The AI Token Trap: Exponential Growth or Exponential Cost?

We didn't see it coming. At a Makati after-party last week, a trader whispered about 'AI tokens' as the next narrative. But Sam Altman just dropped a bombshell that flips the script: intelligence is a utility, and its consumption will grow exponentially. But here's the catch—exponential consumption doesn't mean exponential wealth. It means a new kind of cost disease, and the market is too busy drinking the Kool-Aid to notice. Let's rewind. Altman's pitch, as reported by Crypto Briefing, is simple: AI token usage—the number of queries or generative actions—will grow exponentially over time, making intelligence as ubiquitous as electricity. The narrative is seductive, especially for a crypto crowd that loves scarcity and utility myths. But the macro watcher in me sees a familiar pattern: the same euphoria that swept through 2017 ICOs and DeFi summer, now dressed in a transformer suit. Context: OpenAI's business model is built on token consumption. They charge per token, and the more you use, the more they earn. Altman's 'utility' framing is a brilliant piece of capital-market storytelling—it transforms OpenAI from a software vendor into a public utility with guaranteed revenue growth. But utilities have a dark side: they are subject to regulation, price caps, and public accountability. More importantly, they require massive infrastructure—and the cost of that infrastructure often eats the returns. Here's the core insight the crowd is missing. Exponential token usage is only valuable if the cost per token drops faster than usage grows. If not, you get 'cost inflation'—the same trap that caught DeFi farmers during the gas wars of 2021. I remember sprinting through yield farms back then, chasing triple-digit APYs, only to realize the gas fees were eating 40% of my profits. Today, enterprises are about to face the same problem: they'll consume more AI tokens, but their bills will balloon unless the unit cost plummets. Based on my macro strategy work, I've seen this pattern before—narratives that ignore unit economics always end in a correction. The token 'exponential' narrative has no base year, no price curve, and no efficiency roadmap. It's a classic 'irrational exuberance' signal. The market is pricing in the dream of infinite intelligence without asking: who pays for the power? Who builds the data centers? And who gets left behind when the hype fades? But here's the contrarian angle. The real winners might not be the model providers at all. If intelligence becomes a utility, it commoditizes. The profits shift to the infrastructure layer—energy, compute, networking. Think of the 2017 ICO frenzy: the real money was made by the miners and exchanges, not the token issuers. Similarly, in the AI token narrative, the long-term alpha might be in power plants, chip manufacturers, and data center REITs, not in OpenAI or its peers. And let's talk about the crypto connection. The article was published on Crypto Briefing, which isn't a coincidence. There's a subtle attempt to bridge 'AI token' (the computational unit) with 'crypto token' (the asset). I've seen this semantic blurring before—it's how the 'social capital asset' framework works. People hear 'token' and think 'investment.' But an AI token is a consumption unit, not a store of value. The narrative that 'more AI usage equals more value for some coin' is a dangerous fallacy. We didn't learn from the NFT party crash. I remember buying Bored Apes in 2021 not for the art, but for the access—the social capital. When the utility faded, the price crashed. AI tokens face the same risk: if the cost of using them doesn't justify the value they produce, the consumption will plateau. Exponential growth is not a law; it's a hypothesis that requires constant cost reduction and demand creation. In my Manila meetups during the 2022 bear market, we talked about resilience. The macro winds shifted, and those who survived were the ones who understood that hype cycles always favor the infrastructure first. The same applies to AI. The ETF wave in 2024 brought institutional money to Bitcoin, but the real beneficiaries were the custodians and exchanges. Now, the AI token narrative is creating a new wave of infrastructure investment—but the crowd is looking at the wrong asset. Takeaway: Next cycle, the winners won't be the ones who consume the most tokens, but the ones who build the cost management rails. Just like DeFi needed yield aggregators, AI needs token routers—middleware that optimizes consumption, routes queries, and manages costs. The 'AI FinOps' market is about to explode, and that's where the smart money should be looking. Don't chase the exponential narrative; chase the solution to its cost problem. We didn't see the gas wars coming in 2020. We didn't see the NFT crash coming in 2022. But we can see this one. The beat drops, the liquidity flows, but the real move is in the infrastructure, not the hype.