reticulates

I think that is correct, and exactly why these purported margins are nonsensical. If Anthropic's reported $50 billion in revenue is majority per-token billing, and tokens have a margin of 80%, that would put Anthropic's profit at $30 billion on per-token usage. Where is that $30 billion going?

And conversely, let's look at the amount Anthropic are spending on compute. Anthropic has just started paying SpaceX $1.25 billion per month for compute. At an 80% profit margin that would mean Anthropic is going to be bringing in $6.25 billion per month... that's more than their current reported revenue.

And that's just one contract for compute. We know that Anthropic also pay Google ~$3 billion per month for compute (based on their committed spend of $200bn over 5 years) which is $36 billion per year. At $36 billion per year on compute with 80% margins that would put revenue at... $180 billion.

Add in their spend with Amazon and Microsoft, Anthropic are spending at least $4 billion per month on compute, or $48 billion per year, all but equal to their revenue. If margins on tokens are 80% and an estimated $37.5 billion of revenue is per-token revenue, that needs just $7.5 billion of compute per year, less than $1 billion per month.

The numbers just don't add up. If margins are 80% and they have $48 billion per year in compute spend, revenue should be over $200 billion.

If the 80% margin made any sense whatsoever, Anthropic would be printing money, yet they're losing money, and have only been profitable for one month based on some financial engineering (pre-commitments billed after the fact to reduce their costs during one month).

My guess is margins are closer to 20% than 80%. That's the only way any of the numbers can make sense.

You’re missing the point. Anthropic are bringing more compute online by renting it and then they are creating the demand for it by increasing the limits on fixed cost plans. Anthropic are increasing their spend on compute without increasing revenues. I’m not guessing that, it is part of their announcements!

https://www.anthropic.com/news/higher-limits-spacex

“First, we’re doubling Claude Code’s five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans.

Second, we’re removing the peak hours limit reduction on Claude Code for Pro and Max accounts.“

Every one of these plans is a fixed cost. Anthropic doubled their limits without changing the price. Even if inference is wildly profitable and these plans aren’t subsidized, they’ve just cut the profitability in half.

And the non-plan usage that is being paid for directly is paid for monthly. If they can sell $1 of compute as $10 of inference then they have $9 the next month to spend on more compute. Of course the capital markets would want to give them money if that were true but they would have no reason to take it.

Why would a net 30 business need to borrow hundreds of billions over many years? Anthropic currently spends an estimated $5bn/month on compute so at most they need to float $5bn, but if they’re making 90% margins on compute, that $5bn would be paid for… within a couple of days. Where is the hundreds of billions of dollars?

A good manager doesn’t ask what you want to get better at, they ask what they can do for you. A manager who asks these sort of dumb questions is just going through the motions, you should do the same, say whatever they want to hear and then don’t think about it again. You can, alternatively, try and make them into a better manager by changing the question but they might not be receptive.

The entire economy could be entirely powered by AI and use less compute than is being used today. You're projecting the amount of compute used by coding while subsidized onto other industries, but that doesn't translate.

Speak to some of these legal technology companies and ask them 2 questions:

1. How much is your AI spend on coding? 2. How much is your AI spend on AI within your product?

The answer to #1 will dwarf #2 by orders of magnitude. And that's now, when these companies are still finding their feet, using the most expensive frontier models for their product that are likely overkill (as the product matures, they'll find the right mix of cost vs. capability, whether that's lower cost models from frontier labs, or open weight models).

Put simply, compute does not scale with economic value. A task you bill $500/hour for could be done in 2 seconds by an LLM vs. a task a developer bills $500/day for could take an LLM 10 minutes. Same dollar value, huge disparity in compute.

The only use-cases for AI that are comparable to coding on compute are image and video generation. Unless we end up with an economy primarily made up of companies producing code, image and video, there's literally no way compute needs can keep growing without subsidies.

I think it is hard to overstate just how much "work" is being done by Claude Code and Codex because it is "free" at the point of use. There are millions of newly minted developers prompting Claude and Codex to generate trillions of lines of code every single day because it has no marginal cost, not because it is driving any economic value. And as soon as they're exposed to the real cost, when economic value becomes a factor, they're going to stop doing it (as we're already seeing with companies like Uber).

