Maybe I'm the guy at the party who keeps pointing at the fire exit while everyone else is doing keg stands. I'll own that. But I've been in go-to-market long enough to know that when something feels too good to be true, it usually is. And right now, the economics of AI-powered building feel very good.
There's a whole generation of GTM engineers and demand gen practitioners who have quietly, gleefully cut the SaaS cord. They're spinning up internal tools in Base44. They're building applications in Lovable and Codex. They're replacing four-figure monthly software subscriptions with AI workflows that cost them almost nothing in tokens. They feel like they found the cheat code in the matrix. And honestly? I get it. I've done it. It feels incredible.
But here's what nobody wants to say out loud: we are living in the subsidized token era, and it is ending faster than you think.
The Uber You Already Forgot
Remember the early twenty tens? Everybody was riding around like transportation had been solved. Eight dollar rides across town. Venture capital quietly picking up the tab while companies bought market share with artificially suppressed prices. It felt like a permanent shift in how the world worked. It wasn't.
There was even a scrappy little competitor called Sidecar, launched in 2012, that actually pioneered some of the most interesting ride-sharing technology of the era, including letting riders set their own pricing. Genuinely innovative. Shut down in 2015 after burning through its funding trying to compete against Uber's war chest. Not because the product was bad. Because the economics of the moment were being dictated by who had the deepest pockets, not who had the best idea.
The best idea in a subsidized market doesn't always win. Sometimes the check just stops coming.
When Uber and Lyft finally decided they needed to actually make money, prices went up. Surge pricing became the norm. Everyone who had restructured their life around subsidized rides got a very rude awakening. The infrastructure they'd come to depend on repriced itself overnight. Now swap rides for tokens.
What's Actually Happening Right Now
Anthropic quietly removed Claude Code from the twenty dollar Pro plan in April. Access now starts at a hundred dollars a month. OpenAI launched a new hundred dollar tier the same month, deliberately positioned against Claude Max. Enterprise pricing across both platforms is shifting toward usage-based billing with upfront consumption commitments. The direction of travel is unmistakable.
These companies burned enormous capital to acquire users, build habits, and establish dependency. Now they need to become actual businesses. That transition is not going to be painless for the people who built on the cheap side of those economics.
We've Seen This Pattern Before
This isn't just a tech story. It's a resource story. And we have a deep, well-documented history of what happens when industries build cost structures around resources that seemed cheap and permanent.
An entire generation of American manufacturing was built on cheap energy. The OPEC embargo didn't just raise prices. It revealed how deeply the assumption of cheap oil had been baked into every operational model. Companies that hadn't planned for repricing didn't survive the transition.
Manufacturers built supply chains around cheap Chinese rare earth minerals. When geopolitical tension made that access fragile, companies discovered they'd optimized for a world that no longer existed. The dependency was so deep it couldn't be unwound quickly.
Russian natural gas went from an infrastructure asset to a geopolitical weapon almost overnight. Entire industries that had built decade-long cost structures around cheap pipeline gas found themselves repricing everything in real time.
GTM practitioners are building labor replacement workflows on AI compute that is priced below its actual cost. The subsidy is venture capital chasing market dominance. When the market matures and capital demands returns, the pricing resets. The pattern is always the same.
Abundance creates dependency. Dependency creates fragility. And fragility gets exposed the moment external conditions shift.
The Macro Risk Nobody's Pricing In
Here's where I'll acknowledge the overanalysis charge and make it anyway. The stock market is sitting at valuations that assume a smooth, frictionless AI-driven productivity boom. Markets are pricing in the upside of AI transformation without fully accounting for supply chain disruption, energy shock risk, and geopolitical instability that could strangle compute availability at the source.
Data centers run on energy. GPUs run on chips. Chips run on supply chains that cross oceans that are increasingly being tested by territorial disputes, the slow unraveling of free trade consensus, and instability around the shipping lanes that move the components that power the compute that runs your Base44 app.
If you think token costs exist in isolation from what's happening in the South China Sea, or the Strait of Hormuz, or the global shipping infrastructure that free trade depends on, you are not thinking about this at the right altitude. Continued conflict, energy shocks, supply chain disruption: any one of these accelerates the end of the subsidized token era. All three together could make it arrive violently. The market hasn't priced that in. Neither have most of the people who've built their GTM infrastructure on current token economics.
The Real Reckoning
Here's my uncomfortable thesis, and I'll own the fact that maybe I'm wrong. The GTM engineers and demand gen practitioners who have delighted in building their own tooling, who have felt the genuine joy of cutting expensive SaaS subscriptions and replacing them with AI-native workflows, who have built real competitive advantage on cheap token economics: those people are about to face a reckoning. Not because their instincts were wrong. Their instincts were exactly right. But because they may have built on a foundation that was never permanent.
The question isn't whether the subsidized era ends. It's whether what you built can survive the repricing. Sidecar had great technology and lost anyway. The best idea in a subsidized market doesn't always win. Sometimes the check just stops coming, and whoever is left holding infrastructure built on someone else's venture capital has to figure out what that infrastructure actually costs.
Be a little skittish. It's warranted.
