Solid results from CMG this morning.
https://www.cmgl.ca/sites/default/files/sites/default/files/uploads/docs/CMG_Q4_19_Results_Press_Release.pdf
Also, here's an article from last month about how CMG is using AWS to move to the cloud.
https://aws.amazon.com/blogs/apn/solving-technological-limitations-of-complex-reservoir-simulation-with-cmg-and-aws/
I've been out of the simulation game for a few years now, but if that AWS article is accurate, CMG has had a significant breakthrough. Historically parallelization has worked for simulation, but only to a certain extent. When I ran simulations, my company had a group of virtual machines I could use for parallel processing of complex simulations. However, if you run with too many processors it hurts the accuracy of your results. I don't claim to understand this exactly, but it has to do with the math. By splitting the process of solving a very big matrix into too many processes, the rounding inherent in any computer solver affects the results. That limits the effectiveness of using many cores at the same time.
However, if they've come up with a more elegant/AI based method for dividing up the solver between multiple cores, then massively parallel computing becomes an option. With the much lower cost of renting processing from cloud providers, that could be a game changer for the cost/effectiveness of simulation.
I realize I'm doing a poor job explaining this, but I'm very excited about it. I haven't been able to find anyone in my network who has used this yet. Unfortunately companies aren't exactly spending money on upgrades right now, but I think this has the potential to be very material.
Their historical pricing has been based on the number of licenses, with each processor requiring one license. My work was almost never restricted by the number of computers available, but we had an internal process to prioritize how many licenses one engineer could use depending on the priority of the project. This will increase internal demand for licenses among engineers, and if the cost of computing falls and productivity/value goes up, I could see companies significantly increasing the number of licenses purchased per simulation/reservoir engineer.