05-09-2014, 11:41 PM
Semicon Taiwan 2014---Key Take away
Most equipment vendors at the Semicon Taiwan were quite bullish, with expectations of a 4Q14 recovery in capex spending and strong progress in advanced process nodes (14/16nm and 10nm R&D progress). Advanced packaging is seeing a lot of attention from Foundries and OSATs, while most vendors are counting on the Internet of Things to drive the next leg of growth.
■ Wafer level packaging accelerating, substrate-less solutions in vogue:
Foundries and OSATs are both exploring wafer level packaging options, for both legacy products (conventional wafer level packaging) and advanced products with larger pin count (Fan out wafer level packaging). Several vendors talked about reducing substrate content in advanced packaging – with core-less substrate, using Lead-frames for flipchips, etc. We came away feeling more negative on the substrate market and also worried about the competition between Foundries and OSATs in wafer level packaging.
■ Semiconductor capex likely to bounce back from 4Q14 or 1Q15:
Most equipment vendors expect the current trough in equipment spending to recover from 4Q14 or 1Q15 at the latest. IDM orders (likely Intel 14nm) and Foundry is likely to lead the recovery in the short-term. Looking into 2015, Memory appears to be growing the fastest, followed by Foundries competing for 14nm/16nm, while IDMs should remain largely stable.
■ 8” Foundry tightness–High degree of optimism on MEMS and IoT:
Most market participants expect 8” wafer capacity tightness to last into 2015 and are placing high degree of faith on take-off of MEMS and Internet of Things market to drive continued 8” demand. However, a minority (where we also fall) are concerned about conversion of some large volume applications (Fingerprint sensors, Power management ICs) to 12” which could lead to supply easing out.
■ TSMC’s 10nm progress confirmed, EUV commentary getting better:
Several equipment vendors showed a high degree of confidence in TSMC’s 10nm process plans and confirm an acceleration of schedule, including two or more fabless vendors entering tape-out stage in 2H15. Most vendors are looking at production starts in late 2016 and large volume production in early 2017. EUV progress also appears good with our checks confirming that TSMC also is hitting 600 wafer per day throughput rates with its tools.
■ Apple's next SoC – TSMC likely to do iPads, Samsung/ GF with iPhones?:
Current consensus among industry participants appears to be that Samsung/ GF may win the iPhone application processor socket in 2015, while TSMC could continue to produce iPad APs. However, most metrology vendors agree that 14nm is still seeing yield issues, which could prolong the transition to 14nm.
■ E-Beam Inspection to see more traction at 10nm:
Hermes Microvision (NC) is seeing good traction for 10nm for its e-Beam inspection tools in R&D as well as production stage, with a potential doubling of tool set per customer, compared to 16nm, according to management. Progress on multi-column e-beam is ongoing but tangible demand for in-line deployment is still not yet clea
Most equipment vendors at the Semicon Taiwan were quite bullish, with expectations of a 4Q14 recovery in capex spending and strong progress in advanced process nodes (14/16nm and 10nm R&D progress). Advanced packaging is seeing a lot of attention from Foundries and OSATs, while most vendors are counting on the Internet of Things to drive the next leg of growth.
■ Wafer level packaging accelerating, substrate-less solutions in vogue:
Foundries and OSATs are both exploring wafer level packaging options, for both legacy products (conventional wafer level packaging) and advanced products with larger pin count (Fan out wafer level packaging). Several vendors talked about reducing substrate content in advanced packaging – with core-less substrate, using Lead-frames for flipchips, etc. We came away feeling more negative on the substrate market and also worried about the competition between Foundries and OSATs in wafer level packaging.
■ Semiconductor capex likely to bounce back from 4Q14 or 1Q15:
Most equipment vendors expect the current trough in equipment spending to recover from 4Q14 or 1Q15 at the latest. IDM orders (likely Intel 14nm) and Foundry is likely to lead the recovery in the short-term. Looking into 2015, Memory appears to be growing the fastest, followed by Foundries competing for 14nm/16nm, while IDMs should remain largely stable.
■ 8” Foundry tightness–High degree of optimism on MEMS and IoT:
Most market participants expect 8” wafer capacity tightness to last into 2015 and are placing high degree of faith on take-off of MEMS and Internet of Things market to drive continued 8” demand. However, a minority (where we also fall) are concerned about conversion of some large volume applications (Fingerprint sensors, Power management ICs) to 12” which could lead to supply easing out.
■ TSMC’s 10nm progress confirmed, EUV commentary getting better:
Several equipment vendors showed a high degree of confidence in TSMC’s 10nm process plans and confirm an acceleration of schedule, including two or more fabless vendors entering tape-out stage in 2H15. Most vendors are looking at production starts in late 2016 and large volume production in early 2017. EUV progress also appears good with our checks confirming that TSMC also is hitting 600 wafer per day throughput rates with its tools.
■ Apple's next SoC – TSMC likely to do iPads, Samsung/ GF with iPhones?:
Current consensus among industry participants appears to be that Samsung/ GF may win the iPhone application processor socket in 2015, while TSMC could continue to produce iPad APs. However, most metrology vendors agree that 14nm is still seeing yield issues, which could prolong the transition to 14nm.
■ E-Beam Inspection to see more traction at 10nm:
Hermes Microvision (NC) is seeing good traction for 10nm for its e-Beam inspection tools in R&D as well as production stage, with a potential doubling of tool set per customer, compared to 16nm, according to management. Progress on multi-column e-beam is ongoing but tangible demand for in-line deployment is still not yet clea