Pure-Play versus IDM Foundry: Where will your next design go? (Poll)

[Daniel Nenni]

Having worked with IDMs and foundries through EDA and IP companies for the past 30+ years I have to say IDM Foundries never appealed to me. We covered this is our book "Fabless: The Transformation of the Semiconductor Industry", I think it is important that we have competition and if IDM Foundries are the only ones that can compete with TSMC so be it. My prediction however is that when GF provides a viable alternative to TSMC the IDM foundries will be niche players at best. No matter what the IDM org chart looks like they are competing with their foundry customers. Intel is the clearest example. Intel was partnered with Achronix at 22nm then Intel bought Altera. Xilinx and Altera were at TSMC so this made perfect sense at the time. And now they have purchased Mobil Eye and Nervana. Would anybody in their right mind take their FPGA, ADAS, or AI chips to Intel Foundry? Samsung is much more complicated since they are an IDM, foundry, and electronics giant. If you want Samsung to use your chips you had better use their foundry, right? (QCOM)

TSMC to outpace Samsung in 7nm volume production in 2018
Taiwan Semiconductor Manufacturing Company (TSMC) is ready to enjoy a sure win over Samsung Electronics in the race of 7nm process volume production in 2018, as it has secured orders from over 40 customers for fabricating chips on the advanced process for mobile communication, high-performance computing and AI (artificial intelligence) applications. Both Apple and Qualcomm are among TSMC's major customers, with the foundry house contracted to fabricate all the A12 processor chips for the 2018 new-generation iPhone devices, according to industry sources.

To extend its technological lead over Samsung, TSMC will incorporate the extreme ultraviolet (EUV) technology into its 7nm+ process and step up deployments in the 5nm and 3nm processes.

The company is set to kick off construction of its Fab 18 in the Southern Taiwan Science Park in 2018 as a 5nm production base, and it also plans to invest US$20 billion to build a 3nm fab also in the park, with construction set to start in 2020, the sources said.

The sources continued that TSMC's 5nm process will be an extension of its 7nm process, with application areas still targeting mobile communication, high-performance computing, AI and machine learning. Trial run of the 5nm fab is slated for the first half of 2019.

Samsung countermeasures
On another front, Samsung is aggressively carrying out a spate of countermeasures. Beyond reaching a consensus with the Hwaseong city government over establishing 7nm production lines in 2018 in the Korean city, the company is actively negotiating with US and China customers over new cooperation projects.

Moreover, having spun off its wafer foundry service as an independent business unit in May 2017, Samsung plans to launch 4nm process in 2020 against TSMC's 5nm node, before starting volume production of 7nm process in 2018 and developing 6nm and 5nm processes in 2019, according to industry sources.

The sources cited Samsung's foundry officials as saying that Samsung will strive to grab a 25% share of the global foundry market in five years, up from less than 10% for the moment and compared to around 60% now held by TSMC.

As TSMC has won orders from Apple and Qualcomm for 7nm chip production, Samsung is making extensive contacts with potential customers in various application fields to broaden its foundry service deployments and boost its global market share, instead of focusing on fabricating high-end chips, the sources said.

TSMC to outpace Samsung in 7nm volume production in 2018

 

[Arthur Hanson]

Samsung made a major mistake in handling IP for what is now the world's largest company, Apple. This should give anyone second thoughts about using Samsung. Apple could easily afford their own fab, but wants the breadth and depth of talent TSMC brings. TSMC though is a special case with an unmatched level of expertise in a broad number of fabrication methods. The independent fab spreads cost and expertise over a much larger base for economic advantage for all and will not compete with you, eliminating any conflict.

Note, for full disclosure I have holdings in TSM, AMAT, MU

 

[raghu78]

Having worked with IDMs and foundries through EDA and IP companies for the past 30+ years I have to say IDM Foundries never appealed to me. We covered this is our book "Fabless: The Transformation of the Semiconductor Industry", I think it is important that we have competition and if IDM Foundries are the only ones that can compete with TSMC so be it. My prediction however is that when GF provides a viable alternative to TSMC the IDM foundries will be niche players at best. No matter what the IDM org chart looks like they are competing with their foundry customers. Intel is the clearest example. Intel was partnered with Achronix at 22nm then Intel bought Altera. Xilinx and Altera were at TSMC so this made perfect sense at the time. And now they have purchased Mobil Eye and Nervana. Would anybody in their right mind take their FPGA, ADAS, or AI chips to Intel Foundry? Samsung is much more complicated since they are an IDM, foundry, and electronics giant. If you want Samsung to use your chips you had better use their foundry, right? (QCOM)

TSMC to outpace Samsung in 7nm volume production in 2018
Taiwan Semiconductor Manufacturing Company (TSMC) is ready to enjoy a sure win over Samsung Electronics in the race of 7nm process volume production in 2018, as it has secured orders from over 40 customers for fabricating chips on the advanced process for mobile communication, high-performance computing and AI (artificial intelligence) applications. Both Apple and Qualcomm are among TSMC's major customers, with the foundry house contracted to fabricate all the A12 processor chips for the 2018 new-generation iPhone devices, according to industry sources.

To extend its technological lead over Samsung, TSMC will incorporate the extreme ultraviolet (EUV) technology into its 7nm+ process and step up deployments in the 5nm and 3nm processes.

The company is set to kick off construction of its Fab 18 in the Southern Taiwan Science Park in 2018 as a 5nm production base, and it also plans to invest US$20 billion to build a 3nm fab also in the park, with construction set to start in 2020, the sources said.

The sources continued that TSMC's 5nm process will be an extension of its 7nm process, with application areas still targeting mobile communication, high-performance computing, AI and machine learning. Trial run of the 5nm fab is slated for the first half of 2019.

Samsung countermeasures
On another front, Samsung is aggressively carrying out a spate of countermeasures. Beyond reaching a consensus with the Hwaseong city government over establishing 7nm production lines in 2018 in the Korean city, the company is actively negotiating with US and China customers over new cooperation projects.

Moreover, having spun off its wafer foundry service as an independent business unit in May 2017, Samsung plans to launch 4nm process in 2020 against TSMC's 5nm node, before starting volume production of 7nm process in 2018 and developing 6nm and 5nm processes in 2019, according to industry sources.

The sources cited Samsung's foundry officials as saying that Samsung will strive to grab a 25% share of the global foundry market in five years, up from less than 10% for the moment and compared to around 60% now held by TSMC.

As TSMC has won orders from Apple and Qualcomm for 7nm chip production, Samsung is making extensive contacts with potential customers in various application fields to broaden its foundry service deployments and boost its global market share, instead of focusing on fabricating high-end chips, the sources said.

TSMC to outpace Samsung in 7nm volume production in 2018

Daniel I agree that fabless design companies would be more comfortable working with a pure play foundry like TSMC and GF. Intel and Samsung will be competing with their own fabless customers with products in many markets such as HPC, FPGA, AI, mobile, discrete modems, graphics, ADAS. I think GF has a very good opportunity to become a strong No.2 foundry if they can deliver their FD-SOI and FINFET roadmaps on time and with competitive cost structure (yields). TSMC's leadership position in 7nm with a massive 1 year time to market lead is going to make them an even more dominant No.1 foundry. I think Samsung needs to show extra ordinary execution for atleast 3-5 years to make a significant impact in the foundry business. They have the money from their DRAM/NAND business to fight a long term battle against TSMC. But it remains to be seen if they can execute well and more importantly attract new business from high volume customers.