OpenAI isn’t stopping at models. It’s building the interface of the future. And that puts Sam Altman on a direct collision course with xAI, Google, and every legacy platform still playing by old rules
The scale is impressive, but the economics still don’t line up with the hype. $13B projected revenue sounds huge, but burning $2.5B in cash and spending $6.7B on R&D in six months shows the model is still massively capital-intensive. That’s not SaaS economics — it’s infrastructure economics.
The “everything app” vision might play out, but in the long run stock prices track cash flows. Until inference costs collapse or OpenAI proves durable margins, it’s not Amazon 2.0 — it’s a company subsidized by investors to chase dominance.
For every Amazon or Palantir, there’s a Webvan, Pets.com, or Nikola — companies that scaled hype faster than profits and burned out. The difference is whether growth translates to durable margins before the cash runs out. That’s the hurdle OpenAI hasn’t cleared yet.
Right now, investors don’t care as much about OpenAI’s profitability — it’s the velocity that matters.
$13B a year is a massive number, especially considering how early we are. Most users still haven’t adjusted to the idea of paying for AI services. What we’re seeing is revenue that would’ve gone to Apple, Adobe, Duolingo and other traditional software names.
Once AI agents start booking flights and writing emails on your behalf, the price point will go up.
Velocity is impressive, but history is full of “growth first, profits later” stories that never cleared the profitability hurdle. Revenue shifting from Apple or Adobe only matters if AI margins eventually look like Apple’s or Adobe’s. Until then, it’s a race where cash burn is the business model.
If you follow this logic, you should be buying companies like Google, Meta, and maybe even Salesforce (CRM). I don’t know your risk profile — maybe an ETF would be a better fit.
The point of investing is to spot promising companies before they go public and understand how they shape the broader market. Comparing OpenAI to Apple doesn’t make much sense — especially considering that Apple has been sitting on its cash pile for years, spending it mostly on buybacks and incremental camera upgrades.
Cash burn is completely normal. Some companies reach profitability, some don’t — but that’s not a valid reason to criticize a firm that forced Google to rethink its entire search strategy.
I wasn’t asking for portfolio advice — I was pointing out the economic reality behind OpenAI’s model. There’s a difference between debating ideas and dismissing people.
Cash burn is common, yes. But history shows that which companies survive the burn comes down to whether operating leverage emerges. Google and Meta had software economics; OpenAI is still dealing with infrastructure-level costs. That’s not a small distinction, and it’s exactly what long-term investors should be weighing.
Yes, software margins built Google and Meta. But those companies scaled atop already-mature infrastructure. OpenAI is building the next layer — models, agents, inference, and deployment rails — which is far closer to foundational R&D than to app-layer monetization.
The fact that it’s still burning billions while growing revenue faster than Anthropic, Cohere, or even Palantir is part of the story — not a red flag in itself. It’s a bet that foundational models will become generalized compute layers, and the companies that dominate them (and their orchestration) will extract value from the entire AI stack.
So yes — infra is expensive. But if OpenAI ends up being the “AI operating system,” then those costs become the tollbooth.
That’s a fair framing — the bet is whether “AI as operating system” ends up behaving like a tollbooth or like a utility that gets competed down to low-margin economics. If it’s the latter, then the infrastructure builder takes the risk without ever enjoying software-like returns.
My point isn’t that OpenAI can’t become the tollbooth — it’s that until inference costs structurally collapse or the moat looks durable, the economics still resemble a utility build-out more than a software flywheel. And in markets, that distinction matters.
The scale is impressive, but the economics still don’t line up with the hype. $13B projected revenue sounds huge, but burning $2.5B in cash and spending $6.7B on R&D in six months shows the model is still massively capital-intensive. That’s not SaaS economics — it’s infrastructure economics.
The “everything app” vision might play out, but in the long run stock prices track cash flows. Until inference costs collapse or OpenAI proves durable margins, it’s not Amazon 2.0 — it’s a company subsidized by investors to chase dominance.
Amazon also burned cash, and Palantir and NBIS and CRWV. If you wait until it’s profitable you miss the growth period
For every Amazon or Palantir, there’s a Webvan, Pets.com, or Nikola — companies that scaled hype faster than profits and burned out. The difference is whether growth translates to durable margins before the cash runs out. That’s the hurdle OpenAI hasn’t cleared yet.
Right now, investors don’t care as much about OpenAI’s profitability — it’s the velocity that matters.
$13B a year is a massive number, especially considering how early we are. Most users still haven’t adjusted to the idea of paying for AI services. What we’re seeing is revenue that would’ve gone to Apple, Adobe, Duolingo and other traditional software names.
Once AI agents start booking flights and writing emails on your behalf, the price point will go up.
This will be an expensive friend.
Velocity is impressive, but history is full of “growth first, profits later” stories that never cleared the profitability hurdle. Revenue shifting from Apple or Adobe only matters if AI margins eventually look like Apple’s or Adobe’s. Until then, it’s a race where cash burn is the business model.
If you follow this logic, you should be buying companies like Google, Meta, and maybe even Salesforce (CRM). I don’t know your risk profile — maybe an ETF would be a better fit.
The point of investing is to spot promising companies before they go public and understand how they shape the broader market. Comparing OpenAI to Apple doesn’t make much sense — especially considering that Apple has been sitting on its cash pile for years, spending it mostly on buybacks and incremental camera upgrades.
Cash burn is completely normal. Some companies reach profitability, some don’t — but that’s not a valid reason to criticize a firm that forced Google to rethink its entire search strategy.
I wasn’t asking for portfolio advice — I was pointing out the economic reality behind OpenAI’s model. There’s a difference between debating ideas and dismissing people.
Cash burn is common, yes. But history shows that which companies survive the burn comes down to whether operating leverage emerges. Google and Meta had software economics; OpenAI is still dealing with infrastructure-level costs. That’s not a small distinction, and it’s exactly what long-term investors should be weighing.
Yes, software margins built Google and Meta. But those companies scaled atop already-mature infrastructure. OpenAI is building the next layer — models, agents, inference, and deployment rails — which is far closer to foundational R&D than to app-layer monetization.
The fact that it’s still burning billions while growing revenue faster than Anthropic, Cohere, or even Palantir is part of the story — not a red flag in itself. It’s a bet that foundational models will become generalized compute layers, and the companies that dominate them (and their orchestration) will extract value from the entire AI stack.
So yes — infra is expensive. But if OpenAI ends up being the “AI operating system,” then those costs become the tollbooth.
That’s a fair framing — the bet is whether “AI as operating system” ends up behaving like a tollbooth or like a utility that gets competed down to low-margin economics. If it’s the latter, then the infrastructure builder takes the risk without ever enjoying software-like returns.
My point isn’t that OpenAI can’t become the tollbooth — it’s that until inference costs structurally collapse or the moat looks durable, the economics still resemble a utility build-out more than a software flywheel. And in markets, that distinction matters.
oh no the simpson predicted this, now my favorite substacker EoP!!!