Are foundries trying to inflate wafer prices?

Thursday 29th November 2007, 12:12:00 PM, written by Arun

A couple of weeks ago, TSMC and UMC announced that their capital spending in 2008 would be much more limited than this year, as less new equipment would be purchased for capacity expansion. Many had expected this to be related to expected productivity enhancements and deployment of equipment bought in 2007. But now another theory has surfaced...

Data

  • David Manners reports that Bill McLean, president of IC Insights, believes the industry will be strong in 2008. Demand will thus increase faster than capacity expansion, resulting in (slightly?) higher wafer prices from foundries.
  • The foundries' reasoning is that their wafer ASPs has been going down in recent past, just as many companies are now moving to an asset-light strategy. They are fed up with that situation, McLean claims, and want to change things.

Quick analysis

  • As the industry moves to 45nm, many companies will be increasing their reliance on foundries. This will only become even more true at 40nm and 32nm.
  • This makes next year perfect to try to increase wafer prices a bit: the decisions have generally already been made, and nobody is going to turn back now and give up on its asset-light strategy.
  • Multi-sourcing from multiple foundries, and even multiple processes (not phasing out products on old processes faster than they need to be) may become required.
  • A number of companies are already multi-sourcing aggressively, including Texas Instruments, Qualcomm and NVIDIA.
  • The two former have had many major designs multi-sourced at all three major foundries (TSMC/UMC/Chartered).
  • NVIDIA has recently reduced its reliance on TSMC by also taping-out G92 and G98 at UMC.

As capacity expansion slows down but demand keeps up, the leading foundries are nicely positioned to further improve their financial position. It may be ironic that multi-sourcing becomes increasingly relevant as mask prices and process-related design costs go up, but that's the way things seem to be heading next year.


Discuss on the forums

Tagging

b3d ± tsmc, umc, nvidia


Latest Thread Comments (8 total)
Posted by Geo on Thursday, 29-Nov-07 15:33:06 UTC
Hahhaha. . . good old law of unintended consequences. It's really classic economics in my book. Less competition means higher prices. And "asset light" means less competition.

Posted by Voltron on Thursday, 29-Nov-07 16:39:49 UTC
I don't really think this is anything new or unusual in the foundry business. Any manufacturing company is looking to maximize capacity utilization.

The foundries became much more cautious after the tech bubble burst in 2002/2001. These companies need to be highly profitable in order to continue to reinvest in future capacity. TSMC's margins are huge, which is pretty rare for a manufacturing company, particularly a contract manufacturer. Compare them to circuit board makers and you will note a dramtic difference. Everybody understands that semi plants and equipment are a different ballgame and that high margins are required for a sustainable business. Anyhow, they really spent a lot on capacity just prior to the bubble bursting and they got caught with their pants down. Capacity utilization fell from over 100% to below 50% pretty much overnight. That just killed profits and put some of the foundries' long-term futures at risk. Anyhow, since then the fabs have been very cautious about over spending and have frequently adjusted capex down. Plus they can always get better at doing more with less, just as NVIDIA has with its wafer usage.

Posted by Arun on Thursday, 29-Nov-07 19:40:25 UTC
Well, fundamentally the reason why the gross margins and the operating margins are so large at these companies is that the majority of their costs are capital spending. So it's only visible in net income and through amortization charges etc.

Even after taking amortization into account, it is true that foundries are getting pretty good returns already though - or at least, TSMC does. UMC's returns are decent, and Chartered's are subpar but at least they've been improving in recent years.

Obviously everyone is expecting foundries to increase capacity next year, mostly through "doing more with less" as you said. However, the point is that growth in capacity doesn't look like it'll match growth in demand, despite utilization rates being already very high. So either foundries are expecting the market to be worse than everybody else, or they know their strategy will likely result in supply restrictions for their customers.

And if the answer is that they do know this will result in supply restrictions (which is what McLean implied, and the point of this article), then it means foundries are realizing they are in a position of power next year and will be able to slightly increase wafer prices this way with no real consequence.

Nothing unusual in the manufacturing world overall, but clearly that isn't what happens every year (and especially not in the semiconductor foundry business) so it's still very noteworthy, IMO.

