Here's a bit of a turnaround. Last year all the noise was that the foundries would soon be holding the industry to ransom as the IDMs became increasingly dependent on them.
Now it is expected that the foundries, except for TSMC, will move into loss this quarter, and TSMC, while remaining profitable, will see a 30%+ drop in sales.
So gloomy are the foundries' prospects that the 60 per cent majority shareholder in Chartered, Temasek, is rumoured to be looking for a buyer for its stake.
How different things were last year. The remorseless moves to 'asset-lite' or 'fab-lite' strategies by the IDMs were said to be making the entire semiconductor industry dependent on the foundries.
Repeatedly asked if they were using their newly strengthened whip hand to increase wafer prices, TSMC eventually conceded that they would try to avoid the usual price erosion schedule over a process node by keeping prices up for longer than usual.
Now, it seems, there'll be none of that. The foundries will be scrabbling for orders to fill their fabs and the IDMs will have the whip hand on pricing.
How quickly things change.
TOMORROW MORNING: THE TEN BEST INVENTIONS
I just thought I'd say thanks for managing to get that pricing strategy tidbit out of TSMC last year (don't know if you're the one who managed to make their admit it, if not thanks to both you and that person), that was very interesting :)
I wonder how much the crisis will change those plans though. After all, besides margin-maximization, there are at least two factors at play here: first, fab costs are just getting out of control, and second (more importantly) it's unclear just how much it will be possible to amortize these 40/28nm processes in the long-term.
350/180/130nm are still being used extensively in the analogue, mixed-signal and RF realms. 90/65 will be amortized a lot by connectivity (Bluetooth, WiFi, etc.) and increasingly RF and some chips that are just plain pad limited on newer nodes... There are things you could still be using 40/28nm for in 5 years, but I think nobody really knows for sure and that might be part of TSMC's reasoning.
Another factor I've been thinking about that might play a part in reversing TSMC's strategy - if there's one industry that's being hit harder than foundries, it's fab equipment. Maybe they are being forced to sell new tools (especially immersion ones) at much lower prices than they had expected, and this is playing in TSMC's hand since they had to buy a truckload of them anyway to support the NVIDIA/AMD ramp...
Is there perhaps something you heard about foundries really changing their tune wrt long-term 40/28nm wafer pricing?
Thanks Arun, I think it was me. I'm kicking myself now because the one question I didn't ask Marai Marced yesterday, which I meant to ask, was how this slump in demand will effect wafer prices. Will the policy of holding prices steady over a node be maintained? or will they have to slash prices to fill their fabs? I'm sorry, I grievously failed to get the answer to this one yesterday but I expect there'll be other opportunities and I'll publish anything I get.