All AI companies are valued as if they'll win the AI race. Problem is, only one can win *

* That is, if a lab like Deepseek doesn't release a "good enough" model for the cost of electricity + a thin margin. Then all AI company lose and all consumers win.

> only one can win

Why? I mean, if you assume that the moment some threshold of intelligence is reached it will suddenly explode and self-improve at a pace no one would be able to ever catch up with, then yes probably only one can win. But I think it’s more sci-fi than reality which is often more nuanced and boring.

Late stage capitalism, everyone's curve must exponential growth, yet the amount of customers on planet Earth is limited.

> good enough" model for the cost of electricity

This already exists, in several tiers, depending on what you can afford to run (in terms of VRAM).

Deepseek V Flash 0731 is already good enough for many things.

And that one can only win if it convinces leaders to regulate AI so that opensource models don't get good enough.

They can limit whatever they want within their own borders. Who cares. It’s only ~350m people. The rest of the world benefits while the US dismantles its own relevance.

What does it even mean to "win the AI race" ?

Right now there are many providers providing different tradeoffs, for different usages.

Would "winning" mean releasing an all-purpose better model that will outperform every other companies for all kind of tasks ? Something like AGI but cost effective ?

Or is it winning the consumer-side AI race ? Like making a kind of "super personal AI assistant" the likes you would see in movies and TVs ?

I'm not arguing, I am just struggling to understand what it usually means when somebody says that, it's not like competitors will go bankrupt the second OpenAI/Anthropic/Whatever does something specific.

I think "winning" means being able to live up to their valuations. The market is not big enough for all of these companies to be worth as much as they claim. In that sense they must "win" or experience massive downward adjustment in their values.

In that case I doubt there will ever be a "winner" because these valuations are not based in reality.

So “winning” could mean everyone experiences massive downward adjustment, but your valuation is still highest?

There's probably some holes in the analogy is you stretch it far enough. I'm hoping to "win" as a user by local models hitting a performance breakthrough and not needing any of these companies. I think the OP was generally making the point I mentioned though, that there's not room for all of these valuations, maybe one.

People believe that it will be a market with very few winners.

>What does it even mean to "win the AI race" ?

Creating a world-ruling AGI which feels mild sympathy for its creators, sharing some table scraps with them. Not sure if /s...

And winning the AI race means being the biggest loser. LLMs aren't profitable.

Another outcome is that none win.

Comparing revenue across industries/verticals is not sound.

Mercedes has a ~5% profit margin on this revenue.

SaaS, even AI SaaS is expected to have much higher profit margins at scale.

And on top of that, Mercedes shrunk by 10% revenue over the last year, rather than growing, which makes sense because they missed EVs and are now managing decline and returning money to shareholders.

> Cursor is allegedly worth $60B

Inflated valuation of buyout, based on inflated valuation of SpaceX

Cursor and Mercedes Benz is a weird comparison. One company is working on the future technology while the other one is stuck in the last centuries technology which is being phased out.