The Indicator from Planet Money

Anthropic's IPO S1 reveals $33B revenue, $900B compute commitment, and complex Amazon ties

11 min Listen to the original Episode page
AIIPOS1 FilingAnthropicAmazonAGIVenture Capital

Summary

Hosts and reporter Kevin Roos examine Anthropic's S1 filing, highlighting explosive revenue growth from $41 million in 2023 to over $33 billion in the last nine months. Professor Patrick Battalato notes the company's asset-light balance sheet, with low capital expenditures compared to peers like Meta, but massive future liabilities. The episode details a $900 billion commitment to computing spend, half due within four years, and a $110 billion obligation to Amazon over ten years. Speakers discuss the binary value proposition of Artificial General Intelligence (AGI) and the existential risks disclosed in the filing.

Episode sections

  1. 0:00 – 2:40

    Anthropic's origins and early culture

    Hosts introduce Anthropic's planned public stock sale and recount early history, including co-founder Daniela Amade using stuffed animals for decision-making. Reporter Kevin Roos describes the company's initial academic feel and uncertainty about becoming a business.

  2. 2:40 – 5:07

    Anthropic's pivot to business and revenue growth

    Co-founders left OpenAI to focus on AI safety, eventually building Claude to study safety techniques. The model became popular for coding, driving explosive revenue growth from $41 million in 2023 to over $33 billion in the last nine months.

  3. 5:07 – 6:58

    Analysis of Anthropic's S1 filing and balance sheet

    Anthropic released its S1 filing, prompting MBA classes to analyze it. Professor Patrick Battalato notes low capital expenditures compared to Meta ($83 billion) and a cash-rich balance sheet with liabilities primarily in convertible debt rather than physical assets.

  4. 6:58 – 8:53

    Future risks and Amazon relationship

    Footnotes reveal a commitment to spend ~$900 billion on computing, half in the next four years. Risks include existential AI threats and a complex relationship with Amazon, which is a supplier, customer, investor, and potential competitor, with $110 billion owed over 10 years.

  5. 8:53 – 11:14

    The AGI value proposition

    The core question is whether Anthropic can achieve Artificial General Intelligence (AGI) via recursive self-improvement. Hosts argue that if achieved, it would create a binary choice for businesses: adopt the superintelligence or lose to competitors.

Claims worth verifying

Anthropic made $41 million in 2023 and over $33 billion in the last nine months.

Unidentified speaker ·4:48

How to verify: Check Anthropic's S1 filing with the SEC for audited financial statements and revenue disclosures.

Anthropic spent less than a billion dollars on property and equipment for the year so far, while Meta spent $83 billion.

Unidentified speaker ·5:57

How to verify: Compare the 'Property, Plant, and Equipment' line items in Anthropic's S1 filing and Meta's latest 10-K or 10-Q filing.

Anthropic committed to spend about $900 billion on computing, with half coming in the next four years.

Unidentified speaker ·7:07

How to verify: Review the footnotes and risk factors in Anthropic's S1 filing for capital commitment disclosures.

Anthropic owes Amazon $110 billion over the next 10 years for computing power.

Unidentified speaker ·8:25

How to verify: Check the related party transactions and long-term debt/computing agreements sections in Anthropic's S1 filing.

Anthropic's biggest liability is convertible debt that can turn into shares of stock.

Unidentified speaker ·6:40

How to verify: Examine the liabilities section of Anthropic's balance sheet in the S1 filing to identify the composition of debt instruments.

Anthropic co-founder Daniela Amade used a council of stuffed animals to make decisions.

Unidentified speaker ·0:26

How to verify: Verify via Kevin Roos's reporting in the transcript or external interviews/articles by Kevin Roos about Anthropic's early days.

Anthropic released its S1 filing last week.

Unidentified speaker ·5:07

How to verify: Check the SEC EDGAR database for the filing date of Anthropic's S1 registration statement.

Anthropic's AI systems pose catastrophic or existential risks to humanity.

