Time to rip the band aid off.
I’m sharing this message today because I think it’s important for everyone to hear this.
The age of AI subsidy is coming to an end. And the question is, what are we doing about it?
I measured the total token I used for Claude in February. In 2 weeks, according to calculations the amount I would have used in API costs would have been 4.5k USD.
Claude Max subscription = $200/month
I started paying for two Claude Max subscriptions in March. Every day, I would be running multiple (5-6) parallel workflows and squeezing every token of usage from my 5-hourly rate limits and weekly limits.
Safe to say, my usage per month easily exceeds $10K. Yet I am only paying $400 per month.
They're heavily subsidised. By the venture capitalists and investors all banking on the coming age of AI super intelligence.
I would have been spending >10k/mo in API costs if not for my Claude Max subscription. This is why all the labs are losing money. All these subsidy is not sustainable.
Which should tell you something: the age of AI subsidy is coming to an end.
Those casual conversations with Claude currently costs $20/month when it should have been >$100 per month. Soon, that will happen.
Yet people are taking everything for granted right now.
Not building anything with AI.
And for the most part, not even learning AI at all.
They are not taking advantage of the most paid for public subsidy in the history of the Earth.
The idea of “AI is here to stay” is being conflated with “AI will always be cheap,” and so to the common person the thinking goes: “I am too busy right now so I will learn this when I have the time.”
Or worse, with more naivete, “AI can’t do what I do.”
Meanwhile, AI is getting smarter day by day. The apps they are powered by are getting massively intuitive. And what most people aren’t seeing is that these are all plays to lock you into their ecosystem.
To make you so dependent on them, that when they stop being so generous, the pain of leaving exceeds the pain of staying exactly where you are.
Anthropic is shipping almost every other day. OpenAI is closely following suit. And somewhere in the open source community, OpenClaw maintainers have shipped 5 new major updates in just the previous week alone.
The big labs are racing - tripping - over each other to capture the largest market share that they can by practically giving irresistible handouts and free stuff every single day.
So what happens when the age of subsidy ends, the labs have gotten everyone hooked on their $20/month designers (ahem - Claude design), locked everyone into their ecosystem, and decides its now time to placate the venture capitalists whos been paying for all this?
The most advanced models will be unaffordable and out of reach
The mediocre models will be expensive, making it hard to experiment
The education around AI use will be institutionalized for “safety” purposes
The best AI systems will be used for governance, enforcement, and reserved for the ultra wealthy to enrich themselves even more.
The result: We will have a permanent underclass and a permanent overclass.
(If this doesn't scare you - it should! It certainly scares me!)
We still have hope right now.
AI is still ‘cheap’ - though it is getting more and more expensive.
You can still find many opensourced projects because everyone is still very excited (and generous) about this right now.
Cost of experimentations is close to nothing if not cheap.
But the window is closing really soon on who can actually build what with AI.
A hundred years ago, the wealthy own lands.
A hundred years from now, the wealthy own datacenters.
Speaking of datacenters, this morning I watched a video about how many of them are being built all across the US.
Acres of land are being stripped and rebuilt around neighborhoods. Electricity bills are doubling and tripling. And in return, these datacentres giving hundreds of new jobs while taking away thousands.
But what scares me the most isn’t this sense of impending danger and urgency.
It’s how relaxed - and excited - people are about AI.
A lot of very influential AI bros are very optimistic right now about the future. Guys with hundreds of thousands of followers sharing the news, excitement and possibilities around AI every single day.
What's deeply troubling for me is how everyone who is listening to them is feeling this sense of immense hope about this wonderful future AI is going to bring them.
The AI bros will make money. Their followers will not.
More than 10 years from now, perhaps, we will have that utopian future everyone’s been promised. Healthy, wealthy, and happy lives for everyone. But for many years before that, I can only see something akin to a dystopian future.
We must prepare.
Personally, I just don’t want to end up being on the wrong side of it. I don’t want to see my community being on the wrong side of it, either.
People need to wake up and take this with a sense of immediate urgency to collect capital, skills, and sellable assets with the current window - the ‘golden age’ of AI subsidy.
What we do in the coming months will determine everything.
I am hosting an all-hands-on-deck call I am hosting next Saturday, 9pm Malaysian time for my online community.
The topic: “Agents 101”
You’re invited to join.
If you join, I will catch you guys up to speed on what this agent thing is really all about, what it can (really) do for your business, and how you can leverage Agents without it burning your business down to the ground.
Suitable for everyone who hasn’t yet started utilizing agentic workflows for their business.
[Archived link retired.]
FELIX