How Pieter Levels Passed $10M a Year — With a 94.5% Margin and ETF + Startup Bets
On 23 September 2026, Pieter Levels posted a number the indie internet is trained to flatten. He said he had passed $10 million a year. The line that followed is the one that matters: that figure is revenue plus investment gains, not the businesses alone.
The screenshot people will share is a ten. The sentence under it is a mix. Operating products on one side. Paper and cash from markets on the other. He did not publish a dollar split of the ten. This page does not invent one.
The blog mirror says the post first went up on X. Treat the number the way you treat any public founder ledger: his report, useful as a shape, not a filing.
The shops stayed in the same band
The products did not suddenly become a $10 million company. In the same post he put business revenue in a familiar band: about $200,000 to $250,000 a month, still.
That is the same neighborhood as a public checkpoint from a year earlier. On a July 2025 Cheeky Pint with John Collison, the ops picture was about $3.1 million a year. Different window. Same order of magnitude. The 2026 headline is not a new Remote OK.
He also said investment returns had been larger than the operating businesses for about a year. Again, no split of the $10 million. Only the ranking: gains over shops, for a while now.
The margin is the engine
The other number in the post is the one operators will underline: a claimed profit margin of about 94.5%.
That margin is what happens after he spent 2026 cancelling the middle of the stack. Scrapers, vision APIs, maps, human mods — vibe-coded off the card and onto a VPS, until the leftover list was a handful of vendors. The receipts for that cut are a different story. The short version here is the one he has been writing since July: cancel the subscriptions, keep domains, servers, storage, and model APIs. Lower burn. Higher keep. More cash sitting still.
He described the loop in one direction. Spend less. Costs fall. Profit rises. More of that profit can sit in markets. The percentage return, in his telling, does more of the work than another feature.
This is not a guide to emptying your own card. It is the cost path he attached to the 94.5% line.
Most of the ten is ETFs
Where the extra cash went is the rest of the post. The bulk, he said, sits in simple ETFs — the example he named is a Vanguard S&P 500 fund.
He has walked the brokerage version of that idea before. A 17 July 2026 note on Interactive Brokers splits the tickers the way a passport splits: VOO, QQQ, SGOV, and IAU if you can buy the US line; VUAA, XUTC, IB01, and RMAU if you cannot. That post is his shopping list, not this site’s. He ended it the way this page will end it: this is not financial advice.
The same IBKR note carries the cautionary pair he likes to put next to a win. NVIDIA on one side. Alibaba, in his telling, down about 90% on the other. Startups, he wrote there, should not take more than about 5–10% of the book.
Then he bought the tools he already used
The satellite book is the interesting part, because it is not a thesis deck. It is a usage log with a wire attached.
In 2022 he could not get NVIDIA GPUs. So he bought the stock. A shortage in the machine room became a line in the brokerage.
In 2023, Danny Postma pointed at the inference under the Photo AI-class apps. Levels put money into Replicate, then FAL. The page he keeps for that book, levels.vc, is an angel list he updates himself. He is clear it is not a real fund. The rule on the page is the same rule as the $10 million post: invest in what he already uses.
Replicate, as listed there, later went to Cloudflare in November 2025. FAL stayed on the list. So did names he has attached to products he ships or opens every day.
In 2025 he ran the same question through every app and service on his own machine. The answer that time was Cursor. He has said he joined the Series B in March 2025. In June 2026 he wrote his first real investment exit: Cursor, in his telling, to SpaceX. On levels.vc he frames that deal at $60 billion. That figure is his page’s framing, not a size this article can verify.
December 2025 added another public buy. He said he bought more than $1 million of Google, used ICHN for China exposure, and reduced the NVIDIA line. Same habit: the tool, the bottleneck, then the ticker.
The levels.vc list also names Photo AI, Interior AI, Perplexity, Sync.labs, Immortals, EthSystems, and others. The percentages on that page move. This story does not freeze a live return in a headline.
He does not sell, and he says the crash is coming
Most of the gain, he said, is unrealized. He also said he never sells.
That is the asterisk under the ten. A mark-to-market year can look like a lifestyle. It can also give the year back. He said he expects crashes in some years. A big grain of salt belongs on the whole stack: the $10 million mix, the 94.5% margin, the angel exits, the ETF pile.
None of this is a recommendation to copy the book. He has said the same. This is not financial advice.
Sleep score on the same dashboard
The character beat is the dashboard. He tracks the revenue-and-gains line, the profit margin, and his sleep score on one screen.
That is the one-person company as he has always sold it, taken one step further. The shops, the brokerage, and the body in the same pane. A night of bad sleep sits next to a 94.5% margin the way a GPU shortage once sat next to an NVIDIA buy.
The useful part of the $10 million post is not the invitation to treat a screenshot as a plan. Most makers do not have a decade of PHP products throwing off a quarter million a month, or an angel book that happens to include a tool SpaceX later wanted. The useful part is the settlement he keeps repeating: stay lean enough that the leftover cash has somewhere to go, then do not confuse a high year in the markets with a new kind of product.
The businesses, in his telling, are still the same shops. The new plot is what a 94.5% margin lets a solo operator do with the extra.
A note on the numbers
The $10 million a year in revenue plus investment gains, the about-94.5% profit margin, the about-$200,000 to $250,000 monthly business band, the claim that investment returns outran the shops for about a year, the ETF-and-startup mix, the “never sell” line, and the sleep-score dashboard are Levels’ own reporting, as stated in the 23 September 2026 post. The roughly $3.1 million a year ops baseline is from the July 2025 Cheeky Pint, as he mirrored it. The more-than-$1 million Google buy, the ICHN and NVIDIA trim, the IBKR ticker list, the 5–10% startup cap, the NVIDIA-versus-Alibaba pair, the Cursor exit, and the $60 billion SpaceX framing are from his later posts and from levels.vc as listed there. This article is commentary, not an audit, and not financial advice. If a figure matters to you, read it from him.