OpenAI trades tokens for startup equity

**SPEAKER_1** (0:00)
No one goes to Hanks for his spreadsheets. They go for a darn good pizza. Lately, though, the shop's been quiet. So Hanks decides to bring back the $1 slice. He asks Copilot in Microsoft Excel to look at his sales and costs and help him see if he can afford it. Copilot shows Hanks where the money's going and which little extras make the dollar slice work. Now Hanks has a line out the door. Hanks makes the pizza, Copilot handles the spreadsheets. Learn more at m365copilot.com/work.

**SPEAKER_2** (0:31)
Your summer starts now with Memorial Day deals at The Home Depot. It's time to fire up summer cookouts with the Nex Grille 4-burner gas grill on special buy for only $199.
And entertain all season with the Hampton Bay West Grove 7-piece outdoor dining set for only $499. This Memorial Day get low prices guaranteed at The Home Depot. While supplies last, pricing valid May 14th through May 27th. This only includes supplies, see homedepot.com/pricematch for details.

**SPEAKER_3** (1:00)
This episode is brought to you by Nordstrom. Whether you're packing for a vacation or dressing for a wedding, Nordstrom has head to toe looks for wherever summer takes you. From swimwear and bold color to romantic florals and dresses that make an entrance. Find the right style for every unforgettable moment. Shop new arrivals from brands you love like Farm Rio, Free People, Vince, Waithe and Stodd. Plus free shipping and returns and styling help whenever you need it make it all so easy. Shop now in stores at nordstrom.com or download the Nordstrom app.

**SPEAKER_4** (1:30)
OpenAI is giving $2 million worth of artificial intelligence tokens to all 169 startups in the current Y Combinator batch in exchange for equity.

**SPEAKER_5** (1:43)
Right. Which, I mean, they aren't handing out cash. They're trading raw compute power for actual ownership stakes in these young companies.

**SPEAKER_4** (1:50)
Yeah. And that brings up this fundamental problem for anyone trying to build a business right now. You know, when a startup accepts millions in infrastructure instead of traditional funding, are they securing their future or are they just like signing away their independence to a single platform?

**SPEAKER_5** (2:05)
Yeah. That is exactly the tension founders are dealing with. Sam Altman actually calls this specific strategy Token Maxing.

**SPEAKER_4** (2:11)
Token Maxing.

**SPEAKER_5** (2:12)
Right. Yeah. It's an approach where startups and their engineers just, you know, maximize their use of AI models and credits to accelerate product development and their internal workflows.

**SPEAKER_4** (2:22)
You just saturate your entire operation with AI because the cost is temporarily zero.

**SPEAKER_5** (2:26)
Exactly. But to understand what Token Maxing actually does to a company's architecture, we kind of have to look at what a token functionally is.

**SPEAKER_4** (2:34)
Right. It's the basic unit of text that these large language models process and generate. Like if you're building an app in English, one token is roughly four characters.

**SPEAKER_5** (2:44)
Which works out to about three quarters of a word.

**SPEAKER_4** (2:46)
Yeah. So a prompt of say, 75 words is going to cost you about 100 tokens.

**SPEAKER_5** (2:52)
But that linguistic constraint really alters the math, because that four character estimate only applies to English.

**SPEAKER_4** (3:00)
Right. Because of the training data.

**SPEAKER_5** (3:02)
Exactly.
Tokenization is inherently linguistically biased based on how these models were trained. So non-English text produces a much higher token to character ratio. Well. Yeah. If you're a founder building a product for the Japanese or Arabic markets, the same amount of semantic meaning requires significantly more tokens.

**SPEAKER_4** (3:21)
So you basically pay a tax for not operating in English.

**SPEAKER_5** (3:25)
You really do. Two prompts of identical length and meaning are going to generate entirely different bills depending on the language.

**SPEAKER_4** (3:31)
For a startup trying to build multilingual products, that discrepancy completely alters their burn rate.

**SPEAKER_5** (3:37)
Right. It's a fundamental planning assumption that changes how far that $2 million actually goes.

**SPEAKER_4** (3:41)
Which is wild because the face value of the offer seems staggering at first glance. $2 million per startup across a cohort of 169 companies?

**SPEAKER_5** (3:52)
Yeah, that totals over $330 million based on retail pricing.

**SPEAKER_4** (3:56)
But OpenAI is definitely not spending $330 million to make this happen.

**SPEAKER_5** (4:00)
No, not at all. Those credits are priced at the retail rate they charge developers for API access. OpenAI's actual marginal cost to run those API calls in their data centers, it's just a fraction of that retail price.

**SPEAKER_4** (4:13)
Right, it's kind of like an airline buying equity in a tech startup using frequent flyer miles.

**SPEAKER_5** (4:19)
That's a perfect analogy. The face value. You said this place was steps from the water.