​LCD panel prices continue to rise and expectations from panel producers and set producers continue to see the same for much of the 1st half.  That said, comments from panel producers also cite the fact that some are not producing at full capacity, despite demand, as component shortages are limiting their ability to deliver completed product.  That said, we still expect panel prices to increase this month, with notebook panels up between 2.0% and 2.5%, monitor panels up 1.0% to 1.7%, and TV panels to increase between 2.1% and 3.2%.  While we expect it might be hard to pin down a specific comment, there has been talk that panel producers are looking to see panel prices reach peak prices seen in 2017, however those prices have already been reached this year.
While we expect shortages to keep panel buyers under constraints to meet early 2021 quotas, demand is still being ‘regulated’ by the necessity for stay-at-home scenarios, a result of COVID-19, and while the virus is still at pandemic levels, vaccines and a more stringent public health requirements on a global basis, seem to be having at least some effect on hospitalizations.  With plans for most educational facilities, at least in the US, to be back to in=person classes by the fall, we expect demand to soften during the summer, but we also expect a lag between when demand slows and when buyers feel they can actually negotiate with panel suppliers given the desperation buyers have faced in past months.  Does that push out a slowdown until 3Q or does it happen during the summer?  That is a hard question to answer, but we will present four supply/demand scenarios later this week.  Without giving too much away, the four scenarios are titled, ‘slow burn’, ‘Black Diamond’, ‘New Normal’, and ‘Avalanche’.

​Coherent (COHR) announced that its board found the latest offer from II-VI (IIVI) ‘superior’ to previous offers from MKS (MKSI) and Lumentum (LITE).  The latest offer is valued at ~$6.5b, ahead of the $6b MKS offer and the original $5.7b offer from Lumentum.  This sets the stage for counter offers from LITE or MKS.  The II-VI offer is 8.3% above the MKS offer and 14% above the Lumentum offer.  Is there more to come or does the fat lady start singing?

Way back in June of 2019 we noted that a dispute between Japan and South Korea caused Japan to limit the flow of a necessary etchant (Hydrogen Fluoride) used in the production of semiconductors to be severely limited over reparations in reference to Japan’s forced labor of South Korean citizens during WWII.  The dispute went to the WTO and has been under ‘negotiation’ since then, with an occasional flare-up, usually a result of some new political issue between the two countries.  At the time of the note, we indicated that Samsung Electronics (005930.KS) had begun a program to diversity away from Japanese suppliers to protect itself from such situations and was imploring the South Korean government to help to develop local alternatives.
Now, almost two years later, Samsung is evaluating a local South Korean chemical company, Paik Kwang (001340.KS) as a local supplier of Hydrochloric Acid to reduce its dependence on Japan’s semiconductor etchants.  With Samsung using 80% of Korea'’ etchant output, it is obvious that their dependence remains, with Linde (LIN) of Germany and Japan’s Toagosei (4045.JP) being the company’s primary etchant suppliers, and while final approval has yet to be made, Paik Kwang has indicated that they are ready to mass produce the high quality HCl that is required.
Samsung is not the only one looking to disengage itself from Japan and other global suppliers, although late last year they made investments in four of the companies below totaling $65.4m, after a $97.2m investment earlier in the year in other MPE companies.  SK Hynix (000660.KS), the world’s 3rd largest semiconductor company is also looking to step back from its global supply chain and has been looking to replace a number of semiconductor equipment and material suppliers with local providers.  That said, this is not a task that happens overnight, and local producers, particularly chemical suppliers, have had problems with material purity, but are working with Samsung and others to expand and meet quality metrics although the qualification process can be a long one.  The table below shows some of the areas where South Korean suppliers are looking to become main sources of key semiconductor materials. 
Under the title of “if you can’t beat them buy them”, South Korean film manufacturer KNW (105330.KS) just completed the acquisition of Solvay Korea from Solvay SA (SOLB.BB), a provider of Fluorine and Sulfur Hexafluoride, for just under $50m US.  The six production sites in South Korea will now become local producers of a basic material used to produce Hydrogen Fluoride based semiconductor materials and materials for battery production.

