Flow Engineering raises $50M at $750M valuation to bring AI-driven continuous verification to hardware design

Speaker 2: Stop making excuses. We Up next, we have the founder and CEO of Flow Engineering, Pari Singh, coming back on the show. It's been almost a full year, but we gotta get that gong ready. How are doing? Welcome back.

Speaker 16: Hey. Hey. Good to see you guys. How are you?

Speaker 2: We're doing great, but it seems like you're doing even better. What's the news? What's going on? You raised some money?

Speaker 5: How much did you raise?

Speaker 16: We raised 50,000,000 on a $7.50 valuation.

Speaker 1: Congratulations. John was feeling that one.

Speaker 16: Thank you very much.

Speaker 2: And for what purpose did you raise this money?

Speaker 16: We are gonna invest heavily in AI. Okay.

Speaker 1: We are doing a huge

Speaker 16: amount with AI. AI is completely transforming hardware engineering Mhmm. Just in the way that it's transformed software engineering in the last year. I think when we see design cycles of rockets and airplanes and cars go down from twelve months to twelve weeks to twelve hours, the world's moving very, very, very that is the opportunity ahead of us, and that is what we are investing in.

Speaker 2: Do you like the term AI? Or do you think the next cycle you'll be saying superintelligence? Is that a meaningful designation to you? Or or do you see because I mean, this is a tool, this is a feature I'll

Speaker 1: go first. I don't think it's gonna stick. I think it's don't think we needed a new Question

Speaker 2: for Pan?

Speaker 1: Make going from artificial to super is not suddenly gonna, like, help the the the technology's popularity.

Speaker 2: Yeah. Okay.

Speaker 16: Do think there is a distinction. I'm not sure, like, everyone's agreed on the distinction. But to me, as you say, AI is like a really fundamental tool. Yeah. And when you put a tool in creators and inventors hands, they're gonna invent crazy stuff. Yeah. I do think we're gonna get to the point where AIs and humans collaborate together to design a new class of products. Mhmm. And that those products humans couldn't have done by themselves. Yeah. And when you put AIs and humans together, that's a really, really, really special place to design a whole new wave of hardware.

Speaker 2: Yeah. I think that I mean, putting aside the AI versus super intelligence debate, it's hard to gauge model progress from like deep research reports because those were pretty good. A year ago, they're still really good. But this year, the thing that's felt like it's jumped forward the most has been things in the three d world, the blender modeling demos, the video game development. And that feels like a glimpse into, okay, if it's that good at making video games and modeling architecture, it's probably good for hardware engineering. So how much of this year have your progress has been, okay, the models are advancing and most importantly, aside raw IQ, are they super intelligent, just they got good at hardware this year or they or there was a step function. Or did you feel that this year has that been an important unlock? Or has it felt smooth from the inside of the industry? Because from from my perspective, it was like I would never try to use an LLM with Blender a year ago and now Yeah. It would be my first step to work with Blender.

Speaker 16: So so AI is getting really, really good at hardware engineering. Yeah. When you look at modern products, whether it's a rocket or an airplane or a car, CAD and three d design is one really important element. Mhmm.

Speaker 5: But it's actually probably

Speaker 16: the last element. By the time you are designing the three d geometry of your Wemo, for example

Speaker 1: Mhmm.

Speaker 16: All of the mechanical and the electrical and the sensors and the AI and the software, there's tens of thousands of hours of engineering work that's gone into the product before you get to geometry.

Speaker 2: Yeah.

Speaker 16: And I think we're already seeing AI chip away

Speaker 17: at that.

Speaker 16: I think the world that we've seen in AI software development is as the models get better and better, that they can start to do more and more things. In hardware development, the products that we design are so complex that it requires mechanical engineers, electrical systems, software, regulatory test. It requires all these different types of engineers to come together to collaborate on a system. And on every single one of those, AI has been getting meaningfully better over the last year.

Speaker 2: So at what level of abstraction do you want to interface with a company like Rivian or Joby or Skydio or Stokes Space? Because I imagine within a carmaker, there's the the wire harnessing and the mechanical functionality and all sorts of different systems that need to eventually be integrated. And I could imagine a system overseeing that. And any time a change happens to the electrical system, you can flow that through the rest of of the project to say, oh, well, this is gonna affect this system over here. Flag that. Do do you wanna be the overseer, the sub agent, both? What what what's most valuable right now to modern enterprises?

