Hey Thom –
Agreed – happy to keep following this one.
First update, I asked a very good friend who is probably one of the best and most thoughtful engineers that I know, and he had this to say (we didn’t exchange any thoughts at all before he wrote this, so it’s interesting to see he came to the same “garbage in, garbage out” analogy that I did):
“CAE is a garbage in, garbage out system – which is heavily reliant on engineering talent to get the context correct. You can layer on the complexities of working with many different solvers and software packages. So, the value proposition for an “AI ” system is either being totally generalized OR a specialized system for a certain type of analysis. The former doesn’t strike me as likely anytime soon. The latter is plausible.
Then I think to the market demand- why would companies not be using HPC? High upfront cost and lack of expertise is an obvious one. Lack of available resources is another – even with clear value proposition, finding resources (hires) can be tough. There could also be a place for doing repeated analysis of similar systems (think n variations of airfoil designs). Again, big companies will have the resources to do this themselves, but small to mid may not. Market size is the key question here. It will be the difference between a lifestyle company (consultancy) or a 10x+ growth to acquisition.”
Second update – the founder responded to my questions on StartEngine here. One thing Allan said that raised a few more questions about how viable this will be to sell when targeting SMBs is “2) It’s like Amazon Prime – where you purchase a Prime subscription, and then Amazon helps you buy specific products. (You purchase a GURU subscription, and then GURU orchestrates your payments to the third party ISV & cloud vendors for that job) see “GURU Orchestrates Payments”.
Thus, if I understand the unit economics and pricing correct, it sounds like you will pay the $500 per simulation or $20k per year subscription to Guru, then have to pay the normal cloud compute and HPC fees on top of that. I was thinking that they were somehow going to perhaps leverage being able to purchase compute power in bulk and then allocate it among their customers so that they could pass along savings.
This is a major potential obstacle in my opinion, as it doesn’t really solve the issue of cost of running these analyses (which can be very steep, especially for a small business); it only seems to solve the issue of resources and expertise. And if that’s the case, I wonder more if they’ll be able to get enough initial customers and sales and if they are addressing the real paint point properly.
Let’s keep an eye on this one.