Back in 2007, Havok physics engine was sold to Intel by their developers for $75 M. Are physics engine very "hard" work or fairly simple if one has Phd in Physics and knows a couple of graphics api.
Since most of us will not be able to compete with Havok engine but what best can we do with knowledge from both physics and computer science? 
Also, Intel literally did nothing with Havok as far I know, they sold it to MS for undisclosed amount. May be they wanted to look the code and found it semi to less useful.

Do large companies have this trend of buying anything they see getting heat in market irrespective of analyzing why they would need it for?

IIRC Intel made Havok free to use for commercial uses and this had something to do with CPU vs. GPU physics debate or with trying to undermine NVidia's PhysX back in the day. It was big news when this happened. Microsoft has apparently ended this?

There's a few FOSS physics engines like Bullet and Newton Game Dynamics.

Bullet is kinda more famous of these two since it's in a few AAA games and many others engines and such like Blender, GTA 4 and 5.

Newton Game Dynamics is used by Frictional in their own HPL Engine (for Penumbra, SOMA, Amnesia). First Penumbra game caused some hype due to how they used physics for doors, drawers, combat, etc. (it's a horror game so it made sense to let you slowly open the door partly or make you scramble to rummage through drawers).

PhysX got open sourced for non-console platforms very recently too (proper BDS license, the 3 clause one, which has huge implication, no backsies or EULAs).

I've no idea how hard it is to implement a physics engine personally but it's clearly not that easy and depends on its features like does it have fluids, soft bodies, cloth, etc. anyway. And it doesn't have to be on the GPU necessarily.

And we're spoiled for choice these days, there is no reason for someone (who isn't a huge company with a good reason) to write own physics or do it from scratch, there's no catch with a FOSS library. I'm not sure if Square Enix's Crystal Tools engine has its own custom physics and that engine as a whole had very mixed results.

46 minutes ago, ritzmax72 said:

Do large companies have this trend of buying anything they see getting heat in market irrespective of analyzing why they would need it for? 

Ahh.. It happens sometimes but more often the acquisitions make some sense in other way like Oracle buying Sun or Ksplice or Atlassian buying Trello, IBM buying Red Hat, Red Hat buying JBoss, etc. I can think of tons of ones like these but not of many that are like you said (Intel kind of wasn't IMO for reason I said above). Maybe Facebook buying Occulus or Microsoft buying Mojang would fit these "get something just because it's hot".

Thanks Frex, that was good insight into tech trading. But still, why did intel buy for $76 without doing anything with it and sold it to MS. They are releasing their GPU in some years, may be they gave lessons to their R&D. 

It might have something involving executives. Intel behaviour in the last few years has not been smart as they used to. Their idea was, as previously noted, to counter the advance of GPU solvers. Those basically never surfaced so there was nothing to do - it probably involves users insisting on buying graphics-less CPUs.

GPU design from a piece of software? I don't think so.

On 1/14/2019 at 5:28 AM, FRex said:

And we're spoiled for choice these days, there is no reason for someone (who isn't a huge company with a good reason) to write own physics or do it from scratch, there's no catch with a FOSS library. I'm not sure if Square Enix's Crystal Tools engine has its own custom physics and that engine as a whole had very mixed results.

There may be few reasons to actually develop your own, one of them is accuracy or specific computation required - this is something you won't find in games though, but in simulation software it is actually quite a big deal.

It is also more than possible, that it simulates parts that are not simulated by any of the mentioned libraries (things like particle simulations and interactions).

Another reason is integrating specific items like fluid dynamics computed with GPU, which isn't really supported widely among the physics engines. Implementing interaction between this and another physics engine may actually be harder than one thinks.

And another that comes to my mind (which pushed me once into rigid body simulation) is for learning reasons. I did it once to learn how physics in 3D works, but didn't built on top of it (it is almost years back though). Since then I've used PhysX, ODE and Bullet.