Unidentified speaker ·7:39

How to verify: Read the 'Risk Factors' section of Anthropic's S1 filing filed with the SEC.

Do your own research

  • Review the full S-1 filing for detailed risk factors regarding 'existential risks' and AI safety failures.
  • Verify the $900 billion compute commitment and payment schedule against Amazon Web Services (AWS) contracts.
  • Analyze the convertible debt structure and its dilution impact on equity holders upon IPO.
  • Compare Anthropic's revenue growth trajectory and gross margins against competitors like OpenAI (if public) or Microsoft.
  • Assess the dependency risk on Amazon given the $110 billion obligation and Amazon's role as both supplier and competitor.
  • Check for any changes in the 'co-founder' governance structure or decision-making processes post-IPO.
  • Investigate the specific terms of the 'recursive self-improvement' roadmap and technical milestones disclosed in the technical reports.
Full transcriptexpand
0:00
Unidentified speaker: NPR.
0:06
Unidentified speaker: This is the indicator from Planet Money.
0:08
Unidentified speaker: I'm Wayland Wong.
0:09
Unidentified speaker: I'm Ricky Mulvey.
0:10
Unidentified speaker: Anthropic is planning to sell its stock to the public.
0:13
Unidentified speaker: This event will be a heat check on how well the AI economy is doing.
0:17
Unidentified speaker: But this company that's so well known today wasn't always trying to be a trillion dollar giant.
0:22
Unidentified speaker: And parts of its history may even seem a little silly.
0:26
Unidentified speaker: Like in its early days, one co-founder, Daniela Amade, used a council of stuffed animals to help her make decisions.
0:34
Unidentified speaker: Danielle and her husband used the council for management role-playing scenarios.
0:38
Unidentified speaker: Kevin Roos is one of our new colleagues at NPR, and he's been reporting on anthropics since its early days.
0:44
Unidentified speaker: Their basic belief is like you can sort of attribute psychology to these inanimate objects and you can sort of assign them different character profiles.
0:53
Unidentified speaker: And like one can be the generous, empathetic uh panda, and one can be the like the cat who is like lazy and wants to protect your time.
1:02
Unidentified speaker: I thought this sounded silly until I realized I talked to real animals.
1:07
Unidentified speaker: Who are you?
1:08
Unidentified speaker: Snow White?
1:09
Unidentified speaker: On today's show, we'll tell you how Anthropic quickly became one of the most important companies in the world, what its newly released financials reveal, and the questions that remain as Anthropic gets ready to go public.
1:21
Unidentified speaker: Grab your stuffy, give it a hug, and join us after the break.
1:28
Unidentified speaker: This week on the NPR Politics Podcast, a massive blue wave, a history-making realignment in American politics.
1:36
Unidentified speaker: What will Democrats do with their power?
1:38
Unidentified speaker: Both in Washington and in states around the country.
1:42
Unidentified speaker: Listen as we unpack what a new political era could look like.
1:46
Unidentified speaker: This week on the NPR Politics Podcast.
1:50
Unidentified speaker: I'm Kevin Roos.
1:51
Unidentified speaker: And I'm Casey Newton.
1:52
Unidentified speaker: And together we've been tech journalists for decades.
1:55
Unidentified speaker: Now we're your hosts of Machine Gods, a new podcast from NPR that makes sense of what AI means for your work, your life, and all of our futures.
2:03
Unidentified speaker: Twice a week, we'll be exploring the biggest moves in tech and give you unmatched access to the real titans of industry.
2:09
Unidentified speaker: Listen to Machine Gods from NPR on the NPR app or wherever you get your podcasts.
2:16
Unidentified speaker: Hey, it's Ray from Car Talk.
2:18
Unidentified speaker: You tired of all the depth and thoughtful care that goes into NPR shows.
2:22
Unidentified speaker: Want some good old-fashioned goofing around and stumbling to figure out what's going on.
2:27
Unidentified speaker: Well, I've been taking occasional car questions again.
2:30
Unidentified speaker: You can hear them by signing up for NPR Plus, along with lots of other bonus content.
2:35
Unidentified speaker: Just go to plus.npr.org.
2:39
Unidentified speaker: Time in the AI world moves in dog years.
2:42
Unidentified speaker: While you probably know Anthropic now is a tech big dog, this business is still pretty much a puppy.
2:47
Unidentified speaker: When Anthropic started in 2021, its co-founders didn't even know if they wanted the organization to be a business.
2:54
Unidentified speaker: They were just a group of employees who had left Open AI.
2:57
Unidentified speaker: CEO Dario Amade did not trust Sam Altman and OpenAI's management to develop AI safely.
3:04
Unidentified speaker: They thought of themselves as primarily interested in building safety techniques for AI systems.
3:10
Unidentified speaker: That's Kevin Roos.
3:11
Unidentified speaker: He's one of the co-hosts of Machine Gods, a podcast about AI, which is distributed by NPR.
3:17
