​Korea’s Public Procurement Service is tasked with managing the supply of materials for major building projects across the country, essentially a middle man with vast buying power that can be used by the government and small and medium sized businesses.  The service is responsible for over 30% of the total public procurement market in South Korea and almost 80% of public competitive contracts, so they wield considerable weight in negotiations with suppliers.  As they are responsible for government procurement, including electronics, they hold suppliers to a strict agreed to delivery schedule, and can impose penalties or sanctions against those that do not deliver against those agreements.
This week however, the PPS granted an extension requested by the Government Procurement Computer Association, a group that represents PC and similar CE product suppliers that bid on government CE contracts.  The extension had to do with past and potential delivery dates for PC related products that the GPCA felt could not be delivered on time, due to semiconductor and other component shortages.  The PPS granted and extension of 15 days to delivery dates in order to avoid penalties or future bidding restrictions based on current conditions in Korea. 
While the chairman of the GPCA stated that “In the procurement market, compliance with the delivery date is a strict regulation that is comparable to quality compliance,” the domestic procurement industry is also concerned that while the near-term effects of the component shortage have been dealt with, prices for basic, intermediate, and final CE products continue to rise.  Graphics cards were cited as one of the most affected (see our note – 3/18/21), with prices rising from 200,000 won (~$178 US) to between 500,000 and 600,000 won (~$445 - $535 US), but public procurement contract prices are set under budgets from the previous year, and changes in finished product prices are not easy to come by, which leaves the supplier to eat the cost of the BOM increases, which, according to the PPS, is causing some end-user products to face limited public availability.  While what is noted here is specific to South Korea, we expect similar situations to become more apparent if the semiconductor shortages persist through the remainder of this year.

Chinese OLED panel producer BOE (200725.CH) has had a rocky path with Apple (AAPL).  While the company supplies OLED panels to a number of Chinese smartphone brands, BOE has been attempting to break into the iPhone supply chain, which has been dominated by Samsung Display (pvt) and LG Display (LPL).  Apple’s display criteria, while they might not be the most advanced from a technical standpoint, are quite stringent, with a particular emphasis on yield, given release delays that Apple has faced in the past. 
Last year’s late iPhone releases were a result of the effects of COVID-19, at least according to the company, but Apple has faced release delays in other years due to quality issues, even with Samsung, the leader in small panel OLED flexible displays.  For a time LG Display was relegated to supplying replacement displays for the iPhone, until they were able to prove to Apple that they fully capable of meeting Apple’s technology and volume requirements.
BOE has been in a similar position, supplying replacement displays for the current iPhone, but has been looking to graduate to becoming a full supplier.  Last year BOE was evaluated at least twice, and while they were able to join the Apple iPhone display supply chain as a tertiary supplier, they were not qualified by Apple to be a primary supplier for the current iPhone 12.  There have been a number of comments made by both Chinese and Korean press, indicating that BOE had or had not been qualified by Apple as a primary supplier after further review, with the pendulum leaning toward Apple’s acceptance of BOE for the upcoming iPhone 13, however Korean trade press has indicated that such is not the case, at least this far into the year, meaning that BOE will not be a primary iPhone 13 OLED display supplier.
Mass production for the iPhone 13 is expected to begin next month, and while this does not completely rule out BOE, who could get qualified later this year, it certainly compresses their ability to garner any meaningful volume this year.  Since Apple does not indicate which display suppliers will be producing each iPhone model, or what the size, mix, and features will be for the iPhone 13, we have to make some assumptions as to how the iPhone 13 series will play out this year.
Expectations are that Apple will not change the size of the displays for the iPhone 13 from those of the iPhone 12, which would indicate a 5.4” iPhone Mini, a 6.1” iPhone 13 and iPhone 13 Pro, and a 6.7” iPhone 13 Pro Max, but there are expectations for other changes.  In the iPhone 12 series, touch was embedded in the OLED display for the iPhone 12 Pro and Pro Max, while the Mini and the iPhone 12 used an attached touch system.  We expect that all four, or at the least three of four models this year wil have embedded touch, with the Mini being a possible exception to keep BOM as low as possible. 
We also expect that the iPhone 13 Pro and Pro Max will adopt a refresh rate of 120 Hz., and with that feature will be a TFT structure based on LTPO (Low temperature Poly-Oxide), and alternative to the more standard LTPS (Low temperature Poly-Silicon).  As the higher refresh rate would tax the battery, as the screen is redrawn twice as often, the use of LTPO would offset some of that battery drain, as it is more power efficient.  LTPO is a relatively new TFT process and could be one contributing factor in the assumed decision by Apple concerning BOE, although that would only rule out BOE for the two higher-end iPhones.  We suspect such a decision by Apple, would have been based on BOE’s ability concerning overall yield, rather than a single feature.
Current expectations are that Samsung Display will supply between 120m and 130m OLED displays for the iPhone 13, with LG Display supplying 50m, a big improvement over last year’s ~20m units, for a total of between 170m and 180m by the end of the year, and based on these expectations we ran an analysis of Samsung and LG’s capacity to gain some insight as to what the production schedule for the iPhone 13 might look like this year.  In order to do so, we have to make some basic assumptions concerning capacity allocation, yield, mix, and even whether the iPhone Mini will play a significant role in the iPhone 13 series volumes, given its relatively poor showing so far this year.
We make the assumption t5hat Samsung Display has dedicated the entire output of its A3 OLED fab to production for the iPhone, but based on our yield assumptions, our model would indicate that Samsung would fall short of its expected goal.  That said, we know that Samsung has been working toward refitting a portion of its A4 OLED fab to produce LTPO displays, which when included, would allow Samsung to meet those production expectations.  LG Display, based on capacity at its E5 and E6 fabs should be able to meet current target expectations.
Again, much will depend on yields, even for these two more experienced suppliers, and mix, which, while Apple will set based on its sales expectations, could change based on consumer reaction to model prices.  We do build in a relatively small change in mix in 4Q, with less Mini and more Pro and Pro Max, but that is a variable that could change earlier based on initial sales.  If Apple actually maintains current volume expectations for iPhone 13, it will represent a very substantial part of LG Display’s overall flexible OLED capacity for the year, and while less so for Samsung, we expect it will keep Samsung’s overall small panel utilization rate at a high level for much of the year.  Given a lower yield at the onset of production, the impact to profitability for both would be lower in 2Q, rising in the 2nd half, and we expect that the unit price of the displays, when compared to the combined cost of touch and display, will be a bit lower or the same for Apple, while the overall profitability for both suppliers will be higher, having captured 100% of the touch revenue.