The Apple (AAPL) store has indicated that the iMac Pro is now available ‘…while supplies last, “implying that it will no longer be produced.  The less than 4 year old top-of-the-line all-in-one Mac line, is a 27” high-resolution version of the tower or rack Mac Pro, which does not come with a monitor.  The iMac Pro can be configured with up to 18 cores and 256 GB of DDR4 memory, along with a maximum of 4TB of SSD storage and can support up to four 4K monitors.  The Mac Pro can be configured up to 28 cores, 1.5Tb of DDR4 memory with a maximum of 8 Tb of SSD storage, and can support eight 4K displays.  The Mac Pro base price is $5,999 while the iMac Pro base price is $4,999 and comes with an HD camera.
Video professionals, the target market for both devices, tended to look for power, and while the iMac Pro was certainly a powerful machine, the Mac Pro was the one that was continually upgraded by the company along with the more basic iMacs, which left the iMac Pro as what turned out to be an interim product.  There has been speculation that Apple is just clearing a path toward a new iMac based on Apple’s own silicon design, although most had expected such later rather than sooner, with speculation that Apple is closer to releasing its own Mac silicon sooner than expected.  That said, there are still iMac models available (just not the ‘Pro’) and most professionals would be likely to wait to see how well a new processor might work before jumping into a new iMac Pro, so there is actually little pressure on Apple to fill the iMac Pro gap quickly.  No official Apple event has been scheduled for this month but there has been speculation that Apple could hold an on-line event with updates for the iPad Pro.

Time of Flight sensing in smartphones has been around since 2018, when a number of Chinese smartphone brands included TOF sensors in their camera bundles to enhance imaging and to give applications the depth information they needed to use for animated video overlays.  In 2019 Samsung (005930.KS) picked up on the technology, including it in its Note 10, S10, and A80 series, while other brands such as Google’s (GOOG) Pixel and LG’s (066570.KS) G8 also added the technology, and last year, while Samsung seems to have soured on the technology, but the real believe in TOF has been Apple (AAPL), who introduced FACE ID in the iPhone X in November of 2017.
Apple’s FACE ID system however is a bit different from true TOF systems in that it paints a face with invisible dots and then maps those dots.  Each time a user accesses the device, the system repeats the process and then tries to match the new image against the initial image, and while this is a bit different from the TOF ranging systems used as part of smartphone camera systems, it certainly drew attention to 3D imaging.  In fact, there are really two kinds of 3D TOF systems, and again Apple’s system for 3D TOF is different from most others in the smartphone world.  Most smartphones and tablets that have TOF range sensing use the iTOF (indirect) method for ranging in which a laser emits light that is either pulsed or continuous but is modulated by a second signal.  When the light is reflected by objects, the system collects the light and calculated the difference in phase of the returned signal.  While the physics of this method are a bit complex[1], by calculating the difference in phase between the outgoing and incoming light, the distance of objects can be determined.
iTOF sensors are relatively inexpensive, which makes them attractive as adjunct depth sensors for smartphones with tight BOMs, but they do have some limitations, particularly their accuracy diminishes as the distance from an object increases, so an object 1m away from the camera will have an accuracy measurement of 10mm to 20mm, but when the object is 2m away, that accuracy drops to 40mm, and when the object is 5m away, following the same linear path, the accuracy of the distance measurement decreases to the point where it might affect the AR image overlay placement.
dTOF (direct) sensors, those used in Apple’s ‘LIDAR’ system, do not modulate the laser signal, nor do they measure the phase of the reflected signal.  They directly measure the time it takes for the laser to be reflected and based on the speed of light, calculate the distance.  This technique, which is the same as the types used in electronic range finders in professional cameras, does not have the same accuracy limitations as iTOF, and can accurately measure any distance, which is why the word LIDAR is associated with autonomous vehicle systems where an inaccurate distance measurement could make a big difference when the vehicle is moving at 27m/second (60 mph).  
dTOF also has some other advantages in that it requires less processing, making it faster, and is less prone to errors from object light scattering and objects with poor reflectivity, so why doesn’t everybody use it?  It is expensive.  Despite its simpler physics, dTOF requires a better light source (laser) and a more sensitive receiver (VCSEL), along with circuitry that is also more sophisticated than iTOF, and those requirements, particularly the sensitivity of the receiving sensor add to cost.  That said, sensor producers are constantly working toward higher sensitivity VCSEL array to make the cost structure of dTOF less onerous to smartphone brands that might need a cost conscious approach, but the basic question for TOF, in either mode, is whether it provides a benefit to the consumer, who is the ultimate arbiter for features.
We believe that the enthusiasm for TOF sensors seen in 2019 was an outgrowth of smartphone brand’s desire to compete by adding cameras, however there was and is a limit on how many cameras a user might want, so TOF was added as ‘another camera’.  The end user had little knowledge of what the TOF sensor could do, seeing filter effects and object overlays as things that software did, rather than actual imaging.  As dTOF becomes less expensive, we expect to see mobile devices begin to pick up the TOF pace again, particularly as Apple pushes dTOF into more products.  We expect Samsung to return to TOF once they are able to produce their own cost effective dTOF modules and sensor producer ams (AMS.AG) has shown a new high performance, low power integrated dTOF system that it expects to be in production later this year.