Posted by Voltron on Friday, 30-Nov-07 14:45:12 UTC
Operating margins include dep and am, unless specified as EBITDA. In all accounting systems I am familiar with it is always included in operating expenses on the income statement. This is neither here nor there, but if a capital intensive company didnt include dep in gross margins they would really be through the roof. Anyhow, the foundries have high net profit margins for a manufacturing business which is obviously what I was referring to in the context of reinvestment.

As far as capacity next year, you may have noticed these companies continually revise their spending plans up and down. This goes with what I said in my original post. They learned a harsh lesson after the tech bubble and are much better at adding capacity quickly and as needed. This is just common sense when you have equipment that costs this much. And if you get better at adding it as efficiently as possible you run a better business in the long-term for you and your customers.

TSMC does not want its customers to be supply constrained. That is for sure. It is obvious they are very customer centric. Ask anyone in the industry. But it is impossible to be perfect in manufacturing planning, Prior to the tech bubble they were adding capacity as fast as possible and it bit them very very badly. So now they are more cautious. And they have been for the past 6 years.

Posted by INKster on Friday, 30-Nov-07 14:50:11 UTC
Quoting Arun
Well, fundamentally the reason why the gross margins and the operating margins are so large at these companies is that the majority of their costs are capital spending. So it's only visible in net income and through amortization charges etc.Even after taking amortization into account, *it is true that foundries are getting pretty good returns already though - or at least, TSMC does*. UMC's returns are decent, and Chartered's are subpar but at least they've been improving in recent years.
Yet, i can't figure out exactly why did Phillips bail out of TSMC. :???:

Posted by silent_guy on Friday, 30-Nov-07 17:56:21 UTC
Quoting INKster
Yet, i can't figure out exactly why did Phillips bail out of TSMC. :???:
Because all electronics companies are doing it...
The investments are still increasing and lock up in credible amounts of capital. If fabs aren't your main business and they don't give a clear competitive advantage and you think you can deploy the capital more efficiently somewhere else, it's the logical thing to do.

Posted by Arun on Friday, 30-Nov-07 20:29:31 UTC
Quoting Voltron
Operating margins include dep and am, unless specified as EBITDA. [...] Anyhow, the foundries have high net profit margins for a manufacturing business which is obviously what I was referring to in the context of reinvestment.
Ah yes, this is completely my mistake - sorry for claiming somethin that ridiculous on my previous post, I don't know what I was thinking :)

Quote
TSMC does not want its customers to be supply constrained. That is for sure. It is obvious they are very customer centric. Ask anyone in the industry. But it is impossible to be perfect in manufacturing planning
Obviously they don't want to massively limit anybody's capacity. However, one way to look at this might be that in the last few years, they probably were shooting for 80-90% utilization so as to be able to capture new share if the opportunity arose. Next year, they look to be aiming at 95-100%, and if the upside happens, well wafer prices will go up a bit and margins will be even more optimal.

I don't think anyone is claiming any drastic change in strategy here, but AFAICT, small changes can result in interesting results. And I don't think anyone could deny TSMC and UMC are being rather conservative for next year - and while spending plans do change rapidly, capacity doesn't get installed overnight, so what they predict today does matter for tommorow, IMO.

Posted by Voltron on Saturday, 01-Dec-07 00:12:11 UTC
I think this article might provide a little more incite than the Electronics Weekly piece. I wouldn't take TSMC's claims of productivity improvements as hot air. Plus, because of certain previous outlays, they might not have to spend as much to boost capacity on relatively short notice. Obviously TSMC wants to run as close to max capacity as possible, but much of their success is clearly due to their long-term focus and I really doubt that they would want to constrain customers in order to achieve a few more dollars.http://www.fabtech.org/content/view/5641/


Add your comment in the forums

Related b3d News

Sub $100 graphics at Tech Report
Broadcom purchase AMD's DTV business
Lucid Hydra 100 multi-GPU scaling demonstrated at IDF
The Khronos Group announce Heterogeneous Computing Initiative
3DMark, the Game
PC's Assassin Creed Official System Requirements: 2GB of RAM
Beyond3D's 2007 'Bricks & Bouquets'
Illuminate Labs adds its lighting baking solutions to UE3
MegaTexture in Quake Wars
Analysis: Q3 2007 GPU shipments reach record highs