Anthropic's leaked filing: $42B loss, $518B bill, $2T target

Good morning and happy Sunday, crowd. ☕

Yesterday was about who raised. Today is about what it all means.

This week, a leaked IPO filing showed us, for the first time, what a frontier AI lab really costs to run. A founder publicly fired his VC adviser over alleged spying. And the exit window opened wide for exactly two kinds of companies: the giants, and the ones getting bought by giants.

Grab your coffee. This is the one to read slowly.

1. 📄 Anthropic's IPO filing puts real numbers on the AI arms race

2. 🪜 The exit ladder in 2026: IPO if you're huge, M&A if you're not

3. 🕵️ Factory vs. Khosla: what happens when your VC backs your rival

4. 🧩 The equity mistake most founders make in year one

5. 🏦 VC money keeps coming, and Sequoia just opened its LP playbook

Read time: 5 mins

Elon's Building Something in Tesla's Secret Labs.

Something is being built inside Tesla's facilities that almost no one is talking about — yet.

According to insider sources, Elon Musk has quietly developed a he claims will be "10x bigger than the largest product in history." The target launch date is July 22. And when it drops, the window to position quietly will already be closing.

Most investors will hear about this after the stock has moved. You don't have to be one of them. Our analyst named positioned to ride the launch — with entry guidance, price targets, a bonus 4th supply-chain pick, and a 3-phase playbook for when to buy, add, and take profits.

NEWS OF THE WEEK

Memeburn

For years we've guessed at what frontier AI costs. This week we got the receipts.

Reuters obtained Anthropic's IPO prospectus, and the headline numbers are staggering. Revenue grew roughly 12x in 2025 to about $4.6 billion. The company also posted a net loss of around $42 billion.

But here's the thing most headlines skipped. About $34 billion of that loss is an accounting charge tied to financing that could eventually convert into shares. It's not money Anthropic spent running the business. Strip it out and the operating loss was over $8 billion. Still huge, but a very different story.

The bigger number is what's coming. Anthropic lists about $518 billion in cloud, compute, and infrastructure obligations ahead. It had $20.28 billion in cash and short-term investments at the end of 2025. You don't need a calculator to see why an IPO matters.

Now the part every founder should study. 47% of Anthropic's sales last year flowed through Amazon and Google's cloud platforms, which collected roughly $351 million in fees, or around 16 cents on every dollar of those sales. OpenAI has argued this way of booking revenue inflates the topline. Anthropic says it follows standard accounting because it's the principal in the sale. And the filing warns that many big customers aren't locked into long-term contracts.

There's also a twist you won't see in most S-1s. About 80 of the filing's 261 pages cover risk factors, including warnings that advanced AI could pose catastrophic risks. That's more pages on risk than on the business itself.

The IPO is expected after the November midterms and could value the company above $2 trillion, up from $965 billion in May. One analyst noted that works out to roughly 18 to 20 times sales, which is rich but not crazy for this growth.

For founders, the lesson is about distribution. Even the fastest-growing company in AI routes nearly half its sales through partners who take a cut and who also compete with it. Channel partners can build you fast. They can also own your customer. Know which one is happening.

CNN

Put this week's headlines side by side and a clear pattern shows up.

At the top of the ladder, the window is wide open. SpaceX went public on June 12 at a $1.77 trillion valuation, priced at $135 a share, popped 19% to $160 on day one, and still trades around $147. Anthropic is next, likely aiming above $2 trillion. OpenAI is raising a $30 billion bridge at $1.4 trillion, and the only realistic buyer at that size is eventually the public market.

One rung down, things get shaky. On Thursday we covered Oura postponing its Nasdaq IPO despite targeting up to $14 billion. And this week Monzo was reported to be in talks over a £10 billion sale to Nubank, a deal that could end its long-planned London listing.

Then there's the M&A route. AMD bought World Labs for $8.2 billion in stock. No roadshow, no lockup, no waiting for the window.

So what's actually happening? Public investors want two things right now: mega-scale AI exposure and proven profitability. Companies with either get a warm welcome. Everyone in the middle, the $5 to $50 billion companies with good growth and real losses, faces a market that wants a discount they don't want to give.

That has real consequences for the private markets. Employees wait longer for liquidity. LPs wait longer for distributions. Secondary markets get busier as people sell to the patient buyers. And strategic acquirers, sitting on expensive stock, get to shop for talent at a fair price.

For founders, the playbook is simple. Build as if you'll never IPO, and know exactly who your three most likely acquirers are. If a giant could use your company to save two years of building, you already have an exit strategy. You just need to make sure they know it.

Khosla

This might be the most uncomfortable board story of the year.

Matan Grinberg, co-founder and CEO of AI coding startup Factory, posted on X that he had fired VC Chris Degnan from his role as a board adviser. Grinberg alleges Degnan shared confidential information with Factory's biggest competitor, Cognition.

