Welcome to Startups Weekly. Sign up here to get it in your inbox every Friday.
Borrowing has become more expensive, and profits are harder to come by, which means that 2023 has been a savage year in startup land. PitchBook data suggests that around 3,200 startups — representing a total of $27.2 billion in venture funding — have gone under, with a significant number of startups being in zombie mode: Unable to grow, unable to raise money, but just about limping by well enough to avoid shutdown. Layoffs are happening — also in December — and over the next couple of weeks, a bunch more startups will shut down, so as to not drag out the inevitable into a new tax year. I’ll be looking into this more deeply over the next couple of weeks, so stay tuned.
Also: I was supposed to be writing this newsletter but instead ended up playing the TechCrunch pub quiz for way too long. My score was . . . embarrassingly poor, considering that I’ve literally read every story on the site for the past year to write this newsletter. Still, it was fun — give it a whirl!
When the AIs come marching in
I love it when my colleagues are going super deep into nerd land. That’s definitely one of the hallmarks of Devin’s work from time to time; in this case, he talks about how “Star Trek: Deep Space Nine” fans are using AI to make the old show look better because there’s no official high-quality version. They’re using AI to add details to the original episodes, which is tough and takes a lot of effort — but it’s showing a bunch of promise. Devin concludes that the tech could be a cool way for companies to upgrade old shows, but there are some legal and technical hurdles to figure out. Don’t miss his 3,000-word ode to de-grainification.
The other AI-related nerd-out this week comes courtesy of Ron, who dug into the continued relevance of traditional AI models in enterprises, despite the rise of large language models (LLMs). That makes sense: LLMs are kind of the Leatherman of AI tools: They sort of do everything. I never leave the house without my Leatherman, and it has helped me out of many a knotty situation, but if I’m building a house or repairing a car, I break out the more specialized tool kit.
More startuppy AI news this week:
This really moved me: Just when you thought your online photos were safe, here comes Animate Anyone turning them into eerily lifelike, video deepfakes — because regular old photo fakes weren’t unsettling enough.
G-oops-le: Google’s new AI model Gemini isn’t exactly hitting it out of the park, with early users finding more bloopers than brilliance in its answers. Turns out, even Google can have an “oops” moment in the AI world.
The Pokémon approach to startups: Elon Musk, seemingly never tired of starting new ventures, is now chasing a cool $1 billion for his latest AI escapade — xAI — because why settle for running just a few companies when you can add another AI startup to your collection?
This week in Elon Times
Look, I’m as bored of Elon Musk as everyone else, but gotta give the guy credit for one thing: He doesn’t half attract some attention. Rarely for good reasons, recently, it must be said.
Darrell summarizes the situation in his piece “The end of Elon,” where he — tongue firmly planted in cheek and with the snark meter turned to 11 — dissects the Tesla Cybertruck launch (spoiler: It was a bit of a nothingburger; there’s still much unknown about the truck) and Musk’s, er, unique approach to managing his various ventures — including telling X (formerly Twitter) advertisers to go do something anatomically improbable.
Of course, there was (much) more Musk-related news this week, and if you want it all, give our Elon Musk tag a quick scroll.
What goes up . . . : SpaceX drops $2.2 million on a parachute company, because apparently making parachutes that don’t buckle in space is harder than rocket science.
Keep on truckin’: The Tesla Cyberbeast: Heavy, quick, and falling a bit short in towing compared to its high-priced electric rivals — but hey, who’s counting when you’re driving an angular beast?
Show me the money: X has scored licenses for payment processing in 12 U.S. states, inching closer to Musk’s vision of turning the platform into an “everything app.” With recent advertiser exits and controversies, it seems there’s more drama than dollars in Musk’s grand plan — for now.
After the heyday of 2021, a bunch of startups are crashing to the ground after failing to meet their goals. Let’s have a moment of silence for some of our fallen-from-grace brethren:
To its final zesting place: Going from a zesty $450 million valuation to shutting down, even Goldman Sachs’ backing couldn’t spice up ZestMoney’s survival.
So close: Edtech company Doubtnut learns the hard way that a bird in the hand is worth two in the bush, selling for $10 million after passing up a $150 million deal from Byju’s.
Now, not so fab: From unicorn to extinct: Prefab home builder Veev proves that soaring to billion-dollar status doesn’t guarantee a sturdy foundation.
Top reads on TechCrunch this week
That not enough for ya? Fine, here’s a collection of the most-loved, most-read articles from the past week:
Is it a bird? Is it a plane?: Anduril’s new fighter jet weapon, Roadrunner, lands with the grace of a Falcon 9.
Pour me another one: MIT spinoff Liquid AI thinks it’s time for a change in the AI game with their new “liquid neural network,” because who needs another GPT clone when you can have AI inspired by worm brains and run on a Raspberry Pi . . .
Yeah, but will it wear a beanie hat?: Ex-SpaceX engineers are now saving the planet with a “vegetarian rocket engine,” because apparently shooting stuff into space wasn’t cool enough. Also, were previous rockets full of bacon? I’m confused.
It’s electrifying: GM and Toyota, welcome to the Oops, We Missed the EV Bus club!
Breaking kneecaps, and YouTube records: Grand Theft Auto VI just stole MrBeast’s YouTube crown, racking up more views in a day than a money-giving philanthropist could dream of.