Video · 2026-07-20 · 1h 29m · 6 moments

The Open-Source AI Reality | How Token Costs Will Fall 10X & Usage Will Explode 100X | Lin Qiao

✦ AI generated

timeline · colored by role

Claim◆

The future of frontier intelligence is specialized, private models built on proprietary enterprise data, not a single generalized AGI model, because most of the world's data is private and never used to train general models.

Lin Qiao argues that because the vast majority of the world's data is private and locked inside enterprises, the real frontier of intelligence is specialized, private models built on that data rather than one generalized AGI.

transcript

Lin Qiao: If you think intelligence is a derivative of data, then majority of the data is actually not used for training a general intelligence model. The training data is coming from public internet and the label data. Public internet is very small corpus of data compared with words data. Majority of words are actually private locked inside application locked inside enterprise.

extends · 1provides context · 1rebuts · 1

Claim

In the AI era, having product-market fit no longer guarantees a durable, scalable business, because companies can scale their AI costs into bankruptcy.

Unlike the SaaS era where product-market fit essentially guaranteed a durable business, Lin Qiao says many AI startups now have real customer demand but can't afford to scale because inference costs would bankrupt them.

transcript

Lin Qiao: for startups you know we have great companies they have product market fit customer want to pay them and they really value their product but they cannot scale because once they scale they could scale into bankruptcy. Have you heard about scaling to bankruptcy? So that's a real problem.

provides context · 1

Prediction

The future will not be dominated by a small number of general AGI models; instead there will be millions of specialized models, one per application or use case.

Lin Qiao predicts that rather than a handful of dominant AGI models, the future belongs to millions of specialized models, each tuned for a specific application or business.

transcript

Lin Qiao: I really believe the future will be will not be a few small number of AGI models dominant world. I really believe the future will be it may be scary but I think that's true. It will be millions of specialized model one per application per use case.

explains mechanism · 1extends · 1provides context · 1

Claim

No single company should own or control intelligence, because different regions, companies, and people have different values and need specialized flavors of intelligence rather than one standardized model dictated by one company.

Lin Qiao says it would be dangerous and nonsensical for one company to monopolize intelligence, since general common intelligence must coexist with many specialized, independently-owned intelligences reflecting different values and use cases.

transcript

Lin Qiao: what I don't want to see is there's only one company owns intelligence. I think that doesn't make sense to me because there are different flavors of intelligence. There's this general common intelligence that benefits everyone. Um and then there's a specialized intelligence that actually help us advance in in history.

Claim

Fireworks deliberately avoids building its own chips or owning the full stack, choosing instead to focus narrowly on where it adds the most value and to lean on partners' strengths elsewhere.

Lin Qiao says Fireworks' guiding philosophy is agility and focus — it deliberately avoids vertically integrating into chips, preferring to earn the right to expand the stack only once a workload is proven and massive.

transcript

Lin Qiao: It really depends on the company philosophy. To us, agility is everything and we need to earn the rights of building anything. So focus is everything for us and we want to focus on where we add the biggest amount of value based on our strength and uh we would like to leverage other people's strength to build on top of.

Prediction◆

Token costs will fall roughly 10x over the next three years, and that price collapse will drive roughly 100x growth in AI usage.

Lin Qiao predicts a 10x drop in inference token costs over three years, driven by supply-chain and competitive pressure, which she says will in turn unlock a 100x explosion in AI usage.

transcript

Lin Qiao: in the long term two to three years it should change and that cost will compress uh so overall I can imagine no 10x X uh cost reduction in the next three years and this 10x cost reduction will drive a 100x usage.

explains mechanism · 1extends · 1provides context · 1rebuts · 1supports · 1

Highlight slides

Intelligence Is a Derivative of Data✦ from: The future of frontier intelligence is specialized, private models built on proprietary enterprise data, not a single generalized AGI model, because most of the world's data is private and never used to train general models.The Real Frontier: Specialized Private Models✦ from: The future of frontier intelligence is specialized, private models built on proprietary enterprise data, not a single generalized AGI model, because most of the world's data is private and never used to train general models.Lin Qiao: The Coming Price Collapse✦ from: Token costs will fall roughly 10x over the next three years, and that price collapse will drive roughly 100x growth in AI usage.Cheaper Tokens, Explosive Usage✦ from: Token costs will fall roughly 10x over the next three years, and that price collapse will drive roughly 100x growth in AI usage.

Related episodes

Video · The Cognitive Revolution

Fable's Back, AI Engineer Recap, & SambaNova

→

Audio · All-In Podcast

SpaceX-Cursor Deal, SaaS Debt Bomb, New Apple CEO, SPLC Indictment, Colon Cancer Spike

→

Video · The TWIML AI Podcast

AI Factories Explained: What’s Actually New?

→