5 Comments

Summary:

The pain the recession is currently causing for the semiconductor industry has been well documented, but it may also escalate tensions between chip equipment vendors and their customers. An ongoing debate over the need to invest in the next cycle of manufacturing plants has pitted equipment […]

wafer-insertionThe pain the recession is currently causing for the semiconductor industry has been well documented, but it may also escalate tensions between chip equipment vendors and their customers. An ongoing debate over the need to invest in the next cycle of manufacturing plants has pitted equipment vendors such as Applied Materials, Tokyo Electron, KLA Tencor and others against their customers — primarily Intel, Samsung and the large foundries such as TSMC.

Every decade or so, chip makers have transitioned to building their wares on larger wafers, as a way of producing more chips for less money. However, thanks to the recession and the costs associated with the hoped-for transition, not everyone’s convinced this switch needs to happen. Even if it does, it certainly won’t happen as fast as Intel, one of the most aggressive proponents of this transition, hopes.

Imagine the process of building chips as making a many-layered cake. Currently, advanced fabrication plants, or fabs, build 300 mm cakes, but Intel and others are hoping to make bigger cakes — 450 mm. To do this, they need bigger ovens and bigger silicon wafers. Intel estimates such a shift will allow it to produce more than twice the number of chips processed on today’s wafers. The expected savings would be about 30 percent.

But at the same time, the chipmakers are producing smaller and smaller chips with more transistors on them, essentially cutting smaller and smaller slices from the cake. When you see news releases touting the shift from making 65-nanometer chips to 45-nanometer chips, semiconductor makers are trumpeting those smaller slices. (Unlike with pieces of cake, smaller is better when it comes to making chips). The next step down is 32 nanometers, and chipmakers and equipment makers are spending billions of dollars prepping for that shift and subsequent moves to 22 nanometers.

This emphasis on building equipment to make smaller pieces of cake (“moving down the process node”) and the pain caused by the financial crisis have left chip equipment vendors unwilling — and possibly unable — to think about making bigger ovens. The equipment suppliers estimate that redesigning the ovens and other cake-making tools will cost them more than $20 billion in R&D — money they can’t afford to spend.

Jim Ellis, VP of Global Standards and Technology at SEMI, says focusing on 450 mm is going to have to wait. To focus on both moving down the process node (cutting smaller pieces) and making chips on larger wafers (bigger cakes) would be like fighting a war on two fronts when supply lines are already stretched. Other equipment makers contacted for this story agreed but did not want to go on record for fear of angering their customers, notably Intel, which is still pushing a plan to have a pilot 450-mm production line in place by 2012.

When asked if that timeline still holds given the economic climate, an Intel spokeswoman wrote in an email, “I cannot comment on our suppliers’ timelines, but the statement and release below is all we are disclosing at this time, which does still say goal of a pilot line in 2012.” The International SEMATECH Manufacturing Initiative, a consortium of chipmakers, backs Intel’s plans and Scott Kramer, VP of manufacturing technology says he thinks the suppliers are overestimating the costs associated with those plans.

Be that as it may, suppliers have concluded that the switch to 450 mm isn’t worth it, especially right now. A report issued by SEMI explains their perspective pretty clearly by pointing out that the investments made by equipment makers in 300 mm equipment haven’t been recouped yet, and that the savings to chipmakers come at a cost to the equipment makers. With a transition to larger wafers, Intel wins because its per-chip production costs go down. However, the equipment providers have to invest in the R&D to build these larger, more expensive new machines — machines that only the largest chip companies may use.

In the last few months, the equipment industry has cut a few thousand jobs, and iSuppli expects sales of chip equipment to fall to a six-year low in 2009. According to iSuppli, equipment sales were $54 billion in 2007 and are estimated to be about $35 billion in 2009. Against this backdrop it appears that equipment vendors are ready to fight their customers for survival. But so far, those customers keep insisting they are right.

450 mm wafer insertion image courtesy of Sematech

This article also appeared on BusinessWeek.com

  1. I hope they continue their investments regardless of the economic situation present at the moment, if anything, computer power will be required in larger quantities in the future at reduced prices… I don’t see this changing anytime soon!

    Jon
    http://WoodMarvels.com – Create Unique Memories

    Share
  2. I wonder if Intel might do a end-run here by purchasing a company on the cheap, someone like Aviza Technology – http://www.fabtech.org/editor_s_blog/_a/can_aviza_do_a_tegal/
    and do their own 450 mm. Would be a great time to scoop up someone.

    Share
  3. It’s simple: 450mm isn’t going to happen. The only possible customers are Intel, Samsung, and TSMC. Yes, they’re major customers, but not big enough to recoup the development costs.

    Buying one company isn’t going to help Intel, either — they would have to buy a complete set of front end companies, not just one company, and then they would have to pay for the R&D, which is exactly what they don’t want to do.

    Intel is often wrong. For example, they’ve been pushing EUV for a decade or so, and I’ll bet real money EUV will never make it to a production fab.

    Share
  4. [...] isn’t going to stop, the recession, overcapacity in certain markets and the potential for a wafer upgrade could push some of these companies to the brink. [...]

    Share
  5. Sales were ~$12B for wafer processing equipment in 2009 vs the $35B prediction you were given (which may have include test/assy as well, which combined was still only ~$15B). Suppliers were focussed on staying alive and getting ready for the upturn where 450mm is not required. 25nm Flash has been announced and 22nm logic will be in production by Intel in 2011. The opportunities for 450mm insertion are being passed up at the same time cost for future node advancements will escalate with EUV Litho coming online. Prospects for a wafer size change anytime soon are becoming even bleaker.

    Share

Comments have been disabled for this post