Executives banged-on about keeping internal fab utilization rates high through up & down cycles but used foundries as a major buffer. TSMC etc... would see orders cut to suit the business environment while TI kept its fabs humming as much as possible.

This not only kept margins high but meant TI didn't have to ramp its 300mm fab or spend a few billion dollars on equipment. That little trick also meant it had lower depreciation costs. To say the company ran a tight ship, even in the up-cycles, is a clear understatement.

TI exec's liked it, the financial community liked it and there seemed no way back from there!

From ‘there' I mean not going back to heavy fab investment and going further in becoming a ‘fab-lite' business. A classic example of this strategy can be seen with TI's 300mm fab DMOS6. It was one of the first 300mm fabs to come on stream in 1998 yet has recorded the slowest ramp of any 300mm fab since & it still hasn't reach capacity!

With the closure of KFAB, TI tried to put the spin on this fab being old and therefore something industry observers should treat as a normal in the usual life-cycle of fabs. What TI didn't bother to point out was that KFAB was also responsible for process development work through the 90nm node, rather than DMOS6. It may have been old in years but far from old in process technology!

What I didn't expect and I am pretty sure the financial analysts didn't expect either was the news that TI would STOP logic process development altogether. Though keeping hold of analog development as it really doesn't cost anything, TI is going to rely on its foundries for its process technology roadmap.

Again, TI tried to spin on this, claiming closer collaborations on next gen stuff but I doubt that involves TI giving money to the foundries. NXP is leaving the Crolles2 Alliance in part because its realignment with IMEC and TSMC means only a fraction of the money is required to feed IMEC, while NXP will not spend a penny with TSMC for R&D.

This point makes me think that TI may go to an R&D alliance for certain things in the future, though this was not mentioned at all by TI. Saying that, and dependant on the fab/foundry process liaison team size TI has, they may well become totally ‘fabless' in that respect of process development. It will be interesting to see how that pans out!
Something else TI exec's didn't speak about was the newly completed RFAB in Richardson, Texas. This 300mm fab is currently in mothballs awaiting the nod to tool install for the 45nm devices coming down the line.

Is this fab necessary now?

Perhaps it still is required considering that TI plans to keep current fabs loaded, at least for the time being. With KFAB tools going to analog fabs and KFAB production going to DMOS6 (we assume), the capacity at DMOS6 may get pushed closer to capacity this year than previously planned. If so, then we could actually see RFAB opened up this year and a small line installed and operated for TI's 45nm devices. Remember, please that TI has spent money on developing this process so I can't see them fritting those kind of dollars away without getting some return.

However, TI is making some serious changes and that could easily include DMOS6 being their first and last 300mm fab.

With killing process development, TI is losing an edge over its major fabless competitors. It may have a tech-node lead of about 1.5 nodes but I doubt that will last very long. TI exec's basically admitted this to an analysts' question, which also shows how serious TI is in cutting out process development internally.

Though exec's mentioned that the competitive space is shifting to IC design and so gives TI more ability to focus on this, I doubt the savings will go anywhere near IC design teams. Indeed, TI may start to find the financial analysts a little bit more aggressive in Q&A sessions in the future as they will see higher competitive risk from fabless rivals. TI has been in a luxury position when it comes to having its own process development operations compared to a Qualcomm for example. With that little luxury gone TI will be re-evaluated sooner rather than later!

TI knows that analog fabs are cash-cows and so it would be incredibly unlikely that it will become completely fabless-ever. But it has moved closer to the edge and an edge that others are finding it increasingly difficult not to go over!