Unidentified speaker: Kevin is also the author of the AGI Chronicles.
3:20
Unidentified speaker: And there was some disagreement between the co-founders about whether they should be a company at all.
3:25
Unidentified speaker: Kevin says that Anthropic's co-founders realized that if they wanted to study how safe other companies were being with their AI, then Anthropic would need to build its own models.
3:34
Unidentified speaker: That's how Claude was born.
3:35
Unidentified speaker: Part science experiment, part spite at a former employer.
3:39
Unidentified speaker: Relatable.
3:40
Unidentified speaker: Kevin visited Anthropic in its early days and said that it didn't feel like much of a business.
3:46
Unidentified speaker: They had like one business guy who was in charge of like selling Claude to enterprises.
3:52
Unidentified speaker: And it was a very academic-feeling place.
3:56
Unidentified speaker: There were these, you know, goofy engineers walking around.
3:59
Unidentified speaker: There was a throne at their office made of liquid death cans, like the sparkling water brand.
4:05
Unidentified speaker: It did not seem like a very serious place in terms of like their business aspirations.
4:09
Unidentified speaker: But Anthropics Claude became one of the most popular large language models in the world.
4:14
Unidentified speaker: Consumers started using it, and so did businesses.
4:18
Unidentified speaker: Claude is really good at coding, and that's a strong sell for an AI platform.
4:23
Unidentified speaker: You may pay for AI with a monthly subscription, or your job pays for it, but businesses pay by the pound how much stuff the AI programs are doing for them.
4:32
Unidentified speaker: So you pay for AI like it's Netflix, companies pay for AI like it's steak.
4:36
Unidentified speaker: And using Claude to code complex problems takes a lot of juice, or protein, or steak juice.
4:42
Unidentified speaker: Sure, steak juice.
4:44
Unidentified speaker: Either way, it's one reason why revenue for the company has exploded.
4:48
Unidentified speaker: Anthropic made $41 million in 2023.
4:51
Unidentified speaker: Over the last nine months, it made over $33 billion.
4:56
Unidentified speaker: That's just three years.
4:58
Unidentified speaker: Kevin says this kind of growth hasn't been done before.
5:01
Unidentified speaker: I think it's even been a surprise to some people at Anthropic.
5:04
Unidentified speaker: And now Anthropic is planning on going public.
5:07
Unidentified speaker: The company released its S1 last week.
5:10
Unidentified speaker: That is the document that a company releases into the world to say, hello, potential investors.
5:15
Unidentified speaker: Here's what you need to know about our company before we start selling stock to the public.
5:20
Unidentified speaker: And you know that an event is a big deal when it changes the syllabus for a class.
5:24
Unidentified speaker: This document drop even changed the class schedule for this MBA course at the University of Texas.
5:32
Unidentified speaker: Did anybody look at their statement of cash flows, other than Ryder?
5:35
Unidentified speaker: Wow, good for Ryder.
5:36
Unidentified speaker: I listened to the class.
5:37
Unidentified speaker: Ryder's been killing it lately.
5:39
Unidentified speaker: Patrick Battalato, aka P B teaches that class.
5:43
Unidentified speaker: He's an accounting professor.
5:44
Unidentified speaker: In class, they broke down what the financials revealed.
5:47
Unidentified speaker: PB was surprised at how little Anthropic had invested in capital expenditures.
5:52
Unidentified speaker: This is money that a company spends on long-term bets like building a factory or a data center.
5:57
Unidentified speaker: Anthropic put a little less than a billion dollars into property and equipment for the year so far.
6:02
Unidentified speaker: Meta, for comparison, has spent $83 billion.
6:06
Unidentified speaker: Nor do they have any massive, you know, large balance of like property plant equipment or even intangible assets on their balance sheets.
6:14
Unidentified speaker: So on the surface, Anthropic looks like a lean and mean operator in the AI race, quickly growing revenue while sort of outsourcing the expensive build out to other companies.
6:24
Unidentified speaker: Yeah, the balance sheet, which is the form that shows everything a company owns and owes, shows a company that's cash rich.
6:32
Unidentified speaker: Its biggest liability is this form of debt that can turn into shares of stock, not like lease payments on massive data centers.
6:40
Unidentified speaker: So the biggest item that this company owes is something they can pay for in stock, not cash.
6:45
Unidentified speaker: It's a pretty good situation.
6:47
Unidentified speaker: But Wayland, before you grab your motorcycle and start shooting off bottle rockets while blasting thunderstruck by ACDC, there's another story that's not cleanly told in Anthropic's financial statements.
6:58
Unidentified speaker: Financial statements alone can't tell us the full story of a business.
7:02
Unidentified speaker: We have to use the context provided in the footnotes.
