VCs think, 'Apps are risky, infrastructure is safe,' so they invested in AI infra.
"infra is safe" Hmm, but that wasn't a good idea. because if an open source infrastructure project like TensorZero gets shut down this quickly, won't they start to realize that those investment theories are also risky?
The difficult thing about AI infrastructure is that, unlike other industries, it will not become fragmented. It will likely remain tied to specific big tech models. What does this mean? It means that because AI models are not yet standardized, the infrastructure itself is actually riskier. In other words, the privatization of standards is happening.
The challenge with AI infrastructure is that an independent, stable standard layer has not formed, unlike in other software infrastructure markets such as databases, web servers, cloud, and containers. Over time, those ecosystems developed relatively standardized interfaces and operational layers. But the LLM ecosystem is still evolving rapidly. Models themselves change fast, APIs differ, pricing differs, context windows, tool calling, structured output, evaluation, fine tuning, caching, routing, everything keeps changing.
So even if an infrastructure startup tries to build a common abstraction layer across multiple models, before that common layer can stabilize, big model or cloud providers like OpenAI, Anthropic, Google, AWS, or Azure can just absorb the same functionality directly. In the end, AI infrastructure is at high risk of becoming an attached feature of model providers rather than solidifying as an independent layer.
But if a startup that raised 7.3 million dollars fails this quickly, who would trust and invest in such things? That aside, it seems AI startups are all the rage these days. I also want to learn AI and get funded like that. Does anyone here trust me enough to invest? About one hundredth of that would probably be enough
A few comments.
> VCs think, 'Apps are risky, infrastructure is safe,' so they invested in AI infra.
First off, this isn't even infra in the infra sense of the word. Infrastructure implied something physical, a pure software product can almost never be considered 'infra'. A tool maybe, but not 'infra'.
VCs can also be irrational and driven primarily by personal connections rather than reason. I didn't do a deep dive in this project/leadership, but often who you know is some important than what you produced. There's a reason why a lot of VCs go for the old motto of "I'd rather invest in an A team with a C product; than invest in a C team with an A product".
I also believe the same. Many VCs are obsessed with moat that they clearly got wrong. To me the value created at app layers are so much that gives them the flexibility to diversify their infra layers. Good harnessed do not depend on a specific model provider or memory layer or etc that when it is taken down like anthropic fable they get no risk exposure. Many even after growing train their own model like what cursor did with composer. There’s many more examples in other verticals like manus, superhuman, fireflies, lovable, replit, cursor, nouswise, cline windsurf and kilo but many are concentrated in coding because again I think VCs have preferred this definition of moat.
Due to the echo chamber effect, our opinions get reinforced, which can lead to biased conclusions, so it gives me pause. But your comment is so eerily similar to my own thoughts that I'm writing this reply.
I agree that most people misunderstand the concept of a 'moat' and become obsessed with that misunderstanding. People tend to think that only technical 'coding skills' which they can easily understand constitute a moat. But in reality, the moat is the entire workflow across the product's lifecycle, including coing skills. In that sense, infrastructure workflows are nothing more than 'the most easily replaceable consumables.' The essential purpose of infrastructure is to pursue 'standardization,' which paradoxically means a state of 'zero switching costs' where customers (app developers) can switch at any time to a better API or a big tech built in feature. Pure technology that doesn't latch onto the messy real world domains of customers will inevitably be absorbed without resistance by massive capital.
In some ways, customer lock in at the application layer, or even the fan culture around a product, creates emotional lock in. The end user app that provides a specific workflow integrated into users' daily routines can overcome even technical inferiority through 'experience' and 'emotion.' Technology can be copied, but the user identity attached to a tool is what I think a real moat is.(That is also the reason I love Windows.)
The example you gave, Cursor's Composer, is exactly the case I'm talking about. I think Cursor is inferior, and I don't think its Composer model feature is all that great either. But Cursor has a passionate fan base, and users who choose Composer as the best value for money no longer care about absolute technical performance or benchmark scores. They are captivated by the 'speed of experience' of code being completed quickly as they intended, and the 'frictionless workflow' the tool provides.it's not the company that builds the best AI model that wins, but the company that wraps 'good enough technology' in 'great UX' and dominates users' habits. That is how apps dominate infrastructure, and that's the moat you and I are thinking about.
That said, this conclusion is probably too hasty and has many flaws. Still, your thoughts are so similar to mine that I'm leaving this reply. Thanks for the great comment. Have a good day
Our investors aren't looking for safe, they're looking for a small chance in funding the next Databricks or similar. Most times it doesn't work out unfortunately, but that's part of the game.