Moving day is an intimidating process.  Dozens upon dozens of boxes filled with your stuff, some of which dates back to High School or even earlier.  There is some very important and sentimental stuff in those boxes. Your Farah Fawcett poster from your old room, your Magic the Gathering card collection, your signed “In-A-Gadda-Da-Vida” album from that night at The Whisky, and some of your wife’s less important stuff.  The guys tossing those treasures into a van don’t grasp the significance of your stuff and are talking about some game they saw on the weekend while your stuff is carried around as if it were meaningless trash.
Just think what it would be like if you had to move 24,883,200 boxes, and each box was between the size of a red blood cell and a human hair?  The good news is that Felix the mover could pick up thousands of those tiny boxes at one time and load them on the van, of course, he would have to be able to see them and not destroy them when he picked them up or when he put them in the truck, or when he took them out of the truck into their new home, but that is exactly what has to happen when constructing a micro-LED display.  

We did not just make up that 24.88m number, it’s the number of red, green, and blue micro-LEDs that would make up a full color 4K micro-LED display, and each of those 24.88m LEDs, is produced on one of three sapphire substrates, has to be picked up and placed on a substrate that will be the basis for the micro-LED display.  So whose tiny fingers can do such a job?  That’s a good question because there are a lot of companies trying to figure out ways in which to move these tiny LEDs, and revolutionize the display space.
The reason there is so much activity in the micro-LED space is that it has the potential to become a dominant display technology that could one day be as pervasive as LCD technology that began replacing CRT (Cathode Ray Tube) displays in the 1980’s.  Notice we say ‘potential’ as there are some very difficult technology hurdles that have to be met before micro-LED technology can become more than a showroom oddity.
Aside from the necessity to solve a wide variety of technical issues, micro-LEDs do show promise, at least in a less than mass production setting, and compare favorably with both LCD and OLED technology, as shown in the table below, although we note that the source and circumstances of measurement of these characteristics can change the values considerably, and we note that brightness, measured on a cd/m2 basis is not a good indicator as to the viewers perception of a display’s brightness.  That said, as a general comparison, it gives some idea why display manufacturers are willing to spend time and money to develop the potential for micro-LEDs.