7:05
Unidentified speaker: PB says that you've got to do your homework.
7:07
Unidentified speaker: Anthropic also told its investors that it has committed to spend about $900 billion on computing.
7:14
Unidentified speaker: About half of that is coming in the next four years.
7:16
Unidentified speaker: So its past financial results may not give investors a super clear insight on what's to come.
7:22
Unidentified speaker: Anthropic has already changed so much that like 2024 and 2025 feel like ancient history.
7:27
Unidentified speaker: And it's a business that carries a different kind of risk, not just for its investors, but also for you.
7:34
Unidentified speaker: Yeah, companies have to disclose their potential risks when they release their stock to public markets.
7:39
Unidentified speaker: One of Anthropic's risks is that its AI systems, quote, pose catastrophic or existential risks to humanity.
7:46
Unidentified speaker: At the end of that, they have the boilerplate, which may adversely affect our operations and you know financial performance, which it just seems like, yeah, don't think that was overkill.
7:55
Unidentified speaker: Like if we're not here, don't think anybody's gonna be mad that your margins decrease by 7% or whatever else.
8:01
Unidentified speaker: Yeah, when some robot is holding a knife to my throat, I'm not gonna be like, but Anthropic, your financial performance is suffering.
8:08
Unidentified speaker: No matter what, you protect the shareholders, Weyland.
8:12
Unidentified speaker: So this company has some big questions moving forward.
8:15
Unidentified speaker: One, can Anthropic keep growing its sales to pay for all of the servers, chips, and cooling systems that it plans on using?
8:21
Unidentified speaker: And how does the AI circular economy fit into all of this?
8:25
Unidentified speaker: For example, Anthropic owes Amazon $110 billion over the next 10 years because it is buying computing power from it.
8:33
Unidentified speaker: But it's not a normal company relationship here.
8:36
Unidentified speaker: Amazon is one of Anthropic's suppliers, Amazon is also one of Anthropic's customers, Amazon is also an investor and also potentially a competitor to Anthropic as well.
8:47
Unidentified speaker: Where does it begin?
8:48
Unidentified speaker: When does it end?
8:49
Unidentified speaker: No evidence of wrongdoing here, but a unique, a very unique arrangement.
8:53
Unidentified speaker: And so it's like there's a host of companies that were like there's multiple relationships with Anthropic.
8:59
Unidentified speaker: And the biggest question for Anthropic supporters and really society is can this company create artificial general intelligence?
9:08
Unidentified speaker: Or in this case, AI models that are able to improve all by themselves without much human involvement?
9:14
Unidentified speaker: Kevin with Machine Gods says that reaching this level is possible and would fundamentally change the economy.
9:20
Unidentified speaker: If you hit that cycle of recursive self-improvement before your competitors, and your temporary advantage becomes like durable and compounding because you are always just faster and better at everything than your competitors.
9:34
Unidentified speaker: So if you've ever asked how can these AI companies make more money, one answer is that businesses in every sector of the economy would have to plug into this super intelligent system or get left behind.
9:46
Unidentified speaker: Yeah, you either let godlike superintelligence into your insurance company or your competitors win.
9:51
Unidentified speaker: How are these the choices?
9:52
Unidentified speaker: Those are your only two options, really.
9:54
Unidentified speaker: This is the binary.
9:56
Unidentified speaker: Okay, great.
9:57
Unidentified speaker: So will we get artificial general intelligence solving humanity's greatest problems, or will the music stop before this happens?
10:07
Unidentified speaker: And by the way, Anthropic is also an NPR sponsor.
10:10
Unidentified speaker: This episode was produced by Corey Bridges and Rhea Chowdery.
10:13
Unidentified speaker: It was engineered by Jimmy Keeley and fact-checked by Leila Dos.
10:16
Unidentified speaker: This episode was edited by Julia Ritchie, Kate Kincanan edits the show.
10:19
Unidentified speaker: The Indicator is a production of NPR.
10:28
Unidentified speaker: Midlife brings a lot of changes.
10:30
Unidentified speaker: Your turn with Alice Naubre is a new podcast to help you navigate the twists and turns.
10:35
Unidentified speaker: From health and relationships to creativity and purpose, we can help steer your next chapter.
10:41
Unidentified speaker: With your turn with Alice Naubre.
10:43
Unidentified speaker: Find it on the NPR app or wherever you get your podcasts.
11:08
Unidentified speaker: I said, but at a high cost.
11:10
Unidentified speaker: Find it on the NPR app or wherever you listen to Fresh Air.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3 and interpreted by Qwen3.6-35B-A3B-GGUF:UD-Q4_K_XL. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from The Indicator from Planet Money

