2025-05-07 02:54:33
"does my new startup idea need to have AI in it?"
lots of people asking this rn -- and I'm talking to people about it, via our a16z speedrun program where we invest in brand new startups. The majorty are AI startups so we've been contemplating this question quite a bit (PS. deadline for the next program, starting H2 2025, is this coming Sunday: http://sr.a16z.com)
ok so some thoughts on this question:
- AI as the easy answer to the "why now?"
- "fake" AI startups
- investor novelty vs customer novelty
- the golden age of AI
- why AI consumer apps aren't really a thing yet
- how hiring great talent plays a role
let's start:
- so do you need AI in your new startup idea? probably? maybe? The issue is that we've had over 15 years of mobile apps and a lot of ideas have been tried. So naturally, everybody wants to know: why this idea *now*? Is there something new or special that's changed that makes this idea now possible compared to previous years, assuming that those ideas didn't work out? Now, AI is just a really easy answer because it's a horizontal technology that can apply to every single category of product. Certain categories are being claimed left and right now, particularly in areas like creative tools and productivity, but the rest of the market is actually quite open. So, if your answer to this isn't AI, then you still need something really compelling and interesting to say. Maybe consumer distrust in mainstream media fuels a lot of independent voices, and those people can then create a living that way, like Substack. But you definitely need *something*. And oftentimes that something is AI.
The issue is that we've had over 15 years of mobile apps, and a lot of ideas have been tried. So naturally, everybody wants to know, why this idea now? You definitely need something, and oftentimes that something is AI.
- that said, the funny thing is there are now a lot of "fake AI companies." These are companies that might use AI in marginal ways but claim to be AI-first just to raise money. I see this all the time, along with my investor colleagues. These are companies that are maybe five years old, and all of a sudden they're claiming that they use AI to automate some backend process. This doesn't actually change what the customer sees, but might add some small amount of efficiency. I just don't think of this as a strong AI company because it doesn't actually affect the functionality in a way that gives it an edge. Similarly, people will build customer-facing features, but then the AI feature is some marginal part. These are all kind of not full-on rethinkings of AI. You might ask, well, why are people doing this? The reason is because when you see that there are a lot of AI startups raising money quickly, you want to throw your hat in the ring. Just know that it's easy to see through these. Don't be a fake AI startup.
- And by the way, this isn't just from an investor perspective, but also from a customer one. Customers often go through cycles where, when there's a new technology, they want to try everything. That's why there's so much novelty effect in AI. They see cool products or outputs on social media and want to try them out. In previous years, it was seeing really cool mobile apps that people were installing. Now, this moment in time eventually goes away, but for now, it's a complete growth hack. As a result, the "why now" becomes important for exactly that reason. For investors, of course, they have a similar question, but they're thinking about it more from a market disruption standpoint, and whether you understand enough of the idea to navigate it properly. And they are thinking about other downstream investors, and so on. Thus, the bar for non-AI companies goes up quite a lot -- you better have traction!
- The good news is that we're in the golden age of customer-facing AI startups. For the last two years, every startup was building fundamental tech, often at the model layer. So, the ideal founders were highly technical academic researchers with PhDs. Now that open source LLMs are great, startups are applying this tech to vertical categories like legal, accounting, creative tools, real estate, finance, etc etc etc. Every vertical. Customers just care that it works, and not as much about the underlying tech. They're more interested in security, privacy, workflow integration, and API integrations than the state-of-the-art model. That's great for startups with deep domain expertise, rather than just AI tech. This of course opens the market to the broader cohort of mobile/web founders who can build new UX and businesses that leverage AI in a big way
- However, it's also true that current widespread AI applications have limitations. For productivity use cases, we're seeing a sweet spot where product-led growth is fueled by social media. There's also the ability to expand revenue within a company as a note-taking app or creative tool spreads within a business. As much as I'd love to see mainstream, new takes on gaming or consumer social networks and photo-sharing applications, we haven't seen new native AI versions of these ideas yet. The reason is that the unit economics are getting close, but aren't quite there. If you open a new consumer app and have to fire off dozens of API calls, each costing a bit of money to fulfill the experience, it just doesn't seem to work well in an ad-supported or freemium environment. You really need to get your user costs down to a few dollars a year. When you do the math, using one of the major LLM providers might add up to tens or hundreds of dollars. I know we'll get there, but it's going to take some time. Limitations like this keep AI in high-ARPU environments like productivity/B2B
- Another important angle is hiring. Because of the hype, a lot of smart people want to quit their cushy jobs where they can't do cutting-edge work and move to a new startup where they'll be exposed to new technologies like AI. If you're in a different field like deep tech, maybe this isn't an issue. But in the software world, you naturally want to appeal to those prospective employees. Focusing on AI lets you recruit better people for the same dollars, which is a huge incentive.
To sum it up: if you're building something new, you need a lot of energy to break through and inspire investors, customers, and employees. AI is an easy catch-all to do that right now. If you have something else, great—you might do even better. But if not, the AI angle is too hard to ignore. Of course, this novelty will pass. The mobile platform led to a generation of "mobile-first" startups, which was exciting for a while. Eventually, everything was assumed to be mobile-first, and it was no longer a thing. The same cycle will happen with AI, but we're early enough that it still adds value.
2025-05-07 01:38:33
photos -- what you see when you walk past the a16z SF office rn
fun. quick reminder that the a16z speedrun deadline is this sunday!
2025-05-02 23:11:00
RT a16z speedrun 🧊
best part of my job is running videos like this for approval by compliance
2025-04-30 23:59:03
RT a16z
Join us at #NYTechWeek, featuring events with:
@dhaber
@andrewchen
The a16z podcast
@a16ztxo
@speedrun
And more, from June 2 to 8.
Register now: https://www.tech-week.com/calendar
And follow @Techweek_ for more updates.
2025-04-30 23:41:19
RT Jon Lai
[New] a16z @speedrun request for startups:
Generative MTV 🎶
There’s a Spotify sized opportunity today to build a new generative MTV that can become the defining music platform of the AI generation
At the height of its popularity, MTV was one of the most-watched TV stations reaching >99 million US households. The channel turbocharged the cultural ascendance of stars like Madonna, Prince, & more
AI music models like @udiomusic @SunoMusic have unlocked new ways of creating, discovering, and consuming music. For example, imagine a new music app that can:
- generate an always-on soundtrack that shapes itself to your daily life using context like your calendar, location, and biometrics. For ex. switching from a commute playlist to a work-out playlist on the fly
- generate music video synced to a playlist or listener feedback in real-time, with understanding of the lyrics and musicality of each song. Done well, it would be a multisensory, personalized concert
Some might say that a major music platform or label will build this, but there is a time right now where it could be a new start-up instead. If you’re excited about this space, we’d love to jam!
2025-04-30 01:53:13
a16z is hosting a lot of events at ny tech week in June!
