🕶️ $1B startup, zero salespeople

Good morning, founders 🕶️

Every startup playbook comes with a price list: a sales team before you scale, an equity round before you grow, an Nvidia bill before you ship, a support desk standing between customers and the founder. This week, founders kept crossing lines off that list without the sky falling. Daytona got to $1B with no salespeople. Otter raised $320M without selling a share. Google says you can leave Nvidia without rewriting your code. Ring's founder still reads customer email himself. The expensive defaults are turning optional. So which line on your budget is a requirement, and which one is just habit?

In today's newsletter, we'll get into:

He has a newsletter on beehiiv. So do Codie Sanchez, Jay Shetty, David Begnaud, Colin & Samir, and Joanna Stern.

They could publish just about anywhere. They chose beehiiv to build a direct connection with the people who want to hear from them most. And you can do the same, whether you’re world-famous or just at the start of your plan for world domination.

Get started for free.

Daytona has about 30 employees, a $1 billion valuation and zero salespeople. It does have a basketball court, sort of: it packed 1,200 people onto the floor of San Francisco's Chase Center for its own conference.

The receipt: a $60M Series B led by FirstMark, taking total funding to $91M. Daytona runs sandboxes, isolated computers where AI agents run code and use software, for Anthropic, OpenAI, Nvidia and Cursor. Its revenue run rate grew 25x between January and September, per tech.eu. CEO Ivan Burazin says growth has run just shy of 51% a month for 18 months. Compounded, that's about 1,700x, so we'll stick with the 25x.

The customers, mostly. Daytona runs about 1,100 Slack channels with them, plus dinners, hackathons and meetups across the US, Europe and Asia. Burazin's X account, about 20,000 followers, is the second growth channel; one tweet that hit 2 million views last Christmas helped lead to a partnership with Nvidia. Burazin has run this play before: he founded Shift, a developer conference Infobip later bought.

The pricing does the closing. Sign up, add a credit card, pay for what you consume, and call only when you need higher limits. On consumption pricing, every customer who grows is an upsell nobody had to schedule.

The cap table reads like a founders' group chat: Patrick Collison, Tobi Lütke, Olivier Pomel, Aravind Srinivas, Eric Yuan. Pace doubled its check, Upfront tripled, E2VC quadrupled. Existing investors don't do that out of politeness.

Here's the part that stings for sales-led infra startups: if your buyer is a developer, a credit-card signup and a shared Slack channel can do a first sales hire's job. The catch: it works best when the founder is already famous in the room.

Daytona just agreed to hire its first head of sales. Watch what happens when 1,100 Slack channels meet their first quota.

💡 Investors want startups to lose money again.

After 2023 punished unprofitable companies, the script has flipped. In this episode, Caya breaks down why profitability suddenly matters less, and how AI has reshaped fundraising so that more money is flowing to fewer startups.

We explain the current funding cycle, why milestones matter more than burn right now, how seed and pre-seed rounds quietly ballooned into old-school Series A sizes, and what founders actually need to optimize for if they plan to raise again.

✏️ Your GPU bill just got two new pen pals. On Monday, Amazon shipped TorchNeuron, a way to run PyTorch on its Trainium chips. On Tuesday, Google open-sourced TorchTPU, a version of PyTorch that runs on its own Tensor Processing Units. Both promise you can move an existing PyTorch app with no rewrite; Google calls it a simple configuration change. PyTorch runs about 85% of AI apps, by some accounts, and much of it was built around Nvidia GPUs from the start.

Follow the chip. When you rent an Nvidia GPU from Google Cloud, Google first bought that GPU from Nvidia, so a slice of your hourly bill began life as Nvidia revenue. On a TPU, Google owns the silicon, the data center and the invoice. ✏️ Read that as: the clouds aren't freeing you from Nvidia out of charity. They want the margin Nvidia keeps.

There's a second catch. Moving your app to TPUs also moves it to Google Cloud. The exit from one lock-in opens straight into another. The more public-spirited move is Amazon's: it's submitting code to the main PyTorch project that decouples features from Nvidia chips, which would make every app more portable by default, not just portable to AWS.

Treat it as a promise. TorchTPU's public release is version 0.2.0, and its GitHub repo lists 256 open issues. Even TechCrunch expects no mass exodus from Nvidia.

So here's your move: this week, run one PyTorch workload on a TPU and one on Trainium. Write down cost and speed, then bring those numbers to your next GPU renewal. A credible threat to leave is the cheapest compute discount there is.

✏️ Watch whether Amazon's patches land in mainstream PyTorch. If they do, leaving Nvidia stops being a cloud's sales pitch and becomes a setting anyone can flip.

Scale your growth: Join names like HubSpot, Ahrefs, and Deel. Put your product in front of 85,000+ startup founders. Let's talk.