January 17, 2012 — Day 2 of SEMI’s Industry Strategy Symposium (ISS) 2012 began with weather related issues. The meeting itself was not affected, but several tee times were delayed due to frost. But it was warm inside, where the rest of us huddled around the barrel fires that sent our hopes and dreams for a quick economic recovery up in flames.
Mike Splinter, Chairman & CEO of Applied Materials, kicked off the second day with Applied’s view of what the future holds for us as we accelerate change. He noted that both PV and displays are single-purpose devices, whereas semiconductors can be tasked for a much broader range of applications. The next 5 years will take us through more inflections than the past 15 years. Success will require improvements in R&D productivity. The 300mm transition cost $12B with a 14 year investment recovery time. The 450mm transition is estimated to cost $15-20B, but the investment recovery time remains a big unknown. The top 5 capital spending fabs grew from 36% of the total spending in 2000 to 75% in 2012. The top 5 capital equipment suppliers grew from 54% of total revenue to 67% over the same period.
|Figure 1.Top 5 capex spenders, tool suppliers.|
Mark Thirsk of Linx Consulting opened the materials session with his prognosis on the health of the global electronic materials supply chain. If 450mmm fabs are not built, the number of 300mm fabs required will exceed 300 by 2028; the 450mm transition is required on economic grounds alone. Delaying the implementation of EUV will result in an additional industry cost of $8.5B in 2016 due to the use of less productive double patterning. The semiconductor materials CAGR for 2011-2020 is 12%. The cost of process materials as a percentage of chip revenue will rise from its 2002-2011 average of 1.7% to 3.5% by 2020. Photoresist is no longer performing the ‘resist’ portion of its function, being too thin and having sacrificed etch resistance for pattern definition. There is abundant opportunity for consolidation in CMP slurry, but little chance in photoresist due to strong legacy supplier relationships. Patterning, CVD/ALD and CMP will drive the materials demand growth. The PV materials demand is forecast to be $25B by 2015.
|Figure 2. Global electronics materials industry, 2011.|
Tim Hendry, VP of Intel’s Technology & Manufacturing Group, addressed the advanced products that are enabled by materials development. Early engagement and close collaboration through joint development agreements has been an increasing component of the R&D plan since 130nm. Volume forecasting is becoming more critical to maintain profitability for both supplier and fab. In some materials, barrels and totes are giving way to tankers as consumption volumes increase. Business continuity planning is now focusing on better customizing each material’s reserve based on its supply chain complexity and specific risk levels throughout.
David Lazovsky, Founder of Intermolecular, talked about the collaboration models used by his company to accelerate innovation and time to market. Their process tools are built specifically for the purpose of rapid R&D learning in both wet processes and thin film deposition processes. Given its focus on innovation, it’s no surprise that they have filed 207 patent applications in the past 12 months. He believes the sub-22nm CMOS logic IDM base will shrink to Intel, TSMC, GlobalFoundries and Samsung. The speed of innovation in mobile computing is a strategic match for IM’s growth objectives, as is the increasing dependence on materials and device architecture innovation.
|Figure 3. Performance improvements. SOURCE: IBM Microelectronics, IBM, IC Insights.|
Harvey Frye, Vice Chairman of TEL, moderated the semiconductor markets discussion panel which included Bob Johnson of Gartner, Egan Christiansen of Dell, Will Stofega of IDC, and Paul Semenza of DisplaySearch. Harvey: During the week of December 25-31, 2011, there were over 20M new subscriber activations for Apple and Android smart phones. The number of downloaded apps that week was 1.2B globally, of which 500M were in the US and 100M in China. The Apps Store generates 4% (~$7B) of Apple’s market cap; this is about the same as RIM’s total market cap. Bob: By 2015 the semiconductor industry will derive $375B, almost 2/3 of its revenue, from mobile cloud computing. The internet handles 15 petabytes of new information every day. Over 50% of internet connections today are things (largely public utility related), as opposed to computing or phone devices in the hands of an individual. Implementation of predictive analytics will result in a cloud that knows what you want to do before you do it, and then influence you to do it. Egan: Facebook handles 1M updates and photo postings every minute. The public cloud services market will be $176B by 2015. 43% of the US workforce (63M workers) will be telecommuting by 2016. Will: Smart phone units shipped will surpass ‘feature phones’ in 2015, at 1B each. Paul: As internet-connected TV evolves, it will morph strongly from broadcaster-controlled services to consumer-controlled services. The cloud will replace the TV as the central connection point to which other devices must conform.
|Figure 4. From the Internet of Things to the Internet of Everything.|
Handel Jones, Founder of IBS, forecast some high level IC trends for the next 5 years. We will continue the trend of a 10x increase in wireless bandwidth every 5 years. Demand is strong but delivery is lagging for 28nm HKMG devices; 20nm is expected to ramp even more slowly. Intel is thought to be 18-30 months ahead of the industry in FinFET implementation. Declarations of insolvency are expected for several DRAM manufacturers, which he declined to identify. The projected cost per gate is increasing slightly at 22nm and 14nm, a trend that must be reversed. IC revenue growth will be low as fabs lower prices to protect market share in an unstable global economy. Qualcomm will be a $30B fabless company in 2015. Samsung will challenge Intel for the position of largest semiconductor manufacturer. TSV must reduce cost and stress problems in order to become competitive. 3D packaging can provide a competitive advantage equivalent to a half node.
|Figure 5. Cost per gate trends.|
Dan Hutchenson, CEO of VLSI Research and winner of the 2012 Bob Graham Award, kicked off the 450mm discussion panel with the expectation that while it will clearly be expensive, the 450mm transition will not be as costly overall as the oft-delayed 300mm transition. Spending on 450mm has been $0.5B so far, with another $7.6B needed by 2020. The equipment business, having flat-lined at ~$56B for several years, is not keeping pace the overall growth of the semiconductor fabs; the cost model needs to change to sustain development. Panelists were John Chen of Nvidia, Paolo Gargini of Intel, Randhir Thakur of Applied Materials, Takahashi Abe of Sumco, and Kazuo Ushida of Nikon. John: Transistor cost is not scaling with node for 28nm, 20nm and 14nm; they are all falling to the same normalized cost per transistor, as did 55nm and 40nm. Paolo: 450mm equipment development will take place in 2012-2014 and begin to move into production in 2015. The tipping point will come when litho production equipment is available. Randhir: The 450mm investment is stacked on top of 3 additional nodes that need to be developed for 300mm. Abe-san: Coordinated timing will be critical to achieve a cost-effective rollout of 450mm supplies. Ushida-san: Technology progress will drive technology decisions. CEO’s need to drive the economic decision for 450mm litho. EUV is a technology-gated issue that must be kept separate.
|Figure 6. Suppliers have not shared in the semiconductor industry’s growth.|