Returning to our original thoughts about moving such small chips from wafers to a substrate, there are some other technical challenges that are keeping micro-LEDs from becoming a mass market product.  As mentioned, the LEDs are small, but they are also of three different colors, red, green, and blue, and the materials used to make those LEDs are different, at least for red LEDs, which are based on Indium Gallium Phosphide (AlInGaP), while blue and green LEDs are both based on Indium Gallium Nitride (InGaN)..  This means that each color LED has to be grown on a sapphire substrate in an MOCVD (Metal Oxide Chemical Vapor Deposition) tool, and then sliced into tiny die, and transferred and connected to a display substrate in a pattern that is typically horizontally spaced.  The way in which this is accomplished is currently the biggest puzzle in the micro-LED saga, with almost as many ‘flavors’ as Baskin-Robbins (pvt), and we note the ‘connect’ part, which means that not only do the micro-LEDs need to be placed in RGB sequence, but the spacing and position much match the TFT driving circuitry to which each micro-LED is connected.
Typically larger LEDs can be connected to circuity via a metal wire that is soldered to a pad of conductive material, however that option is not available due to the large amount of substrate real estate it would consume, reducing the possible number of LEDs, along with the potential issues of what might happen to such thin metal wires during transfer and soldering.  A more common method for connecting smaller LEDs would be flip-chip bonding, where a contact on the LED is bonded to a ‘bump’ of material that is in-turn bonded to the TFT driving circuit.  This is certainly a more effective way to attach micro-LEDs to a substrate and TFT driver, but remember that this must be done almost 25m times for each display, and using current bonding tools, this can take many hours or even days, making the cost of production astronomical.

So, regardless of the speed at which micro-LEDs can be transferred, display engineers must consider the absolute accuracy of LED placement, which compounds the complexity of the transfer issue.  But wait, there’s more…   Even at the high rates of speed that specially designed pick and place tools can move micro-LEDs, there are some ‘in-between’ steps that complicate the process further.  We noted that each of the three micro-LED colors are grown on a sapphire wafer, and while their small size allows for hundreds of thousands to be produced on a single wafer, each of those ‘smaller than the thickness of a hair’ chips needs to be ‘singulated’ (cut) from the wafer and must remain upright and in position to be ready for the transfer process.  In some processes the micro-LEDs are transferred to a ‘tape’ that can be fed into a mechanical transfer system that places them on the display, but the time it takes to move the individual die to the tape must be included in the total production time, and that can be a time consuming intermediate process.
One way micro-LED engineers have come up with to eliminate that intermediate step is by using an Elastomer Stamp, a process championed by X-Celprint (pvt), that is said to be able to transfer 1m die/hr.  The stamp is commonly made of PDMS (Polydimethylsiloxane), a silicon-based polymer that is used in shampoo to make your hair shiny.  This material, or something similarly ‘sticky’ is pressed against a section of the singulated wafer and picks up hundreds of micro-LED die at once.  The Stamp then moves to the display, where the stamp is slowly peeled back, leaving the micro-LEDs on the display substrate.  This is an effective way to move large numbers of micro-LEDs, however the die must be pre-arranged to both match the receiving TFT pattern, and if it is an RGB display, they must be spaced to leave room for the two other color micro-LEDs, which then need to be transferred separately.

