Imagination Technologies IP licensing and execution

Entropy,

I hope you read PMs more frequently now

 

Is it easier to get GPU engineers by posting job ads or by buying up graphics companies to get both the IP and the engineers?

 

Obviously the first. There are of course exceptions when ARM bought Falanx for an undisclosed sum at the time; considering Falanx was small and didn't have any sales success up to then I could remember and their NET assets were worth $315000 I don't think it was "expensive" for them and I don't think Falanx had all that many engineers back then either.

 

MUCH easier to just hire people.
Unless you specifically want the IP-portfolio, customers or distribution network of a company, buying a whole organization is an awkward proposition. And you don't have to post any job ads, you can just quietly head hunt. Apple building a gfx-group didn't get any attention in the press until they hired a large group of ex-ATI folks at once roughly a year ago.

 

Them buying Intrisity caused quite a wave amongst fruitco fans and the neverending notion back then that they'll develop their own processors from scratch was more than just tiresome. Thank God it's all over now.

In between we also had the "where will Apple manufacture next" sagas and how the ultra-strong Samsung would kick supposedly Apple's butt etc etc. I repeated over and over again that it ISN'T in both sides best interest for Apple not to continue to manufacture at Samsung and that story is not by any means directly connected with any legal dispute or patent nonsense the two are battling in courts.

I honestly often wonder why people just don't think of the most simple things especially when they're right in front of their eyes; obviously it attracts far more interest if you overdramatize things

 

Well that's why I said billion, double their market cap can make a lot of objections irrelevant.

 

That's not handling everything in house, that's just implementing IMG's macro.

If they don't want to depend on IMG they will either have to make their own architecture or fork an existing design. The former would take fucking ages, the latter would probably take fucking ages as well to create anything better than the original company if they try do it with new engineers.

I'm not saying they want to handle everything in house, I'm just saying that if they really do, just buying the IP and the engineers who know it inside and out at the same time (ie. the entire company) makes more sense.

 

Agreed, if you look at the market 3 years ago, there were a number of companies which seemed genuinely reliant on IMG (Intel + TI but also ST-E/Renesas/etc) but the situation has changed. Most of these companies have either exited the market or diversified their 3D IP supply base.

There would still be some obstacles for anyone wanting to buy IMG but I expect that they would be surmountable nowadays, especially at the right price. I don't consider it likely personally but I wouldn't dismiss it out of hand either.

 

For more yes; yes their stock value is pretty fucked up at the moment but that primarily IMHO because they have a lot of new projects IP products that aren't yet bringing any profit in. That should not mean that if they play their cards right they couldn't rebounce to the stage they were in the recent past where their market cap was approaching $2b.

Scratch that then because I thought you mean that it takes ages for IMG's GPU IP to get integrated.

It would be mighty weird if a semi signs a multi-year multi-license deal and even allow to be named in it, only to turn around and devour the company afterwards. I'm not aware of what Apple could actually pay in royalties but is it way more than say $40-50Mio on an annual basis these days? Hard to believe considering their annual royalty revenue. Invest a billion for that and on top of that to have to sustain N more engineers over the years?

 

IMG's stock price simply went back to where it belongs. IP is just a really shitty business to be in, ARM being about the only notable exception. It may be a bit painful to swap one GPU for another, but ultimately, GPUs are still very much replaceable, so there's not a whole lot of lock-in.

I don't know what kind of new IP you're hinting at, but it's hard to think of something that's more complex than a GPU. And if it's simpler, it's even harder to get some money out of it.

 

Well the first sentence with the last paragraph builds an entertaining oxymoron. If you don't know what the company is actually selling these days, then any conclusion as to where the stock's place is, is definitely justified.

Other than that ARM also decided to sell GPU IP quite some time ago and IMG is trying hard to attack a fraction for ARM's core business too with CPU IP. Whether they'll reach their goals there (which seem quite bold) is another question, but they are offering the majority of IP one could need for an entire SoC.

 

Instead of calling oxymoron, why not clarify exactly which kind of ground breaking IP will shake the world and their revenue with it?

