What the 28-Nanometer Shortage Means for the Chip Industry
Near-term winners include ATMI and KLA-Tencor.
Several leading chipmakers, including Qualcomm (QCOM) and Nvidia (NVDA), have announced that they will have trouble fully satisfying demand in the coming months because of capacity constraints for cutting-edge 28-nanometer chip wafers. Taiwan Semiconductor (TSM), or TSMC, the world's largest foundry and technological leader in 28 nm chip production, simply doesn't have enough production capacity on hand to fulfill all its orders. The shortage is likely to cause a scramble in the chip supply chain--foundries will race to expand capacity and buy new equipment, while chipmakers will look at alternatives, such as using 40 nm chips or switching to competing foundries, in order to fulfill demand.
How We Got Here
Initially, rumors swirled that TSMC's production yields for 28 nm chips were subpar and that technological roadblocks were the source of many customers' frustrations. But more-recent comments indicate that 28 nm yields are acceptable, but TSMC simply doesn't have enough production capacity to fully meet demand.
TSMC expects to sell out of 28 nm chip production and see 19%-21% sequential sales growth in the second quarter. One of its foundry competitors, United Microelectronics (UMC), also expects a 15% sequential increase in chip unit shipments next quarter. An interesting dynamic should play out in the coming months, as UMC won't have 28 nm production capabilities until later this year, while it will take TSMC several months to put more 28 nm capacity in place. UMC could end up with a terrific 2012 if it were able to win this race and steal some short-term chip orders by Qualcomm (or others) away from TSMC.
Foundries tend to be in the sweet spot when demand outstrips supply, as they are able to fill more of their manufacturing capacity and ensure healthy pricing for wafers. The next steps for foundries involve hefty capital expenditure investments to expand capacity and take on additional orders. Along these lines, TSMC announced that its 2012 capital expenditure budget will expand by at least $2 billion to $8.0 billion-$8.5 billion, with the bulk of this increase pertaining to both 28 nm and future 20 nm chip production. Historically, foundries have overexpanded after shortages and quickly entered cyclical downturns where they faced unused capacity, soft wafer pricing, and lower profitability. However, because demand for 28 nm chip production has been so robust, especially from mobile processor makers, we see less of a risk that TSMC will overexpand this year.
We believe that TSMC, in particular, has been able to use its immense scale (the source of its narrow economic moat) to its advantage in recent years, as it has become more cost-prohibitive for smaller foundries to keep up with the latest generation of chip production. Only a handful of foundries (TSMC, UMC, Samsung, GlobalFoundries) have a chance to stay on the technological forefront in the years ahead. Before the shortage, we believed TSMC would be the go-to foundry, while Samsung and GlobalFoundries would not only look to strengthen their current foundry relationships (such as Samsung's with Apple for its A-series processors, or GlobalFoundries with Advanced Micro's (AMD) computer processors), but also strive to displace UMC for second-source chip orders. We still tend to believe that our original thesis will play out, but we have greater concern that the 28 nm capacity shortage at TSMC may have strained its relationships with customers, thus leaving the door open for clients like Nvidia and Qualcomm to actively look to other foundries, including Samsung and GlobalFoundries, for a larger portion of their 28 nm production later this year. Further down the road, the supply constraints may also encourage major customers to look to TSMC's competitors as key suppliers for their 20 nm chip production.
Chipmakers Affected by the Shortage
We believe that much of the 28 nm shortage stems from a flood of new orders coming in from smartphone chipmakers. Traditionally, chipmakers at the front of the line for new manufacturing technologies include FPGA (field programmable gate array) firms like Altera (ALTR) and Xilinx (XLNX) and PC graphics processor (GPU) chipmakers like AMD and Nvidia. With 28 nm, however, semiconductor titan Qualcomm entered the fray, seeking smaller, faster 28 nm chips for its upcoming line of mobile processors and LTE-compatible baseband chips. We suspect that Qualcomm's LTE baseband chips will be used in Apple's upcoming iPhone 5, and that Qualcomm will need to rely on 28 nm manufacturing in order to provide Apple with sufficient energy efficiency from these baseband chips, as poor battery life has been a common customer complaint for previously launched LTE-based smartphones. We suspect that mobile processor makers will also strive to join FPGA and GPU chipmakers at the head of the pack for future generations of leading-edge chips, such as 20 nm, which should come out in 2014.