Through your YC connection and legal work, you have access to a lot of very successful people working in every industry: ask those 2 questions, how much are they spending on coding with AI vs. how much are they spending on AI in their product? You're going to find that even the most aggressively AI-integrated products are spending pennies on their product's AI usage compared to the dollars they spend on their AI coding.

We're at peak compute demand, and it is almost entirely driven by coding, which is subsidized. The real economy isn't like the creative and chaotic make things and see what sticks world of coding. The real economy is boring, routine, regimented, task oriented, you hire people, train them, they do the tasks, you make some money. Most of the economy could be replaced with a few semi-intelligent macros.

Yes, AI is coming to every industry, you're right, but it isn't going to explode compute demand, it is going to make these industries more efficient, it is going to reduce costs, not shift them to compute, because $1 of compute can do more than $1,000 of a human in most industries.

We can come back to this comment in a couple of years. I bet we'll be using less compute then than we are today.

but your contention is they are profitable on inference. Why would they need to raise money to get their hands on compute if they're making money on compute? There's no upfront costs for Anthropic, Anthropic don't own or build compute infrastructure, they just rent access to compute owned by someone else, such as the $15bn/year SpaceX deal they signed recently (which they used to create more demand without increasing revenue).

https://www.anthropic.com/news/higher-limits-spacex

How does future demand translate to spend?

Anthropic aren't building out the data centres themselves, they're renting/leasing/borrowing from companies that are doing the actual spend on building out infrastructure. And the data centre companies aren't spending their own money, they're borrowing too (hence Apollo investing in data centres). Anthropic are paying SpaceX ~$1.25bn/month right now for access to more compute, that's $15bn a year, more than what these supposed margins would require in total spend (based on current revenue estimates).

https://www.anthropic.com/news/higher-limits-spacex

The SpaceX deal is a great example of Anthropic creating demand, i.e:

We’ve agreed to a partnership with SpaceX that will substantially increase our compute capacity. This, along with our other recent compute deals, means that we’ve been able to increase our usage limits for Claude Code and the Claude API.

They committed to spending $15bn per year with SpaceX and then increased limits for customers on fixed cost plans, creating more demand without any increase in revenue.

So, sure, it's not necessarily that they are raising money because they are unprofitable, but no alternate explanation makes any sense. The argument that could maybe made in favor is based on announcements like this one:

https://www.anthropic.com/news/anthropic-invests-50-billion-...

Today, we are announcing a $50 billion investment in American computing infrastructure, building data centers with Fluidstack in Texas and New York, with more sites to come. These facilities are custom built for Anthropic with a focus on maximizing efficiency for our workloads, enabling continued research and development at the frontier.

You might conclude from that, Anthropic are financing Fluidstack's build out, but they're not.

https://x.com/fluidstack/status/2079250004510728559

Just after that announcement, Fluidstack raised $830 million to build out data centres, none of the money coming from Anthropic. Fluidstack are currently rumored to be raising another $1bn. Anthropic's "$50 billion investment in American computing infrastructure" is just committed spend on renting compute from Fluidstack, a commitment that Fluidstack then use to raise money to actually deliver it. If Anthropic making money hand over fist, they wouldn't need to raise for committed spend.

And thus we return to the original question, how does future demand translate to spend? Actual handing over of dollars?

Your contention is they're spending it on what, exactly? Leaks put OpenAI's training spend at single-digit billions so that can't be the machines they're building, and they (OpenAI + Anthropic) are famously renting/leasing/borrowing compute through varying-degrees-of-circular deals... so what's the machines they're building?

Looks like it’s behind a paywall. I’ll take their word for it that semi analysis now estimate it to be 80%. That makes my point even stronger, if that number is true, where is the money? The report says that Anthropic generate over $50bn in revenue so at 80% margins that gives $40bn in profit. Where is that money? If they’re generating $40bn in profit, even after accounting for very high employee compensation and training costs… they should have tens of billions in profit, yet they’re out raising tens of billions instead. Where is the money going? And if only 20% is their actual inference costs, where are all these compute providers going to make their money? The world is at compute capacity on, what, $10bn in revenue?

