Ditlev

Me: I think we should set ourselves a slightly ridiculous target before the end of the year. X: How ridiculous? Me: $250 million in signed customer contracts. In 57 days. X: You only started selling in March. Me: We signed $135M in October. X: So you need to almost double that. In less than two months. Including Christmas. Me: Yep. X: And you think you can do that? Me: No idea. There’s a pretty good chance we won’t. X: So why do it? Me: Thought it would be fun. And if we make it, the team shares a $1M bonus. X: And if you don’t? Me: Then we fail. Quite publicly, actually. X: Publicly? Me: Yeah. Built a website with a live scoreboard so everyone can watch. X: You do realise most people wait until they’ve achieved something before announcing it? Me: Where’s the fun in that? :) 250m.live #BuildingInPublic

We've given ourselves 57 days to sign $250 million in customer contracts. And if we make it, our team shares a $1 million bonus. Split equally, everyone, no exceptions. This might be a terrible idea. But where's the fun in playing it safe? #BuildingInPublic, right? :) We only really started selling at @hosted_ai in March this year. In October we signed $135M in customer contracts. So naturally, we decided to see if we could do something even crazier before the end of the year. That's $4.4M a day. Every day. Including Christmas. And rather than announcing the result afterwards, we're doing the whole thing in public. Live scoreboard: 250m.live Starting from zero today. No idea whether we'll make it, but it'll be fun finding out. And if you're in the market for AI infrastructure, or the software to build and run your own AI cloud, I suspect you'll find our commercial team unusually motivated to get a deal done before December 31 :) Oh, and most of the team haven't heard about this yet. So if you're reading this and you work at hosted.ai... surprise! :) Wish us luck!

Jeg havde ikke regnet med, at jeg skulle ende med at forsvare @zeniastampe ... but here we are. Jeps, jeg er dybt uenig med hende i stort set alt, hvad hun sagde om AI i Debatten. Jeg synes, hun grundlæggende undervurderer teknologien og den enorme forandring, vi alle står midt i. Og jeg synes, hun er lidt af en utopist, når det kommer til, hvor meget vi kan styre den udvikling. Men ved I hvad? Der skal sgu også være plads til utopister. Til mennesker, der insisterer på, at verden bør se anderledes ud, også når vi andre mener, at de er helt galt på den. Og den kritik, hun efterfølgende har fået for sin opførsel? Den er jeg mindst lige så uenig i. Jeg har set debatten, og jeg synes ærligt talt, det er sørgeligt at se, hvor meget vrede og hån hun er blevet mødt med på LinkedIn og alle mulige andre steder. For jeg synes faktisk, det var befriende. Her var en politiker, der tydeligvis brændte for noget. Som blev vred, frustreret og måske også lidt for personlig. Som ikke hele tiden tænkte over, hvordan hendes næste sætning ville se ud i et 20 sekunders klip på sociale medier. Et menneske, med andre ord. Et menneske, der var villig til at risikere at ligne lidt en idiot på landsdækkende tv for at forsvare noget, hun oprigtigt tror på. Det har jeg faktisk ret stor respekt for. Og jeg tør også godt sige, at jeg har en mistanke om, at der er noget køn på spil her. Når jeg ser ord som "skinger" brugt om hende igen og igen, kan jeg ikke lade være med at tænke på, om en mand med samme temperament ville være blevet mødt på helt samme måde. Måske. Men jeg tvivler. Vi bruger så meget tid på at brokke os over politikere, der er medietrænede, polerede og aldrig rigtig siger noget. Men så snart en af dem smider manuskriptet og viser lidt ægte temperament, står vi klar med høtyvene. Selvfølgelig skal man kunne kritisere en minister. Også for hendes tone. Men der må altså være forskel på at være uenig med nogen og på at gøre dem til genstand for en offentlig gabestok. Jeg vil til enhver tid hellere diskutere med en politiker, der tager fejl med hele hjertet, end en, der aldrig tør mene noget af frygt for at tage fejl. Så Zenia, vi bliver nok aldrig enige om AI (eller meget andet). Men respekt for at du tør stå ved det, du tror på. Og i øvrigt også respekt til @thorborg for at gøre præcis det samme. Han er jo kendt for at sige tingene, som han ser dem, også når det larmer lidt. Det er en stor del af det, vi elsker ved Martin. Måske skulle vi også kunne rumme det hos Zenia? Og måske er det lidt ironisk, at det mest menneskelige, jeg har set fra en politiker længe, kom i en debat om kunstig intelligens.

