> people underrate how precarious a position OpenAI is in.
Most people assume OpenAI is both making money and the only American AI company of note. So...yeah, agreed. Though I think you'll put me in this bucket, too.
> It's much more like Dell vs Lenovo vs HP.
How so?
Migrating from a Dell to a Lenovo is the same process as migrating from a Dell to a Dell. Same platform, even he same manufacturers in many cases, just a different brand. There's little functional difference.
> The data gravity isn't actually that real.
Whereas their chatbots develop a relationship with the user. Even if you export your chats and could import them elsewhere, if you want *your friend* and not *some new friend*, you have to stay put.
> First mover advantage doesn't mean you automatically win.
ofc.
> AI isn't a free service
ChatGPT has always had a generous free plan, and ~98% of their half a billion MAUs of ChatGPT are on it. 4o is already way more than most people need, evidently.
> , and people want the best. (Especially when the models get really good, people will definitely want the best.)
> They will cross compare and shop around.
I don't believe see any evidence that this is at all true outside of nerdy populations and heavy professional users, and I wonder if you've seen evidence to the contrary (re genpop).
> Especially when you have to pay.
But you don't. I'm an enthusiast and I don't even pay with all the SOTA and near-SOTA models available fo free (like Gemini).
> There aren't actually any real network effects.
Disagree. "I was talking to Chat" is starting to become a thing broadly. "I was talking to <some other AI>" is not a thing I have ever heard someone outside the bubble say.
> It's not a social product.
Disagree:
1. It's engineered to be friend hyperstimulus --which it is very good at doing!
2. Talking about ChatGPT chats is a social activity.
3. Sama is building ChatGPT into a social network.
> There is an unbelievable amount of competition.
> Western oligopolies often tacitly pricefix to maintain mutual margins (like in cloud), but the Chinese will definitely undercut them.
B2C, you can't undercut "Free thing everyone you know uses".
B2B, Chinese providers are going to have a really hard time breaking in for many reasons, extremely over-broad trade sanctions among them.
> And autists like Zuck and Elon are also willing to defect to gain market share.
Everyone is offering ChatBots for free already, so...
> OpenAI is at a massive hardware disadvantage vs Google.
> Their AI is way more expensive to create and they have less money to start with.
True. These are real problems.
> OpenAI is at a massive infrastructure disadvantage vs Elon.
It can't be that massive considering they serve more users and have trained a bigger base model.
> He doesn't have the megapacks to smooth electricity. He doesn't have the skill. It's Elon, the strongest talent in the world at infra various the random nitwits at Oracle.
Agree Elon has advantages here. But Sam has found all the compute they've needed so far.
> OpenAI is at a massive cash disadvantage vs Meta.
True facts. That said, the man can pitch, and a lot of money out there is real dumb.
> Zuck is an autist who is repeatedly willing to burn his company to ground and crater his stock by 75% spending insane amounts of money on wild bets.
I wouldn't count out Zuck by any means, or Elon, but neither of these guys have models people want to use.
> Also, my personal opinion -- but I think Sam is a conniving but ultimately mediocre talent and intellect.
Engineering talent? Yes. But at gaining power, he's a savant.
> He has never succeeded at anything. Not at loopt or anything impressive at YC.
Taking over YC after Loopt while being flagged by your boss as extremely power seeking is itself an achievement.
> The reason openai succeeded was because of the incredible early concentration and monopoly of talent because it was a very luxurious and open environment compared to Google. A lot of the talent was recruited by Elon.
You could say that credibly about the first few years. At this point...Elon's been gone a long time, the turnover has been huge, and ChatGPT is still the Xerox of AI.
> I think Sam got lucky in huge part. Right place right time.
Agree to disagree, I guess. Not that there was no luck but also...he's been beating the turboautists you were lauding above who have far more money and bigger preexisting platforms.
> Do acquisitions like Jony Ive and Windsurf really inspire confidence?
He didn't acquire Windsurf, and yes, partnering with Jony Ive to make hardware is a sensible path to more moat.
> He's cavorting with idiots like Masa And *Silent Gen* boomers like Larry Ellison.
I don't know who Masa is, but Larry Ellison can build him datacenters. (Silent Gen and Boomer are mutually exclusive).
> He's bleeding talent like a sieve. Ilya. Schulman. All those Chinese guys and o1/o3 guys.
This is a real problem. And one open weight models could help with now that the doomers have left the building.
> Everyone good is looking for a big payday.
This is demonstrably not true --a bunch of people refused.
> Can anyone name a single researcher left at OpenAI besides Noam Brown? (yeah Keller Jordan too I guess)
Yes.
> I frankly think Altman's a lightweight compared to Wenfeng, Elon and even Zuck.
In what domain? Engineering? Definitely. AI specifically? Zuck and Elon have been deep in the weeds with Llama and Grok and it didn't obviously help. Wenfeng is amazing but compute strapped.
> Maybe closer to Dario and Google n Chinese labs will start getting a huge spigot of Huawei chips starting in Q3 2026 apparently, further brutalizing the economics.
OpenAI doesn't compete in the Chinese market, and (frankly absurd overreaching) trade sanctions limit the thread to OpenAI specifically (since they aren't in China anyway).
> The ability to compete in future rounds of this tournament will get exponentially more expensive. 5 billion in 2024. 100k H100s. 10 billion in 2025. 100k GB200 NVL72 40 billion in 2026. 400k GB300 NVL72 And so on.
This is an area where Chinese competition national competitiveness frame probably helps them, tbh.
> These 5 giants will start slowly falling one by one as people realize how commoditized AI really is.
APIs are a commodity, yeah. Businesses care about this but most won't use Chinese models. So, they can leave OpenAI for Anthropic, Google, or Mistral.
B2C...the models are a commodity, but they aren't the product. The product is the product and OpenAI has the 'best' (darkly) productization.
I could see OpenAI biting the dust eventually. But how would they *all* go down? 'Merica is not going to move over to Chinese models, Google Meta run very lucrative behavior modification businesses, and OpenAI can pivot to that whenever they want (to the extent they haven't already).
> Probably in this order.
> Anthropic (possible talent / Interpretability advantage but both might be total vaporwave/overrated)
I would love to see this. They have very wealthy ideological power-hungry network. But maybe!
> Meta. (money)
Meta has a money printing machine. But one scenario I can sort-of imagine is something like: Llama 4's successor crashes and burns, their stock price tanks, and Zuck scales back his ASI ambitions.
> Openai. (first mover advantage)
IMO they'll be able to keep raising as long as AI progress keeps up.
> xAI. (Money and infra)
They do have money and infra. No customers, though, and Elon doesn't seem to be in top form recently. Like...could this Elon have built SpaceX? IDK but I'm skeptical.
> Google. (Money and chips) (Many of the things I said could turn out wrong of course...)
I agree they're the most secure player. So much compute, talent, money, data, and userbase...it's really their game to lose (and long has been if you ignore the blip from ChatGPT through Gemini 2 or so).