We don’t “know” because they haven’t released any numbers but the most optimistic estimates (which many people believe are very very optimistic) put it at 60%: https://newsletter.semianalysis.com/p/anthropic-growth-and-b...

The simple question to ask is, if it is so profitable, where is all the money going? If Anthropic have 90% margins on API usage and API usage is $50bn+ in revenue per year, where is the $45bn going? Why do they need to raise so much cash, constantly?

Where are you getting that 80% figure from? Even semi analysis, the most aggressively optimistic analysts, put it at around 60%.

https://newsletter.semianalysis.com/p/anthropic-growth-and-b...

You’re talking across the issue. The demand is real because it is cheap. The demand is being generated by OpenAI and Anthropic selling inference below cost on fixed price plans. If everyone was paying the actual costs then demand would fall through the floor. The legal tools you’re looking at use barely any compute. They’re not driving the compute demand. You can validate this by asking how much they are spending on API usage. A company spending $100,000 per month on a frontier model via an API is the equivalent of… 10 or so OpenAI and Anthropic fixed price plan customers. Are these legal tools spending hundreds of millions per year on the frontier models?

no, even if we assume their margins are 90% (they are not) they are still losing money because the $200 plans allow for tens of thousands of dollars worth of inference and a huge number of users are milking every cent across multiple accounts. Every “reset” OpenAI and Anthropic do is setting money on fire.

If it were true that they’re making money hand over fist they wouldn’t need to raise tens of billions of dollars every few months.

https://xcancel.com/i/article/2076078865060151465

Yes but how much of that compute shortage is from demand that is subsidized? We’ve seen companies like Uber drastically cut how much they are willing to spend on AI because they are paying actual usage costs, while at the same time OpenAI and Anthropic increase the limits on their fixed cost plans for individuals meaning people not paying usage costs are using it more and more… doesn’t this show that the compute shortage is because OpenAI and Anthropic are paying for it, not their customers? And the moment OpenAI and Anthropic stop paying for it, demand will collapse.

The $200 you’re paying is heavily subsidized, you need to consider the real inference cost which could be closer to $10k if you’re using the maximum permitted by your plan.

You’re back to the grandiose claims. The most important set of solutions in AI? Just as intercoin was the most important solution for money and qbix was the most important solution for community. You’ve been doing the same shtick for 20 years. Yawn. A month / year / decade from now when you’re still doing the same grandiose shit with zero traction, refer back to this thread. You are clearly intelligent but wasting all your potential on churning out grandiose-but-garbage ideas that you would know are the wrong thing if you spent more time listening and less time broadcasting. You’ve been shouting into the void for so long, maybe it is time to reflect on why that is. But then I guess if you’re not nauseated and ashamed by your ChatGPT conversation you may be too far gone https://chatgpt.com/share/691a9edc-b5b8-800a-99a3-c32d9abccf...

I’ll give you a million dollars if safebot is ever mentioned by someone credible. You don’t seem to realize that you are 1 of millions of people doing exactly the same thing, I have read so many of these revolutionary safety ideas, the only thing worse are all the god awful “memory” solutions people keep “inventing”.

Advice and information subreddits have gone to shit because of AI usage. A large number of people seem to think that when someone asks a question, what they really want is not someone with direct knowledge, but instead someone to relay the question to ChatGPT and post the result as if it is their own hard earned knowledge and insight. I have no idea about the quality of this research but in the real world (well, real-ish, as real as Reddit can be) it is stark, people aren’t just refusing to say “I don’t know” they’re actively seeking out opportunities to pretend they know things.

And yet your next comment after this was, once again, self promotion, proving my point.

Safebots isn’t a contribution. You have just reinvented tools but bundled it with thousands of words of AI generated slop and terrible AI generated “ideas”. Everyone with Grok invents a new “computing substrate”. Yes, “substrate” is such a Grokism that I can immediately identify you use Grok for “thinking”.

