next layer in a stack India has been extending for a decade. The same instinct that built India Stack is the one that closes the AI gap.
63 Million Reasons This Matters
India has roughly 63 million MSMEs. They are not waiting for the future to arrive — it has already arrived for the labs and the enterprises. They are waiting for it to reach them, in a form they can use, in a language they speak. You can read that as a tragedy. I read it as the largest distribution problem and the largest opportunity of this decade sitting in the same place. The businesses that win the next ten years won't be the ones with the most advanced models — everyone will have those. They will be the ones who solved distribution: who built the dependency layer that turns "the future exists somewhere" into "the future works here." That is also why the agent trust layer is India's next move.
The Takeaway
Gibson's line is usually quoted with a shrug. I'd flip it. The future being unevenly distributed isn't the problem to lament — it is the work to do. Capability is the easy part now. Closing the distance between where the future already exists and where it doesn't yet — that is the trillion-dollar job, and it is the whole game. Chapter 6 of The AI Agent Economy makes the full case for why distribution, not model-building, decides who wins. So the next time someone drops that quote on stage and the room nods, ask the harder question: who's building the distribution?
Frequently asked
What does 'distribution is the business' mean for AI?
It means the durable value in AI is shifting away from raw capability and toward the layer that delivers capability to people who can use it. As models that generate code, analysis and language become abundant and near-free, the scarce — and therefore valuable — thing becomes the trust, identity, attestation, language and workflow infrastructure that turns 'the future exists somewhere' into 'the future works here.' Capability spreads on its own; distribution is built deliberately.
Why is the distribution gap an opportunity rather than a problem?
Because the gap is where value accrues. The capability already exists in the labs and large enterprises, but it has barely reached India's roughly 63 million MSMEs in a form and language they can use. That distance between where the future exists and where it doesn't is not a temporary glitch on the way to everyone catching up — it is the largest distribution problem and the largest business opportunity of the decade, sitting in the same place.
Related reading
From the same content cluster.
Cluster pillar
India Built UPI for 1.3 Billion. The Agent Economy's Trust Layer Is Next.
Why India's infrastructure DNA positions it to own the layer that makes capability land.
Related post
When Intelligence Is Free, Where Does Value Go?
The economics of abundance — when generation costs collapse, value migrates to the scarce dependency layer.
Related post
Why India's DPI Template Maps Directly to the Agent Economy
Aadhaar, UPI and ONDC are the working proof that open distribution at population scale is the winning pattern.
Glossary
Glossary: Dependency Layer
Canonical definition — the infrastructure that decides whether capability actually reaches anyone.
From the book
The AI Agent Economy — Book 1
The full thesis, developed across ten chapters and fifteen falsifiable predictions.