The Pricing Model
Is the Confession.
Palantir's CEO went on national TV and asked the question Silicon Valley won't answer: "If it was so valuable — say I can make you $1 billion tomorrow — wouldn't I say, I'll make you $1 billion and I want 30%? Why are they charging for tokens?" The clip went viral. Half of it is exactly right. Half of it, I'll push back on — using my own business as the evidence. The line between the two halves is the most useful due-diligence test I know.
What he actually said
On CNBC, Alex Karp said enterprises are "livid," "paying for tokens that create no value," while AI companies are "stealing their weights and alpha." And then the pricing question above. Two honest footnotes before we use it: Karp competes with the companies he was attacking, and Palantir's own stock rose 9% after the interview. Also — Palantir doesn't take 30% of its customers' profits either. It charges contracts. Remember that; it matters in a moment.
The half that's wrong — inputs are priced per unit, honestly
Nobody asks why the electricity company charges per kilowatt-hour instead of taking 30% of the factory's profits. Electricity is an input. Its value depends entirely on what the user builds with it — the same kilowatt-hour runs a bitcoin miner losing money or a chip fab printing it. Charging per unit isn't a confession of worthlessness; it's the only honest way to price something whose value the seller doesn't control.
Tokens are the same. I know because I buy them — deliberately. I've written before about why we moved from AI subscriptions to API: pay for what you use, at cost you control. The same token that writes a worthless email for one company runs our entire content, commerce and assistant infrastructure at IMA AI. The token didn't decide that. We did. When the buyer owns the outcome, per-unit pricing is exactly right — and outcome-share pricing would be absurd. Would you give your electricity company 30% of your business?
The half that's devastating — for a different target
But redirect Karp's question away from inputs and toward promises, and it kills.
"This course will make you rich — RM1,997." "Our agency guarantees your GMV." "This tool will 10x your sales — RM99 a month." Every one of these prices an outcome claim as a flat product. And here Karp's logic is airtight: if the seller truly controlled that outcome, flat pricing is irrational. They'd take equity. They'd take revenue share. They'd use it themselves. I made exactly this argument about AI courses — if the method prints money, the operator runs the method; only the guru sells it.
My own books, opened
I'll hold IMA AI to the same test, because people assume we're an AI services company. We're not. I don't sell AI services. I invest them.
We don't run an open-door agency. There's no public price list and no sales team chasing leads. I choose who becomes a customer — and for anyone outside that choice, the price is premium, no discount, take it or leave it. That's not arrogance; it's the pricing test applied to myself. My capacity belongs to my own outcomes first, so a fee has to outbid my own upside. Conviction prices like that: a stake, or a premium that compensates for not taking one.
Our actual model is Karp's answer put into practice. We partner with an existing player in an industry — their domain knowledge, their licences, their relationships — we bring the AI layer, and together we go compete in that industry to disrupt it. We don't invoice the partner for the AI. We share the outcome we're claiming the AI can produce. When we were good at livestream commerce, we invested in our own operations instead of selling the skill. Now that we're good at AI, we invest it into industries the same way.
If a seller truly believes their AI makes businesses rich, this is what their P&L looks like. Watch for it.
The buyer's checklist
Ask what's being sold. Input, craft, or promise? Only the third should scare you.
Ask why they're selling instead of using. For inputs and craft there are honest answers — the value depends on you, or the craft is the business. For an outcome-promise there is no honest answer at a flat price.
Watch where their own money goes. The seller's capital allocation is the truth serum. A guru whose income is course sales, an agency that guarantees GMV but takes no equity, a "passive income" system sold for a fee — the pricing model has already confessed.
And flip it on yourself. If you're certain of an outcome in your own business — invest, don't sell it.
The bottom line
Karp's question deserves the attention it got — just aimed correctly. Token pricing isn't the scandal; it's utility pricing for a utility, and the smartest thing a business can do is buy the input cheap and own the outcome itself. The scandal is everyone who prices your success as their flat-fee product. When it's too good to be true, the tell was never the promise. It was the price tag.
Published by IMA AI — July 2026. We don't sell AI services publicly — we invest them into industries we choose, with partners we choose. If you think your industry is one of them, you know where the door is.