​A small Korean start-up LC Square (pvt) has come up with what it believes is a way to modify the elastomer stamp process that will speed up and improve the accuracy of the transfer process.   The die are removed from the wafer ‘en masse’ via laser lift-off and transferred to another wafer where the laser selects only those chips that match the correct location on what the company calls an ‘interposer’ with the process repeated for the other two colors.  This creates a RGB pattern on the interposer which matches the TFT driver pattern on the display substrate.
The company then applies an ACF (Anisotropic Conductive Film) between the interposer and the display substrate and through heat and pressure, the ACF becomes the bond between the RGB patterned interposer and the TFT substrate, which allows the entire interposer to be placed at one time with all of the micro-LEDs in place and connected.  LC Square provides either the three RGB interposers with chips, or the final interposer with all three colors placed, along with the ACF so it can be transferred by the panel producer.  Of course, the practicality of such a process will depend on the cost, the metric all other transfer techniques are judged by, but the company insists that it can separate 1.2m micro-LED chips from wafers and create an interposer in 4 minutes, with a placement accuracy of +/-2µm at a yield of 99.99%. 
One possible solution, interim or otherwise, is to use a single color blue micro-LED for all sub-pixels and use quantum dots to convert blue to red and green in 2/3 of the sub-pixels.  While this is similar to the way LG Display (LPL) produces large panel OLED TVs, the use of quantum dots would eliminate the brightness loss that reduces OLED TV brightness when using phosphor based color filters.  Quantum dots do not block other colors as phosphors do, but shift the wavelength (color) of the light in a highly efficient manner, and the creation of a quantum dot color filter could be accomplished using ink-jet or similar mass production technology.  Such an idea solves the problem of selective placement of RGB micro-LEDs, but still leaves a big open issue as noted below.
Again we note that there are quite a few other technologies being explored in relation to producing micro-LED displays, but our look at just one technology and a variant, points to not only how much research is being done in the area of micro-LEDs, but just a few of the challenges facing the commercialization of the technology.  Each new system has to face not only the question of speed but the even bigger question of what to do with those micro-LEDs that do not work once they have been placed.  Even with an accuracy of 99.99%, a 4K display would have 2,489 micro-LEDs that do not work and would have to be removed and replaced, a far more arduous task than the initial placement, which would likely cost as much or more than the transfer process itself.  More to come.
 

DRAM[1] prices jumped 26.7% in April on an monthly basis (average price on 4/29/21), while NAND Flash[2] remained flat for the month. The DRAM increase was the highest since early 2017 as potential shortages and increasing demand for notebook PCs has encouraged PC component buyers to increase orders to ensure stock in 2Q, when inventory levels are expected to decline on increased notebook orders and server demand.  Current expectations are that DRAM prices will rise again in 3Q as inventories are worked down and sales during the holiday season are anticipated.
 


[1] DDR4 8Gb 1Gx8 2133MHz
 

[2] NAND Flash 128Gb 16Gx8 Multi-Level
 

​Last year’s iPhone 12 series saw both a delayed announce date of October 13 (typically September 20) and a further delayed release date for the iPhone 12 Mni and the Pro Max, which were released on November 13, 2020, 20 days after the iPhone 12 and iPhone 12 Pro.  While the December quarter was still a record for overall Apple (AAPL) iPhone sales, which included previous models, a normal selling period could have boosted sales further.  We expect this year’s iPhone 13 will be back to the typical release date and possibly sooner.
The delay last year was based on a suppose number of issues, with Taiwan Semiconductor, Apple’s foundry for the A14 bionic processor that is the basis for the iPhone 12 family, and last minute 5G antennae module changes, which meant requalification.  This year however, while there have been rumors that Apple has reduced some component orders during 2Q, for the existing iPhone, there have been a number of component suppliers who have hinted that they are getting ready to begin component production in June , which would indicate a return to Apple’s more normal iPhone release schedule.
Some of the component suppliers have indicated upcoming changes to cameras and it is well expected that some of the new iPhones will have a 120Hz. refresh rate, which implies an LTPO backplane, supplied by Samsung Display, LG Display, and possibly BOE (200725.CH), but most important would be TSM’s production schedule for the A15, Apple’s upgrade to the current A14 processor.  Apple has contracted much of TSM’s 5nm capacity and expectations are for processor production to begin in May, which would fit with a more normal release schedule.  While we have seen stories about how any number of potential component suppliers are gearing up to become part of the iPhone 13 supply chain, we take those more as expectations rather than actual order based expansion, but all in, it seems that Apple is back on schedule for a normal iPhone release this year and if they can produce as popular a device as it seems the iPhone 12 seems to be, they could set some new records again this year.