Speaker 16: I I think you called it. So the single biggest problem in hardware companies is systems integration and verification. Yeah. Let me give you an example. Let's say you're designing a reasonable rocket Mhmm. And you make a tiny change. Let's say you take the stage one, and in stage one, you take one of the engines

Speaker 1: Mhmm.

Speaker 16: And you make a tiny change to the injector. Well, that has this chain reaction of changes across the fluid flow and the combustion and the thrust and the mission profile. Each one of those will change other things, which then change other things. And it's that third, fourth, fifth order impacts is where we trip up, and and that's where failures come from. Mhmm. So the single biggest problem in hardware development is actually integrating systems together. And because it's been so expensive and so painful to do that, the systems engineering organization could only really afford to do it every six, every nine, every twelve months. It's a really, really expensive process to to do full on verification. But what AI enables for us is this idea that we can move from manual verification, which is once every year, to continuous verification, just like we have in software engineering with CICD. And when you bring that paradigm over, engineers can make changes in software, in CAD, in Git, in SIM on a nearly hourly basis, and the AI can understand the ripple change of that impact and flag issues the moment they happen.

Speaker 2: Yeah. I want this for architecture. I want an architect to be able to move a wall in an architectural drawing and an agent goes and reads the laws and see if this needs permitting. And like actually merging the law which is written in text files with CAD which exists in the three d space like that feels like glue for AI such that you have sort of like, you know, red, yellow, green stoplight as you're building or as you're designing the structure.

Speaker 15: And that

Speaker 2: feels like a lot what you're doing. What about

Speaker 1: How how do you think the the world is gonna need to adapt and supply chains are gonna need to adapt to, to to the actual design process and hardware engineering process speeding up by, you know, 10 x and then and then a 100 x, which feels feels very inevitable at this point. But but oftentimes, like, the during the during the actual process of designing the system, you're also figuring out how to, you know, source all the all the different sort of inputs to the to the final product.

Speaker 16: Yeah. So so this is, a really, really important point. Hardware companies are vertically integrated. They're more vertically integrated than they've ever been before, but they still need to partner with suppliers at a very deep level. So even the most vertically integrated automotive companies have hundreds, if not thousands of suppliers, and the requirements for what they need is changing on a nearly weekly, if not daily basis. And that core collaboration is happening not just across the company, but across the partner too. For a number of our largest customers, customers designing everything from data centers to EVs

Speaker 2: Mhmm.

Speaker 16: To drones in the defense industry, they're actually working natively with their partners on flow. So that core collaboration changes that the partners will make will flow into their systems, and then the AI agent will understand the implications of that change in real time. And what used to take weeks or months before can now be done in seconds. And and the thing that John mentioned is so important. Like, architecture is great because it's really obvious what the regulatory requirements are. But if you look at any one of our customers, customers like Andor, Joby, Stokespace, Rivian, there's hundreds or thousands of regulatory requirements that need to be kept up to date across mechanical electrical software, and it's a really huge problem to do that today manually.

Speaker 2: Yeah. What's the state of integrating a systems integration integration? Is it faster? Or or do you have to rip and replace some other system? Or can you sort of drop in, drop on top of all the systems? Is computer use speeding that up? Are there MCP servers for some of this stuff? Because it feels like you're very much operating in the real world, and you can go and reverse an API from some old piece of software. But what what does it actually take to onboard a new customer these days?

Speaker 16: Yeah. So amazing news for Flow and and probably what drove the raise. Today, 96% of our customers come to us inbound. Wow. And they're able to get up and running within thirty days. Wow. The legacy players are looking for, like, a nine to twelve month period of about thirty days. And AI is absolutely driving that adoption in itself. Where we typically land is on something called requirements management

Speaker 2: Sure.

Speaker 16: Which is this sort of typically old school stodgy tool base Yeah. Which houses all of the requirements and specifications. Mhmm. And the reason that is so important is it's it's the bit of information that all the design data in the company will flow back to. So if you're making changes in CAD or Git or simulation, they they come back to verify the requirements because that is your CICD process. So we are the system of record for requirements and verification Mhmm. In these hardware companies. That is that is our wedge. But what we are becoming is this much more important core collaboration layer Mhmm. This context graph for the AIs and the humans to be able to work together in in a seamless way.

Speaker 2: Well, congratulations on the progress. Congratulations on the new round. Thank you

Speaker 1: getting a new snapshot of all the progress. Yeah. It's great.

Speaker 2: Fantastic progress.

Speaker 1: Congrats

Speaker 16: Great to meet you guys. Thank you so much.