Here's where it gets messy. Khosla Ventures has invested in both Factory and Cognition, and in Cognition since at least early 2025. Then Vinod Khosla himself jumped in, seemingly defending Cognition, the larger portfolio company. And Keith Rabois, a partner at the same firm, publicly took Factory's side.

Same firm. Two partners. Two portfolio companies. Opposite sides of a public fight.

We can't say who's right on the facts. But the structural problem is real and it isn't new. When a fund backs two companies in the same category, the founder who matters less to that fund's returns is taking a risk. Information doesn't need to be "stolen" to be useful. A board adviser who sits close to both companies simply knows things.

In AI coding, where products change every few weeks and every roadmap detail is a competitive weapon, that risk is magnified.

For founders, this is a governance checklist, not gossip. Before you take money or add an adviser, ask directly which competitors they've backed or advise. Put information rights and confidentiality terms in writing, and limit what advisers see to what they need. Board seats and adviser roles feel like relationships, but they're legal positions. Treat them that way from day one.

And for investors, the reputational cost of a public fight like this almost always exceeds whatever a portfolio overlap was supposed to gain.

Here's a quiet mistake that costs founders more than any bad round.

Most first-time founders get equity backward. They're generous with advisers who show up once a month and stingy with the first employees who show up every day.

The better approach flips it. Go to the top of the range for hires one through three, and be careful with almost everyone else. Your first engineers and operators are taking real career risk on an unproven company. They'll build the product, set the culture, and likely recruit the next ten people. An adviser who takes a call every few weeks is helpful, but rarely irreplaceable.

This connects to two other ideas making the rounds this week. First, top-down management fails with great talent. The best people don't follow orders just because you're the boss, and they can leave whenever they want. Equity is part of how you earn their judgment, not just their hours.

Second, the guerrilla hiring playbook. When you can't match Big Tech salaries or brand names, you compete on ownership, speed, scope, and mission. Meaningful equity for early hires is one of the few weapons a startup has that a giant can't easily copy.

And notice how this ties back to Factory. An adviser with a small equity stake and a large stake elsewhere can have very different incentives than you'd assume. Equity aligns people only when it's meaningful relative to everything else they own.

For founders, the takeaway is concrete. Audit your cap table this week. If your advisers collectively own more than your first three hires, you've likely got it backward. Fix it in your next grant cycle, before it shows up in who stays.

Carta

While founders worry about exits, the money to fund new companies keeps arriving.

Headline, an early backer of Mistral, closed a $400 million fund as European AI seed rounds passed $1 billion. Friends & Family Capital, co-founded by Palantir's CFO from 2011 to 2017, the executive who helped grow it into a $20 billion business, raised $300 million for its fourth fund. And DIG Ventures, from MuleSoft's founder, closed a $120 million Fund III to back 30 AI infrastructure startups.

The pattern: specialists with sharp theses and deep networks are raising easily. Generalist funds without a clear edge are the ones struggling.

Then there's the move everyone in venture is talking about. Sequoia partner Pat Grady recorded and publicly released a briefing Sequoia gave one of its LPs, Boston College. LP meetings are usually the most private conversations in venture. Publishing one is a statement: Sequoia is confident enough in its strategy to show its work in public.

Here's why that matters beyond Sequoia. Venture firms increasingly behave like media companies. They publish, they post, and they build audiences, because founders now pick investors the way customers pick brands. Transparency is becoming a recruiting tool for the best deals.

For founders, two takeaways. First, there's capital available, but it's concentrated with investors who have a specific point of view. Pitch the fund whose thesis you fit, not the biggest name. Second, watch what top firms tell their LPs when they go public. It's the clearest signal of where they'll deploy next.

⚡ Quick Hits: Worth Your Sunday

🇧🇪 Belgium's tech sector raised €1.09 billion in the first half of 2026, but most of it went to a single company.

🇮🇳 Bessemer Venture Partners added Mayank Khanduja as a partner in its India practice.

🎓 The IIT Madras-Unicorn India Frontier Fund announced a ₹450 crore first close.

🤖 Anthropic shipped its new Opus 5.5 model just ahead of its expected IPO, as competition with OpenAI heats up.

📉 Product marketers are getting rejected despite strong skills, mostly due to broad positioning and forgettable interviews. Sharpen the pitch and show how you'd solve the company's actual problem.

👀 Startups hiring now: Rillet (AI accounting), Alpaca (trading infrastructure), Legora (AI for lawyers), Tenstorrent (open AI chips), and Hark (personal AI hardware).

💬 What This Week Tells Us

The AI economy is being financed at a scale we've never seen, and the bill is now public. The winners are getting bigger, the middle is getting squeezed, and governance is getting tested in real time. In a market like this, the edge isn't access to capital. It's knowing exactly where you sit on the ladder.

Question for you: Would you buy Anthropic at a $2 trillion IPO valuation, or wait to see what the market does after the midterms? Hit reply. We read every one.

Hunt accordingly. ☕

Jose, Alberto & the Founderscrowd Team