If you spent just a little more time reading and a little less time self-promoting you would know that you are 12 months behind everyone else, human in the loop approval flows for tool execution are standard. The examples of LLMs doing destructive things are because users provided blanket approval.

https://ai-sdk.dev/docs/ai-sdk-core/tools-and-tool-calling#t...

“I have contributed quite a bit — not i n the form of snark, but in the form of actual SOLUTIONS that I built and put on github as open source.”

Code that nobody has ever used. HN is a discussion forum, you haven’t contributed to discussion. You have always been a slop merchant, quite impressive that you were slopping before LLMs, maybe you are an internet innovator after all.

The ego of America to think it is the only place where innovation can happen is pretty funny. Maybe America’s hubris and hostility towards talented foreigners is driving talent towards other countries? China doesn’t need to embed bad actors and spies in Anthropic and OpenAI, the U.S. is sabotaging itself just fine.

America was at the forefront of technology because it attracted the best talent from all over the world. The talent is now staying at home, and America is discovering that America is nothing without immigration. OpenAI and Anthropic are filled with Chinese employees because some of the most talented engineers are Chinese, not because they are spies. Unfortunately, that’s the last generation of Chinese talent that America will get, the latest generation are staying in China.

Americans should be ashamed of what they have fumbled, not conspiratorial about what China is doing.

go back through all of your comments from the last decade, group them by comments that promote/mention one of your projects and comments that do not. I’d ballpark 95% of your comments have been self-aggrandizing self-promotion, it is exhausting to read. I guess today was the day I said something when I saw someone else suffering through another one of your comments. You might want to create healthier technology but you have contributed nothing to HN in almost 20 years, almost impressive enough to be an achievement unto itself. I’d eat my hat if I could find a substantive comment you’ve ever written that wasn’t self-promotion.

Read enough of EGreg’s posts and you’ll immediately recognize that this site is definitely his but at least LLMs stopped him posting about qbix all the time, a sweet mercy

You said revolution, that’s barely incremental. If you had googled “apartment electricity usage calculator” you would find many options that provide exactly the information you want without risk of hallucination.

it’s definitely a revolution in access to information

Did you mean: Google?

I think you have it backwards. Poor people cannot rely on their parents because their parents are poor too. Every person I know who grew up rich relied on their parents until… well, forever. Privilege is being able to rely on your family.

https://en.wikipedia.org/wiki/Mechanical_doping

We don’t know what’s inside these bikes!

YC has tens of thousands of founders, more are probably at Google and Facebook than Anthropic and OpenAI. I’m not sure a sample of 105 founders shows… anything?

As others have said, this is an age old idea that has been tried dozens of times and sometimes with very big financial backing (there was another attempt on HN just a few months ago). My personal view is that this idea is fatally flawed and will never work. That said, every idea sucks until it doesn’t, sometimes it’s all about the idea in the moment, maybe right now is finally the moment for this idea? So don’t be discouraged by all us naysayers, maybe your fresh perspective in this moment could make the difference.

"Corgi builds insurance structures that allow us to best serve the needs of our customers. For technology companies, operating a technology liability line through a Risk Retention Group is not unusual, improper, or exotic; it is a standard insurance structure for specialty liability risks where similarly situated businesses benefit from tailored underwriting, specialized coverage, and risk alignment. The suggestion that Corgi customers are unknowingly taking on “balance sheet risk,” member-assessment risk, or responsibility for unrelated insureds’ liabilities is false.

Your draft’s statement that “Corgi will help you share that risk,” combined with the question whether customers understand the risk of other companies in the group, does not merely describe RRGs in the abstract. It falsely implies that Corgi leaves customers exposed to open ended financial liability for other insureds. That implication is defamatory and false. RRGs are regulated insurance carriers subject to financial, reserve, governance, and regulatory requirements. They are not informal risk sharing clubs where policyholders unknowingly become responsible for each other’s balance sheets.

The RRG structure unique to Corgi. Major insurance groups use different insurer structures for different classes of risk because different risks are best served by different structures. Berkshire Hathaway, which the draft itself invokes, has affiliated insurance operations involving Risk Retention Groups in specialty liability markets, including medical and legal professional liability. That underscores the point: RRGs are a widely recognized insurance structure for specialty liability lines, and allow insurers to provide more tailored coverage options rather than issuing a generic policy.