Let me repeat: GPUs are about the most complex IPs one can think off and even there they don't manage to make a lot of money. Ray tracing is not exactly the kind of stuff people are waiting for and the only reason people have used MIPS lately is because of legacy reasons or because they were cheaper than the competition.

Enlighten me about the next Great Leap Forward?

 

How long does it take to open the company's page and brush through it?

Probably why they expanded in about every direction possible including CPU IP.

Both the Caustic and MIPS acquisition cost money and it'll take quite some time until those investments start to come back in. Likewise Raptor ISP IP and what else they're dealing with these days.

If they manage to get 25% of the processor market in the timeframe they've marked it's not something you can brush over that easily. Yes it's a huge risk and yes no one knows if they'll come as far, but there's more potential in IMG than just a pissat half a billion. Not when at the same time ARM is an "exception" and I assume it's over 20b market cap is justified and yes that is an oxymoron then.

Maybe if ARM would present a wee bit less uncompetitive practices IMG would have an easier time overall and yes let me hear that I don't know what I'm talking about. You want to get "enlightened"? Find a religion; prayers should help with exaggerated sarcasm

 

It seems it's a seller's market for experienced GPU engineers.

So there would be a bidding war situation and even if you won, the engineers you get might be encumbered by their knowledge of IP from their former companies?

That is you pay for their knowledge but their knowledge might include proprietary IP that you wouldn't legally be able to leverage?

 

Nope patents belong to the companies the patents have been granted for. You pay for their experience primarily.

 

Way too long! It's not as if it says on the front page "this is why we are currently down, but the following stuff is going to fix all that."

They are going to try to make money with high hanging fruit to compensate for the lack of revenues in the low hanging fruit?

(Note: I don't consider IMG a failure in any way. For an IP company, they are remarkably successful and they deserve respect. It's just that they are in a business that's fundamentally shitty.)

Again: ray tracing nobody cares.
MIPS is for bottom feeders.
I know next to nothing about ISPs: is it more than a bank of highly parallel ALUs and some histograms? Is it sufficiently specialized to add functions that makes one stand out and that's hard to come up with by others?
Whats the whatever else? "communications and cloud IP" ?

25% of the cheap bottom feeder market. Are there reasons other than cost to not choose an ARM processor?

When ARM is the only company in existence in this business that has this kind of market cap then, yes, that's an exception by definition. As for $20B being justified: no need to build straw mans. I never said that.

I have no idea what you're talking about. Not saying if it's true or not, I don't know. But if they do this, it's a scandal. A scandal, I tell you!

Hey, you're the one who seems to be a believer in IMG. I'm a sceptic as ever...

 

My expectations have been always lower then what they've achieved over the years and still are.

Yes the IP business is crapalicious, but in the ULP market it's the only thing that seems to make sense. If then Intel would also stand better chances than ARM.

As for ISP IP and what not I cant be bothered to brush through endless of pages of their PR stuff to find the links but they acquired a small firm for it, just as they acquired Caustic.

Ray tracing let's who'll be the first to license IP once it's available and we'll see the true potential of MIPS once the first significant licensing deal appears.

They are currently selling less GPUs than before due TI quitting the market amongst others, but they'll recover in about a year or so.

 

Everyone "cares", but there's a chicken-egg problem for us to solve first.

A tad short sighted, but if you're just talking about phone and tablet APs then I can see where you're coming from. We're going to try and build something compelling there.

It's a lot more of a problem to solve than just parallel math and some histograms, but you're right in that it's not necessarily anything approaching black magic. After all, most ISPs are in-house today. But that's slowly changing as they get a lot more complex and consume more bandwidth.

I haven't observed cost being the dominant reason for MIPS licensees to take the IP, but those reasons aren't something I can comment on publicly.

Sounds like we need an "IMG Execution" thread in the semiconductor industry forum. I'll create one.

 

Rys,

Shouldn't you leave a reference in the former thread that content has been moved here?

 

While I have no clue what IMG is planning, I hope the first IP wave at least will be standalone RTU IP. I can't imagine what I'd need as a user RT in a smartphone.