Global foundry war breaks out, extended from the previous 28nm to 16nm / 14nmFinFET processes. 28nm process can be regarded as the turning point of the semiconductor industry. Because according to the size reduction has been driven by industrial progress so far to change dramatically, usually every two years ahead of a process node, reducing manufacturing costs by about 50 percent of the rhythm, began to slow down to 28nm or less, but in some cases the cost will rise. Reflecting the industry level, the enterprise to the power forward has a smaller size than ever before. Many top IDM companies, started asset light strategy (fablite) from 28nm, embrace foundry. This led to a flourishing global foundry.
Especially after Intel announced a few days ago, Samsung 14nm production will make the battle for foundry 16nm / 14nm order to become the new focus. Global foundry first camp, originally had TSMC, UMC pass Luofang De (Globalfoundries), now IDM super giant Samsung and Intel have joined the ranks of the foundry, leading foundry in the first camp battle situation presents complicated.
Apple and Qualcomm order to weathervane
Samsung and GlobalFoundries 14nm alliance may advance into mass production, Qualcomm and Apple orders will become focal points of contention.
70% of global OEM customers from fabless, but according to the current situation, in which the high-pass and two large orders Apple has become the focus of contention. Reflecting the semiconductor industry momentum is shifting from PC mobile terminals, including mobile phones and tablet computers, the whole industry supply chain was a big change.
Qualcomm chip and rely on patents to provide global smart phone, 2013 sales had reached $ 24.8 billion, of which one-third comes from licensing fees. Ddaily2014 in May, according to data in the global mobile processor market, Qualcomm share were as follows: 2011 48.7% 2012 42.9% 2013 53.6%. Apple corresponding 14.4% respectively, 15.90% and 15.70%. This is because Apple uses its own processor chip design only in mobile phones and tablet PCs, Mac computers still use Intel chips.
Qualcomm has undisputed dominance in the mobile processor industry, although there are already other fabless manufacturers for the formation of a high-pass pressure, but it is still very obvious advantages, the major flagship brand mobile phones almost exclusively use Qualcomm chips. Qualcomm’s current high-end chips Xiaolong 805,64 8-bit core, using 20nm process.
Apple also holds the hands of large orders from Samsung OEM before the recent design of its own 20nm process has partially moved TSMC A8 chip processing. TSMC is obviously want to continue to build on this, for 2015 its 16nm production line A9 processor for Apple’s foundry. But the situation is complicated, both competitive relationship, there are also technical issues. No one knows, in 2015 Samsung and GlobalFoundries 14nm alliance may advance into production, with the advantage compared to TSMC’s 16nm process. Chang also admitted that even this step may be behind TSMC, it is pinning its hopes on the 2016 battle of 10nm process. However, Samsung and GlobalFoundries Union may be insufficient capacity plague, adding to the variables for the next competition. The TSMC 20nm process already has production capacity of 70,000 / month. No matter what, Qualcomm and Apple orders will be focal points of contention, but Apple’s principle is not put all your eggs in one basket.
OEM battlefield who can win?
TSMC foundry in dominating the situation, despite the near future can not shake it, but it will be eroded.
On behalf of the industry’s growth can not be achieved overnight from which TSMC’s unshakable dominance. Especially in 2009 in charge of the company since the second Chang, using chilling active investment expansion strategy, the total investment in the period 2010 to 2013 amounted to more than 300 billion dollars, making advanced process technology continues to advance, again reinforces the leading foundry status and get results very gratifying.
2013 TSMC total capacity of about 1.3 million per month (8 inches), including 28nm capacity to produce 130 000 (12 inches), the global market share in terms of sales amounted to 80%. And it’s 28nm ramp is very fast, the fourth quarter of 2011 has just started its 28nm steps quarter was $ 150 million in sales to the end of 2012 have been accounted for $ 17 billion in annual sales of 24 percent, reaching $ 4.08 billion , accounting for nearly 37% of $ 20 billion in sales by the end of 2013, reaching $ 6.5 billion.
To 60,000 wafers per month capacity to calculate the average price of TSMC 20nm process wafers system is estimated at $ 6,000 per tablet in the fourth quarter of 2014, with 28nm wafer average price (about $ 4,500 to $ 5,000) compared to has greatly improved. The estimated cost of production FinFET wafer 16nm / 14nm is about $ 4,000 per piece, with gross margin of about 45%, the sales price was $ 7,270 per piece. If TSMC 20nm process for accurate prediction, share the point of view of its 20nm process as a whole, will reach 95% of the world in the fourth quarter of 2014.