Several fabless chipmakers have announced that the 28 nm shortage will limit each firm's ability to fulfill chip demand from their customers and will place a cap on each of their revenue growth in the June quarter. Many firms also indicated that they'll need to keep a watchful eye on 28 nm capacity in the second half of 2012. Ultimately, we believe that fabless chipmakers will adequately navigate through the 28 nm shortage in 2012 and we don't anticipate making material revisions to our fair value estimates for fabless chipmakers based on the shortage.
Altera intends to stick with TSMC through the shortage and may sell older generations of chips to its customers. In contrast, Xilinx believes that it has enough 28 nm chips on hand to survive the shortage (that is, until TSMC puts more capacity in place) because it is a lower-volume customer of TSMC's and uses a slightly different manufacturing process. We don't believe the 28 nm issue will have a long-term effect on the rivalry between these two FPGA firms, but we think that Xilinx has more at stake in the 28 nm FPGA generation, and we're encouraged to learn that the shortage won't put it at a disadvantage.
Qualcomm has long maintained a diversified foundry strategy, and it will seek to source 28 nm chips from other foundries. Qualcomm has also touted its strategy of developing integrated processors that combine the applications processor (used to run apps and phone operating systems) and the baseband function (used to connect the phone to the network), but in light of the shortage, it will try to sell discrete chips to customers in order to piece together a solution that is comparable to its integrated 28 nm Snapdragon chips.
In the GPU chip segment, AMD and Nvidia have new graphics processors that are manufactured on TSMC's 28 nm process. AMD has been seeing strong demand for its latest Southern Islands GPUs and has so far been able to meet customer commitments, but management said supply constraints have limited the upside opportunities that the firm could have achieved with the product. Nvidia appears to also be in a tough position, as robust demand for its new Kepler GPUs is exceeding supply. Nvidia's management initially blamed poor 28 nm manufacturing yields at TSMC for the Kepler shortages, but now says that actual capacity constraints are limiting the supply of the chips. While it anticipates that availability should improve, the shortage issue probably won't be resolved until later in the year.
Mobility Trend Will Drive Demand for Cutting-Edge Chips
The proliferation of smartphones and tablets is driving demand for faster chips that are more power-efficient, which are enabled by the most advanced semiconductor manufacturing processes. Additionally, some of these chips, particularly mobile processors, are having more features and processing cores being added to them, resulting in larger die sizes and higher per-unit costs. Chipmakers like Qualcomm can offset this trend by aggressively staying at the forefront of Moore's Law (in which Intel cofounder Gordon Moore said the number of transistors on a chip will double approximately every two years), as the smaller semiconductor circuitries that can be achieved with cutting-edge manufacturing technologies provide them with the ability to shrink chip sizes and lower unit costs.
The foundry production ramp at the 28 nm node is in the early innings, as only a handful of chipmakers, all of which use TSMC, currently have products based on that manufacturing process. As TSMC and the other major foundries alleviate the 28 nm supply shortages by adding manufacturing capacity, more fabless chipmakers will introduce chips manufactured at the 28 nm process. Aside from smartphone and tablet unit growth, the buildout of the cloud and communications infrastructures to support the mobility trend will also boost demand for 28 nm production, as chip suppliers to these opportunities will also look to take advantage of performance and power efficiency enhancements. We think the technology trends of mobility and the cloud are placing a renewed importance in the semiconductor industry on pushing Moore's Law.
Chip Equipment Market Will Benefit from Additional Foundry Spending
The 28 nm supply shortages at TSMC should benefit the semiconductor equipment market, as foundries will raise their capital investment levels in order to expand their capabilities at that manufacturing process. Most of the original forecasts for the front-end wafer fab equipment market in 2012, including those from Applied Materials (AMAT), Lam Research (LRCX), and market research firm Gartner, had been in the $30 billion-$35 billion range; we had estimated that spending would come in at the lower end of that range. However, we believe capacity constraints in the 28 nm process could result in an additional $3 billion-$4 billion in foundry wafer fab equipment spending for the year.