The Indicator from Planet Money

How a tribal water deal could put a price tag on the Colorado River

Episode discussion: Federal proposal to fund tribal water infrastructure in Colorado River basin faces political opposition

Speakers’ claims: The episode discusses a federal settlement proposal aimed at addressing the lack of reliable clean drinking water for approximately half of the homes on tribal land, citing the Hopi people as an example. It highlights the complex dynamics of the Colorado River basin, which supports 40 million people and 30 tribes, noting that while tribes hold legal rights to a quarter of the water, they lack the infrastructure to utilize it. The Navajo Nation's geographic position straddling the Upper and Lower Colorado River divide is identified as a complicating factor for water leasing. Economists argue that the proposed deal could enhance economic efficiency by allowing water to shift from low-value agricultural uses to high-value urban uses. However, Upper Basin states including Colorado, New Mexico, Utah, and Wyoming oppose the agreement due to fears that it sets a precedent for water transfers to Lower Basin states.

Water RightsTribal InfrastructureColorado River
12 min0 assets0 claims to verify
The Indicator from Planet Money

Three influencer types and how they generate money

Episode discussion: Podcast analyzes three influencer monetization models: hucksters, entertainers, and gurus

Speakers’ claims: The episode explores the creator economy, citing IAB data that 1.5 million US residents are full-time creators. It introduces Stanford Professor Angèle Christin's framework categorizing influencers into hucksters, entertainers, and gurus based on revenue streams. Hucksters rely on brand sponsorships, entertainers chase platform revenue shares through viral content, and gurus monetize directly via subscriptions and merchandise. The hosts note that despite industry growth, a CreatorIQ survey reveals most creators earn under $10,000 annually and face significant burnout.

Creator EconomyInfluencer MarketingSocial Media
11 min0 assets6 claims to verify
All insights