Why did licensing deal with Chartered and IBM fail?

[yubjun]

In light of what's happening with Samsung and GF, I thought we could take a step back to see why similar licensing model between IBM and Chartered really did not work out. Does anyone have details on why Chartered kept licensing from IBM since 2002, yet, they kept losing money, eventually selling themselves to GF?


My thought is that Chartered didn't really have a choice since they had no R&D capability, and IBM was the only company who would license them advanced tech-nodes. And perhaps the licensing cost was too high, or they were not able to ramp-up in time. If so, i wonder why they did not develop R&D to do it on their own.


I'm curious to find out what happened with the licensing model between IBM and its customers. Does anyone have some inside view on this?

Thanks

T

 

[Idontcare]

Chartered's problem was simply one of scale (wafer volume) versus their overhead (including licensing costs) relative to their competition, target customer, and the fact they had no R&D to differentiate their product offerings. Competing on cost alone only works if you have the lowest cost structure in the business.

At the time, Chartered faced competition from the low-cost foundry SMIC in the process nodes that mattered to the customers that Chartered was targeting.

Its not IBM's licensing model itself that is/was the problem. Rather its existence is a symptom of a much larger, and more problematic issue.

Namely that no company that is in a position of strength has willingly joined the IBM fab club. It is a method of last resort on the path towards irrelevance.

When you find yourself, as decision makers of a major IDM or foundry, throwing in the towel on your own internal development team and joining up with an externally controlled consortia, you are throwing in the towel but just not quite acknowledging it fully at the time.

The IBM consortia exists as a collection of second-place holders across the industry, an area where the losers retreat to lick their wounds and hope to live to fight another day. Meanwhile the winners (Intel, TSMC, Samsung) continue to just get stronger.

And when you have a consortia of second-in-class businesses teaming up together to develop tomorrow's solutions by committee, to combat the same opponents who themselves are best in class because of the very internal R&D engines that they continue to fund and grow in the meantime, it sets the stage for exactly what we have been witnessing these past 5-7 yrs.

A slow demise of any business that owns a fab which doesn't maintain control of their own destiny by maintaining control of their own R&D.

That was Chartered's downfall. And it is a story that will be repeated in time.

 

[Phynaz]

IDC, I am SO glad you are back!

 

[yubjun]

IDK, awesome thanks.

So basically you are saying that IBM themselves had second class process technology, and therefore, whoever licenses it will also only get second class process technology. And this is why IBM - Chartered licensing model did not work as they could not compete against best in class TSMC.

Do you think that Samsung - GF model could potentially work better than the IBM - Chartered? I would think that even if GF licenses best in class (let's assume for a second that Samsung is first in class, although it is probably not at the TSMC league), they still have issue of having to continue to license it, (higher cost), and always being held hostage by the licensor. But besides this issue, I wonder if GF really can provide the same tech-node to customers as the articles have been writing.

 

[Idontcare]

Do you think that Samsung - GF model could potentially work better than the IBM - Chartered?

There are a handful of major differences here with Samsung-GF which work in their favor versus that of the IBM-Chartered situation.

One is that Samsung operates and maintains a successful foundry business in a way that IBM never achieved even at its peak in the foundry space. IBM created the fab consortia only when it became painfully (economically) obvious internally that their foundry efforts were failing and they needed a "Plan C" to run with. (Plan B was to open up their fabs for foundry work in the first place)

The second major difference is that GF has the kind of fab capacity and customer counts needed in order to drive wafer volume to levels that can better handle the licensing overhead in a way that Chartered could have only hoped for.

By far the biggest handicap that GF was handed in its creation was the fact that AMD dismantled their internal R&D team one node too soon and joined the fab club as a transitional step to becoming fabless.

Had AMD maintained their internal design team and spun off GF with that inhouse capability then the GF that exists today would have had a very different competitiveness.

Its a loss that frankly I don't see how GF can recover from. No one in the history of the semiconductor industry has successfully rebuilt a process node development team after having divested themselves of such internal capability.

To rebuild that core competency, and then set that group (which would be starting from scratch in terms of knowing how to successfuly develop advanced process nodes) on a 4yr project to come up with a 7nm node (best case timeline right now for GF) simply defies all reasonable bounds of logic and comprehension.

So what does GF do from here? They are stuck between a very hard rock and a very hard place, the only way out is to buy someone else's pre-existing R&D team (preferably one that has a history of success in the market and not one that is being sold because it has a history of not being successful in the market) or being bought by someone who is interested in their fab capacity and customer lists (very much the reason Chartered was targeted and acquired to make this topic come back around on itself ).

 

[yubjun]

So what does GF do from here? They are stuck between a very hard rock and a very hard place, the only way out is to buy someone else's pre-existing R&D team (preferably one that has a history of success in the market and not one that is being sold because it has a history of not being successful in the market) or being bought by someone who is interested in their fab capacity and customer lists (very much the reason Chartered was targeted and acquired to make this topic come back around on itself ).

I see. What do you think about this high-tech R&D center . Do you see it being much of a game-changer, when it gets compelted?

 

[Phynaz]

Is the R&D center even being built? Fab 8 is two years behind schedule.

 

[Idontcare]

I see. What do you think about this high-tech R&D center . Do you see it being much of a game-changer, when it gets compelted?

I fully expect them to do their best to rebuild an internal process node development team. They need to and they have been making all the motions to suggest that is their long-term game plan.

It is a challenge though, finding all the right engineers and managers necessary to re-create a fully functioning node development team...only to then wait 4+ yrs before the first fruits of their labor could be expected to come to the production environment.