I am a product marketer. My world is go-to-market strategy, user journeys, and positioning. For most of my career, if I had an idea for a product flow or wanted to test a concept with users, I had to brief a developer, wait, iterate through feedback loops, and hope the final output matched what I had in my head. That process could take weeks. Now with AI, I can build the prototype myself without coding. Not a rough sketch, an actual working prototype, sometimes in the same afternoon I had the idea.

And what used to be separate workstreams, building the product, documenting it, creating the marketing materials, crafting the use cases, now feed into each other. I can build a prototype, screenshot the flow, write the copy around it, and have a landing page up before the end of the day. That kind of speed used to require a team.

I am not alone in this. Developers I know are writing their own documentation in minutes instead of hours. The walls between disciplines that used to define what each of us could and could not do are coming down fast. And with them, the barriers around who can access jobs, capital, and the tools to build something from scratch.

But what made that possible for me is part of a much bigger shift happening right now, and not everyone is going to benefit from it equally.

The stakes are bigger than most people realise

In mid-March, Jensen Huang stood on stage at GTC 2026 and told the world he could see at least US$1 trillion in AI infrastructure spending through 2027. Anthropic CEO Dario Amodei, speaking recently in Bangalore, said AI adoption in India has doubled in just three to four months. Andrej Karpathy, one of the people who helped build the foundations of modern AI, admitted he has not typed a line of code himself since December.

Also Read: Ethical implications of using AI in hiring

These are not just impressive numbers; they are signals that the rules of the next economy are being written right now, mostly by a small number of players with very large infrastructure budgets. In simple terms, tokens are the units AI runs on and gets paid for. Every response, every piece of analysis, every bit of generated content is measured in tokens, and a gigawatt data centre costs around US$40 billion before a single chip goes in. The countries and companies that can build those factories will define the cost of intelligence for everyone else. In this new economy, your token budget is becoming as critical as your cloud spend, and if that cost gets set by a handful of players in one part of the world, everyone else ends up buying at a price they had no say in.

This is what infrastructure inequality looks like in practice, and it is worth understanding what game you are actually playing.

The part that gets missed

Dario made a point in Bangalore worth paying attention to. He said Anthropic does not come to markets like India looking for consumers; they want to work with local builders who actually understand their own market. Every two or three months, a new model release opens up something that was not possible before.

OpenClaw, the open source agentic AI framework that Jensen described as the most downloaded open source project in history, surpassing Linux in weeks, makes this even more concrete. Karpathy called it the operating system for agentic computers, the same role Windows played for the personal computer. A developer anywhere in the world can now build on the same foundation as one in San Francisco.

The infrastructure layer requires billions to build, which means it is dominated by players with the deepest pockets, but the application layer, meaning the tools and products built on top of AI, is still wide open. That is where the real opportunity sits for founders and builders in this region. That window is real, but only if you know it is there.

The real barrier is the on-ramp

Most people assume the barrier to AI adoption is access, that if you just had the right tools, you would be fine. But that is not what the data shows. 64 per cent of Southeast Asian sellers cite high costs and time as major obstacles, and while 41 per cent of SMEs say they are adopting AI, only five per cent are actually using it in a meaningful way. The barrier is not access. It is the on-ramp.

A small logistics company in Southeast Asia, with five people and no dedicated tech team, recently started using AI to handle customer communications, route queries, and generate weekly ops summaries. What used to need a part-time coordinator now runs largely on its own, and the founder ended up not hiring the person she had budgeted for. That is what the tool working as promised actually looks like.

But getting there took three weeks of trial and error, a developer friend who helped with setup, and the willingness to push through a lot of frustration. Three weeks and a developer friend are not things everyone has.

Also Read: A new era of automation: Establishing best practices for intelligent automation and generative AI

This is the gap that does not show up in press releases or keynote slides. The tools exist, but most people still cannot figure out how to get started. Karpathy talks about spending sixteen hours a day in what he jokingly calls AI psychosis, basically an obsessive state of directing multiple AI agents at the same time, each working on different tasks, while he reviews, adjusts, and keeps them all moving. That is what mastery looks like right now, and that gap does not close on its own.

What actually needs to change

So what does closing it actually look like? Some of it is already happening. Google’s Stitch update in March 2026 means a founder who cannot afford a designer can now generate a full UI, interactive prototype, and design system in under an hour, for free, with no design skills required. Figma’s stock dropped 8.8 per cent the day the update was announced. The market saw the shift before most people did, and this is exactly the direction things need to keep moving: tools that start from what you want to achieve rather than assuming you already know how to build it.

That is why I think this category of operator-first tools matters, including what we are building with Fuseful Workflow Studio at Morpheus Labs. Most automation tools still assume you know how to build the system. Fuseful starts from the business outcome instead, built with operators in mind, not engineers.

But tools alone are not enough. The average enterprise now runs more than ten AI applications, yet 76 per cent report negative outcomes because the tools do not connect, and nobody was trained to use them together. Anthropic has a team they internally call the Ministry of Education, and that is not a trivial signal. Companies serious about equity need to treat capability-building in their users the same way they treat feature development, not as an afterthought but as the actual product.

And the last piece sits with local builders. Dario is right that Anthropic cannot and should not build for every vertical. The real opportunity for domain-specific, market-specific, culturally-grounded applications sits with the people who actually know those markets. Funding those builders and not cannibalising them when they find success is what building equity by design actually looks like in practice.

The big labs will keep building, and the infrastructure will keep scaling. That part is not really up for debate anymore. What is still up for debate is what gets built on top of all of it, who gets trained to use it, and who gets funded to try.

Those decisions do not belong to Jensen or Dario. They belong to every founder, operator, and builder in ecosystems like this one. And we are still early enough to get them right.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The post The AI wave is real, but it won’t lift everyone appeared first on e27.