[1]
​

Foxconn Industrial Internet (601138.CH), a Foxconn spin-off that is focused on ‘industrial production’ and also functions as an arm of Foxconn’s massive device assembly facilities in China, signed an agreement with the Zhoukou Municipal government in Henan to build the Zhoukou Science & Technology Industrial Park, a $231m project that will add to Foxconn’s assembly facilities on the Mainland.  According to local media, the two phase project will employ up to 30,000 people when completed.  The expansion is said to be a part of Foxconn’s relationship with Apple, with Foxconn’s complex in Longhua Town, Shenzhen being the largest iPhone assembly facility.  That factory complex, known as “Foxconn City” covers over 1.2 mi2 and has 15 factories, dormitories, 4 swimming pools, its own TV network, a bank, grocery store, restaurants, and a hospital,, and is said to employ just under 500,000 workers, many of whom live in the dorms.
While we expect Foxconn is serious about building this new factory project, especially considering they have other facilities in nine Chinese cities, we harken back to the wonderful world of Foxconn in Wisconsin, where Foxconn is now promising to ‘revolutionize the automotive industry’ in cooperation with Fisker Inc (FSR), comparing their collaborative efforts with the work of Isaac Newton, and even hinting that they could even manufacture the cars in Wisconsin.  Foxconn’s massive project in Wisconsin, that served as a photo-op platform for former President Trump and then Wisconsin Governor Scott Walker has never established any kind of production, even failing to produce cutting edge coffee kiosks and ventilators during the early COVID-19 days, so we are always a bit skeptical when Foxconn makes promises, as are those 13,000 workers in Wisconsin who had been counting on the jobs promised by the company.
Aside from the 2,500 people who were forced to sell their property when the project was started (eminent domain), here’s what Foxconn has promised since the project was started, despite having built only a few small buildings on the site, which remain mostly empty:

• Gen 10.5 LCD fab – Never Built
• TV assembly facility – Never Built
• Foxconn North American Headquarters – Bought buildings in Milwaukee – No Employees – Now leased to others
• Innovation Centers – To be built across Wisconsin – Buildings bought in Green Bay, Racine, Madison, and Eau Claire – No employees
• $100m grant to University of Wisconsin in Madison - $700k has been granted
• Joint development of Ginseng farming in Wisconsin – No word since 2018
• Coffee Kiosks - Project dissolved
• Ventilator Production with Medtronic (MDT) – Nothing
• Google server assembly – Still waiting

In most of our notes on Micro-LEDs, we have focused on a particular aspect of the micro-LED process, that of transferring these extremely small LEDs from a die or carrier to a substrate, as the speed and accuracy of this part of the process is a large determinant in the overall cost of micro-LED systems.  In those notes we discussed a number of ways in which moving such vast numbers of micro-LED die could be accomplished, from simplistic pick & place techniques to highly sophisticated laser-based systems.  In those discussions, we noted that some systems were passive and some were active, while some required pre-processing and post-processing steps. 
Passive systems transfer die.  The transfer each die and have no way of knowing whether the die being transferred is good or bad.  In such truly passive systems, the die that have been transferred have to be individually tested on the final substrate and any die that do not meet spec have to be removed and replaced.  The time for the transfer, post-testing, individual die removal and replacement, and retesting, all contribute to the time it takes to produce a 100% fully functioning micro-LED display and obviously its cost.  Transfer processes such as stamps and roll-to-roll are such passive processes, and while they are far less expensive tools than active systems, they reworking that they require must be added to the total system cost, with the reminder that a typical 4K micro-LED display would have 24,883,200 LEDs that are between 50um and 100um in size, which makes them about the same size as a human hair.
Active transfer systems are different in that they require a way to discover which die do not meet spec before they are transferred, eliminating the expensive and time consuming rework process.  This is typically done in the semiconductor space with a probing tester that attaches to the leads of a chip or IC and tests for various electrical functions.  This is not a problem when you are testing a relatively small number of complex semiconductor devices, but not effective when you need to test millions of relatively simple but very small LEDs.  In order to create a KGD map (Known Good Die) LED tool vendors are adapting methodologies for testing such large numbers of LEDs using high resolution optical cameras, but even these, which typically focus on one die at a time, are to slow for such small die and such large numbers.  Newer systems that can view much larger clusters of die are becoming available to speed up the process, but even these systems are only looking at the physical characteristics of the LED die, which can recognize physical issues but not specification faults such as luminance or color issues on individual or groups of die.
Newer systems have attempted to take die mapping even further, using lasers to ‘stimulate’ each die, which causes each die to radiate a heat pattern that can be mapped.  While this does not give absolute specification data, it does show a completely different picture of the die, as shown in Fig. 1.  This data is converted into a ‘map’ of the die, which many of the active transfer tools can use to skip the transfer process for those die that have been excluded from the KGD list, which avoids the post-transfer rework steps.  Again, the time for creating the KGD map has to be added to the overall process time and cost of the micro-LED, and the fact that just ‘skipping’ the bad die during transfer can slow the process down must also be considered, but as these transfer tools continue to evolve, they reduce the cost of the process and bring the technology closer to becoming competitive with other display modalities.
 