(Also, we raised the capital in 2024 and didn't burn most of it.)
First of all, I respect your decision. I apologize for speaking too hastily about your choices. What I was trying to do was simply talk about how incredibly fast AI infrastructure changes. I also understand your respect for investors looking for the next Databricks. But the reason I wrote what I did is because the confidence expressed in the README ended so early. That said, such confidence isn't necessarily a bad thing. Isn't it said that victory belongs to challengers like you, not cynical people like me? I feel bad about that part. Still, I have no intention of withdrawing my skeptical view about whether AI infrastructure can succeed as an open source startup. I'm very sorry it came across as if I was mocking your failure. That was not my intent. I was simply trying to leave a comment saying that AI infrastructure has a different direction from traditional infrastructure.
Thanks, appreciate the follow-up. It's certainly still to be determined if OSS AI infra will pan out, but I hope it does!
I wish you success. I wasn't trying to mock your failure, and I'm truly sorry if I made you feel bad. I don't want to be a cynical person who mocks others' challenges. I apologize again for that. I was just expressing that something doesn't seem right from my current perspective. I hope your next challenge goes well.
I mean it. I'm sorry once again
> VCs think, 'Apps are risky, infrastructure is safe,' so they invested in AI infra.
I think you're really overgeneralizing what "infrastructure" means in this case.
Infra is perhaps somewhat safe but realistically it's a really low margin capital intense business long-term unless you can lock-in customers with hundreds of services like AWS. So not a lot of space for a huge ROI.
> are all the rage these days
Are they? Overall it seems kind of tame compared to 2020-21 since VCs are somewhat risk average outside of a few outliers. Funding looks much more concentrated these days.
You're right. Looking at recent indicators, there are more stable investments than I thought. But please understand that, as a human, I haven't achieved ROI in terms of marriage, relationships, a stable job, etc., so my perspective might be mixed with a bit of envy
Tell me you haven't talked to a VC.
A better model for VCs is: companies are finding tons of budget to allocate to new AI spend. Besides the labs, who is going to be able to capture some of that spend while they're actively looking to spend it?
Nobody at the seed stage is investing in things they think are "safe". They are investing in things they think have huge upside.
Sometimes people don't realize that 'professional' ideals and 'reality' are different.
What you're talking about seems like 'ideal' investing, not real world investing at all. Of course, the VCs in your country and the VCs in my country are different.
It's like in software, where everyone says you should write maintainable code within the norms, but in reality, most people don't do that
that investing in 'potential' is the basic principle of VCs. They call it the power law. But when you look at actual investment portfolios, it seems quite rare for people to follow only that principle. I guess you don't think so. Of course, I agree that ideal venture investing follows the power law. But in real world investing, there are pragmatic investors who operate somewhere between the ideal and reality. We always project ourselves onto the 'ideal,' but I don't think there are only people who are immersed in that ideal. Of course, no VC would invest in someone like me. I've met with VCs three times in my career, but they all turned me down. Haha.
Anyway, I wasn't trying to mock your profession. Here's what I think. Most VCs and investors have their own success formula. There will be VCs who succeeded by investing in infrastructure. But the question is whether that same success formula applies to AI startups right now. Of course, from your perspective, it might look like 'this clueless kid is just being cynical without knowing anything.' I partly agree. But that's not the core of my argument.
What I'm trying to say is that those success formulas themselves need to be reconsidered.An insider from up there came out and talked about the next 'Databricks,' believing that's the kind of potential they're looking for. All of them do. Everyone wants to be the first investor in a goldmine. I don't think this is just about greed
The question is whether the traditional infrastructure investment logic holds here. I think most current AI infrastructure tools are closer to 'temporary patches' that exist before the functionality gets internalized.
Let's say infrastructure is like a concrete building. Traditional IT infrastructure basically has a standards committee, and once that committee sets things, changes are extremely rare. It's a kind of 'lake.' But AI infrastructure right now is different from one to another; even the ecosystems differ—the Chinese ecosystem is different from the US ecosystem. It's a flowing 'river.' I just think the question is whether the old grammar can be applied in this situation.
You probably have more money, more investment experience, and more success than I do. I only have a lot of failure. But apart from that, the issue is simply that 'potential' in growth potential ends up being data measured against past examples, and the question is whether that data still holds up now. Anyway, I might have been slightly sarcastic earlier, so I apologize for that. Someone as successful as you, please bear with it a little.