The draft’s statement that Corgi “innovated with AI in a regulated industry by cutting corners” is also false and defamatory. Corgi raised millions pre-revenue and spent nearly two years building and obtaining regulatory approvals for its insurance operations, including approvals and requirements relating to reserves, pricing, liquidity, governance, and compliance. That is the opposite of “cutting corners.” Any allegation that Corgi used AI to evade regulatory approval, underwriting standards, reserve requirements, pricing controls, liquidity controls, or other compliance obligations is false.

Any article suggesting otherwise, including by implying that Corgi misleads customers, conceals the RRG structure, exposes policyholders to undisclosed balance-sheet risk, or uses RRGs and AI to evade proper underwriting or regulatory obligations, is false and highly damaging.

To be clear, if you publish these false statements or defamatory implications, Corgi will sue you personally and will pursue all available claims and remedies against you and any other responsible parties. Corgi has enforced its rights before and will do so again. You should not mistake this for an abstract legal reservation."

From their head of legal.

Is a startup gets insurance for something they couldn't get insurance for elsewhere and then Corgi goes belly up, the startup is our their premiums but otherwise in the same place.

That’s not accurate if the startup is making a claim against their insurance. And even in cases where the startup hasn’t yet filed a claim, losing insurance that cannot be replaced could be catastrophic. Corgi are insuring things that are uninsurable elsewhere. If your startup relies on being insured against hallucination risk, and you lose your hallucination risk insurance, then what?

For all we know, there are multiple risk groups under the hood for different risk types/profiles to insulate mispricing of different policy types.

There aren’t and there can’t be. A Risk Retention Group requires that all insured parties are equal members, Corgi cannot divide customers up based on risk profile. The entire premise of a Risk Retention Group is the risk is shared across all members. Hence, it is wildly unsuitable for how it is being used by Corgi.

Honestly, I feel like startups don't buy insurance at all unless customers ask, there's just nothing meaningful there to insure.

Newfront built a $1.2bn business on startup insurance.

If you fuck up that badly you're probably just going to go out of business even if the insurance check comes through.

We live in a brave new world. Startups are doing more and more politically and socially risky business, like AI medical advice. You are assuming that a startup being sued and losing their insurance is whatever, just shut down the company, but the founders are at risk, and losing legal coverage mid litigation has a material impact on their ability to defend themselves. Good lawyers can keep founders out of prison, no lawyers cannot.

Yes, I am the author.

Yes, insurance is mostly a box ticking exercise for most startups. My concern with Corgi is that even after accounting for how unimportant insurance is to startups, they are so blasé about their underwriting that for the small number of startups that will eventually need the protection insurance offers, there is a substantial amount of exposure to Corgi going under.

A typical startup needs D&O insurance to satisfy their investors -> the startup approaches Corgi -> Corgi's sales team negotiate more comprehensive insurance that covers many more of the risks that the startup faces that would not be insurable by traditional underwriting -> the startup uses their insurance coverage to justify risk taking.

Historically, the type of risks startups took were of little legal consequence but that has changed with AI. The social and political appetite for taking down AI companies is only getting stronger. We're already seeing OpenAI and Character.AI subject to multiple lawsuits over teenage user suicide.

All it takes is a single large judgement against a single Corgi insured company to liquidate the whole Risk Retention Group and then any ongoing litigation that Corgi was covering, is suddenly uncovered, and uninsurable elsewhere. The potential fallout from a startup losing coverage mid-litigation could be substantial when that litigation is government sponsored, the corporate veil isn't very useful when a government is looking to make an example of a company.

Multiple Corgi customers are already involved in expensive litigation and while I believe that is not covered by their Corgi policies because it predates Corgi's launch, it is a sign that expensive litigation is well within the realms of possibility for their customers.

Corgi is even worse than just circular revenue, their entire insurance business is a house of cards: https://reticulating.substack.com/p/ycombinators-corgi-insur...