It can be seen TSMC unshakable dominance reasons: First, the yield up to 90%, about 70% of opponents might; the second is to provide support to customers have about 6300 IP patents, if the industry dubbed a “library “like; the third is the capacity in place quickly, such as 28nm production capacity of 130,000 per month, three times the GlobalFoundries.
TSMC foundry in dominating the situation, despite the near future can not shake it, but will be eroded, that it is unlikely to sustain more than 48% of the high gross margins. Both of which compete for 16nm / 14nm 28nm foundry there. 28nm is still the main battlefield, because the global market size of this piece is about $ 8 billion – $ 10 billion. Advantages Samsung and GlobalFoundries alliance is to propose a new FD SOI’s 28nm process routes that require lower power chips attractive. In addition, their price will be lower than TSMC foundry.
As UMC and SMIC 28nm, even though they claim to be ready right away, but limited production capacity, as well as technical climbing time, you want to get a big breakthrough, it takes time.
In addition, it is undeniable Intel is a potential rival. Because the process technology level of technology and research and development investment point of view, it certainly is the world’s most advanced. However, I want to turn to OEM Intel is not an easy thing from IDM mode. On the one hand is a lot of capacity to turn foundry, Intel ideologically yet final determination under the average gross profit margin of up to 68 percent of Intel processors, foundries margin may not be so high. On the other hand, it lacks the necessary supporting OEM’s IP, it’s not ready to get it ready in the short term. So Intel executives recently revealed its intention to maintain only a handful of high-margin by OEM customers.
Industry structure Prospects
In 16nm / 14nm sector, TSMC, Samsung, Intel and other home, and do not have to win to grasp.
And 28nm foundry industry is different from future global 16nm / 14nm and below foundry pattern Prospects. Because the current state of the global semiconductor industry is this: every two years forward from one process node technology, the theoretical value is 14nm in 2013 and 2015 10nm, 14nm Intel in fact production was postponed to the fourth quarter of 2014, compared than normal, extending the two quarters. TSMC is smart, claiming 2014 is 20nm volume production in 2015 is the year and volume production of 16nm point in time. Samsung Electronics launched advanced foundry process 14nm FinFET application processor (AP) test products, will be provided to Qualcomm, Apple, Advanced Micro Devices (AMD) and other major customers, but now their production capacity, Samsung only produce 10,000 ~ 15000, 35000 Grosvenor square moral. Plus two monthly total was 50,000.
Even more critical is that the current 10nm process technology are under development, including Intel, TSMC and Samsung included, still away from real production, which has a lot of variables. The most optimistic forecasts, 10nm manufacturing process but also to the 2016 production. In addition, after the 10nm process, how do they go down, it is not very clear, including EUV difficult to predict when it is ready, 10nm lithography process when cost and 193nm technology and other alternatives gate material is not yet fully in place.
According to Gartner’s view, the near term, in the amount of 1 to 2 years FinFET production, global foundry capacity requirements will not exceed 50,000 per month. By 2018, there will be no more than 250,000 monthly market demand. And this demand has two large foundries have enough, so now many foundries have entered frontline FinFET technology, the future will find someone speechless.
According to an analyst at Pacific Crest Securities in perspective, from the scale of investment dollars, 16nm / 14nm FinFET technology investments 10,000 capacity to be $ 1.27 billion of investment, an additional 20,000 need $ 2.5 billion investment.
There is also a problem can not be made through the technical aspects of who can really pass, that the yield can sync it? According to the current situation, no matter the amount of TSMC’s 20nm Intel’s 14nm production or production, have emerged to postpone the phenomenon, there exist problems which yield. Therefore, the future of technology Who really needs to observe clearance.
Objective analysis, Intel has a certain advantage. Because it’s from 2011 began using 22nm FinFET technology nodes, it has been the second generation, and experience relatively more. Otherwise originally Altera TSMC old customers can not happen, the situation suddenly turned to embrace Intel’s 14nm chip foundry. Also, the recent Matsushita Electric also under orders to Intel. Enough to prove that it may head higher in the technical aspects. But it is undeniable that there are many problems in the Intel foundry. So in 16nm / 14nm in this section, regardless of TSMC, Samsung and GlobalFoundries Union, or Intel, Which do not have the certainty of winning.
In summary, the future of the global 16nm / 14nm and below foundry pattern Prospects.