A huge congratulations to the whole team at @Creandum on their new $750M fund! They led our last round at hosted·ai and I couldn't be happier to have them on board. Carl and team are great people, humble, genuinely supportive, and a real pleasure to work with. If you're an early-stage founder and Creandum wants to talk, take the meeting.

Funny choice of animal. @elonmusk predicts "guppy-level death" for neoclouds. Guppies are actually one of evolution's great survivors. They're small, reproduce quickly, adapt remarkably fast, and have managed to thrive in waters full of much bigger predators. Sound familiar? The big clouds have billions in debt, enormous fixed costs and infrastructure decisions that take years to reverse. The smaller neoclouds can move faster, adapt to changing demand, experiment with new business models and exploit opportunities the giants don't even notice. Sure, plenty of guppies will die. That's how evolution works. But the species survives. And sometimes the little fish are considerably better adapted to the environment than the sharks. My money is on the guppies...

I feel lucky. I’m 53. I’m building my seventh company. And somehow, I think I’m living through the most interesting decade of my life. Everything is happening at once. AI, energy, geopolitics, the whole compute build-out. It would be easy to feel overwhelmed by it. Mostly I feel the opposite. I get to wake up every morning and work on problems I don’t know the answers to yet. After 25 years of building companies, that still feels like a gift. I get to live and work across Asia, the US and Europe. Malta one week, Bangalore the next, then a call with the US at an hour I’d rather not admit to. It’s tiring, and I wouldn’t swap it for anything. I get to work with my son. Few things in my life have made me prouder, or kept me more honest. And I get to build another company with people I’ve known and worked with for years. We only started selling hosted.ai for real in March. This month we’ve signed more than $100M in contracts. Actual customers, actual GPUs doing useful work instead of sitting idle. Most of this team has built companies with me before. They didn’t have to do it again. They did anyway. That means more to me than I probably tell them. None of this was guaranteed. Plenty of people smarter than me have had worse luck, worse timing, or simply didn’t get the health to keep going. I try to remind myself of that. Because the one thing I really don’t want is to get so used to this life that I forget how bloody lucky I am.

I've got 1000+ B200's ready to hand over, today @ 5.90/hr/GPU. >> [email protected] <<

X: you're selling RTX PRO 6000s for $199 a month? Me: yep. Flat rate. Monthly. On packet·ai X: the Blackwell one? 96GB? Me: the DC edition. 96GB GDDR7. X: that can't be right. Those things rent for $1.50-2.50 an hour everywhere else. Me: I know. X: so at 24/7 usage that's... less than 30 cents an hour? Me: $0.27 if you want to be precise. X: how? Me: same answer as always. Multi-tenancy. hosted·ai runs the scheduling and overcommitment. Average VRAM usage across our GPUs is under 20GB. On a 96GB card, that's a lot of headroom. X: so you're overselling? Me: we're utilising. Airlines oversell seats. Cloud providers oversell CPU and RAM. We do the same with GPU - but smarter, because we actually measure and schedule it. X: and the $199? Me: it's the Digital Ocean moment for GPUs. Simple pricing. No calculator. No reserved instances. No "contact sales." You just pick a GPU and go. X: who's this for? Me: developers. Tinkerers. Small teams that need serious VRAM but don't want to deal with hourly billing anxiety. Fine-tuning, inference, dev environments - 96GB handles a lot. X: what's the catch? Me: limited nodes. We're not AWS. But what we have is real and it's running. X: weren't you selling these at $0.66/hr before? Me: we still do. But some people just want a flat monthly number. No surprises. Like renting a flat vs paying by the night. X: you said you got 3 sales in the first hour last time Me: before we were even ready. First deployment had to be handheld. That told us something. X: what? Me: that simple pricing wins. People don't want to do GPU math. They want a number they can budget for. X: where? Me: packet.ai/blackwell

Even though I miss the ‘old’ @Radiolab it’s still the best thing you can put in your ears: open.spotify.com/episode/11q…

Can I show what we’ve built at hosted.ai?

best AI fake I've seen yet...I mean, it's gotta be AI, right?

Hvad med at prøve det her næste gang du har middag med vennerne? Super lækker og nem mad fra Jesper Vollmer og chefmade.dk

Do I know anyone in or around Sidi Ifni in Morocco?

Raising less than $500k for your startup? At Greenroom.io we're getting a lot of $1-3M capital rounds listed, but I'd like a larger percentage of even earlier stage startups. Add your startup, or tag someone who you know could use the help. It's free.

Are you currently raising a seed round? I am building greenroom.io and want to help 100 startups raise capital in 2023. Add your startup and let's talk. No cost.