TSMC has already raised its capital spending outlook for 2012 to $8.0 billion-$8.5 billion from about $6 billion in response to its 28 nm capacity constraints. About $1.4 billion of the firm's incremental spending will be for additional 28 nm manufacturing, while roughly $0.7 billion will go toward development at the future 20 nm node. We think the 20 nm development spending is quite telling of the fact that the smartphone/tablet trend is pushing the foundries to aggressively pursue advances in semiconductor fabrication technologies.
We believe the other major foundries are looking to take advantage of the 28 nm shortages at TSMC to gain market share. In particular, we expect Samsung to raise its full-year capital expenditure outlook to build out additional 28 nm foundry capacity in order to woo some of TSMC's customers. Samsung, the world's largest memory chipmaker, has ambitions to expand its small but emerging foundry business beyond manufacturing the A-series smartphone/tablet processors for Apple. At the beginning of the year, the firm announced 2012 capital investment plans of KRW 15 trillion (about $13 billion) for its semiconductor business (both memory and foundry) in 2012, up from KRW 13 trillion in 2011, with the majority of the increase going to expanding the foundry segment. We suspect that Samsung views the 28 nm capacity constraints as a great opportunity, served on a silver platter by TSMC, to capture foundry market share, and there are rumors that Qualcomm, Nvidia, and AMD are looking to the firm for at least some 28 nm chip production. As a result, we think Samsung could raise wafer fab equipment spending by about $1 billion-$2 billion to steal 28 nm business from TSMC.
For the time being, we believe UMC and GlobalFoundries have enough capacity to supply any increases in 28 nm demand without having to increase capital spending. UMC maintained its 2012 capital expenditure forecast of $2 billion several weeks ago, while GlobalFoundries, which plans to spend $3 billion this year, will probably have enough 28 nm capacity to serve additional customers after recently waiving an exclusivity agreement with key customer AMD. Previously, AMD was required to manufacture all x86 computer processors at GlobalFoundries, but a new wafer supply agreement between the two firms allows AMD to look to other foundries to produce those products.
Suppliers to Foundries Stand to Benefit
The increase in 28 nm-related equipment spending will provide a boost for the semiconductor equipment industry. Of the companies we cover, we believe process diagnostic and control tool supplier KLA-Tencor (KLAC) stands to benefit the most, given its exposure to technology-related investments. As foundries continue to increase cutting-edge 28 nm capacity, they will require new PDC tools to maximize manufacturing yields at the new process node. More important, we think foundries will be pressured to ramp up 28 nm manufacturing as quickly as possible because of the current shortages, which will make PDC even more critical. Foundries will need to invest heavily in PDC tools in order to be able to eliminate defects and increase yields quicker, which in turn will allow them to move to volume manufacturing on their new 28 nm production lines faster.
We believe other chip equipment winners will be firms with large exposures to TSMC and Samsung. The two firms combined to account for 39% of revenue at etch tool supplier Lam Research during the firm's fiscal 2011. TSMC and Samsung are also major customers of Applied Materials and contributed to 22% of total sales in Applied's fiscal 2011. The two companies made up 28% of 2011 revenue at Novellus, the deposition tool supplier that is being acquired by Lam.
Finally, though it is not an equipment supplier, we believe semiconductor fabrication materials and consumables supplier ATMI (ATMI) is well positioned to benefit from the stronger-than-expected demand for the 28 nm chip fabrication process at foundries and the additional 28 nm manufacturing capacity that will be put in place to alleviate current constraints. The firm stands to profit from the 28 nm manufacturing ramp through its high productivity development platform, which allows ATMI to create customized materials tailored to specific customer needs for their chip fabrication processes. The firm's HPD capability has become increasingly critical to enabling advances in semiconductor fabrication technologies, as chipmakers are becoming ever more dependent on new custom-developed materials to help them migrate down Moore's Law. The development and ramp of the 28 nm manufacturing process at foundries provided ATMI with its initial revenue from HPD materials in 2011, and we expect rising 28 nm manufacturing activity over 2012 to drive continued growth in this emerging opportunity for the firm. HPD-related revenue accounted for 4% of ATMI's revenue in 2011, and we currently project that it will reach 8% this year. However, with such robust demand at the foundries for 28 nm manufacturing, we think HPD sales at ATMI could exceed expectations.
Brian Colello does not own (actual or beneficial) shares in any of the securities mentioned above. Find out about Morningstar’s editorial policies.