Mini-LED backlights are a step above the traditional backlights used in LCD displays.  Typical backlights, which can be edge-lit or direct (see below), with edge-lit more typically used on entry-level displays.  In both instances an opaque ‘diffuser’ sits against (edge) or on top of (direct) to spread the light across the display.  Both types have their advantages, but high-end TVs must compete with OLED TVs that have extremely high contrast ratios that are a result of OLED’s control over every pixel individually.  Typical LCD backlights have a limited number of LEDs in their backlights, which gives them little image granularity and far less contrast.
Over the last year, backlight manufacturers have been working with chip suppliers to reduce the size of the LEDs used in LCD backlights, to the point where they are able to incorporate thousands of LEDs in a backlight array, rather than a few hundred.  This gives the set far more control over what areas of the image are dark or light, increasing the overall contrast of LCD displays.  These ‘mini-LED’ backlights are primarily used in high-end displays, but are moving down the CE display product chain into monitors and laptops, and while 2020 was the first year when mini-LED displays were commercialized, we expect 2021 to be a year when they become a featured part of displays and will provide a better image and competitive differentiation.
There are a number of panel producers that are offering LCD displays with mini-LED backlights, with Samsung Display (pvt), LG Display (LPL), AU Optronics (AUOTY), Innolux (3481.TT), BOE (200725.CH), and Chinastar (pvt) being the most visible, with some developing in-house product and others buying mini-LED arrays from LED producers/packagers Ennostar (3414.TT), Sanan (600703.CH), HC Semitek (300323.CH), and Seoul Semi (046890.KS).  As a relatively new business line for most, producers are working with a product that has no standards.  There is no standard mini-LED size, no standard LED count or no standard LED spacing, and that has kept it a more ‘customized’ business, with designers trying to balance all of the factors to meet specs from panel producers.
The cost to produce mini-LED backlights is high, between 30% and 50% higher than entry-level direct lit backlights, but still lower than the cost of equivalent OLED displays, but as mini-Led suppliers gain experience and refine process, the cost has been declining.  Ennostar, the holding company for Taiwan based LED producer Epistar (pvt) and packager Lextar (pvt), has been producing mini-LED backlighting since late 2018 and has developed a number of ‘standardized’ mini-LED backlights that are ‘stock’ rather than having to develop custom products for each customer. 
Driving the growth in mini-LED usage a number of applications where the high contrast or high brightness of mini-LEDs can justify their higher cost, and Ennostar has indicated that it has 2021 order visibility through 1H and is running at between 80% and 90% utilization, with 50% of its mini-LED production capacity booked by Apple (AAPL).  Apple is already using mini-LED backlighting, albeit at a relatively low LED count, in its XDR Pro Display Monitor, but a continuous stream of rumors about the company’s adoption of the technology at the notebook level seem progressively more realistic.  Chinese TV producer TCL (000100.CH) has already released a number of TV models using mini-LEDs and a number of laptops using the technology have already been released, particularly the ‘Creator’ series by MSI (2377.TT).
As LED producers and packagers expand mini-LED production, costs will decline and production bottlenecks will be eliminated (see our 1/19/21 note), driving mini-LED backlighting into mid-tier products.  That said, while TCL has received considerable technology press for its rapid deployment of mini-LEDs in its TV line, the adoption of the technology by Apple will have a very significant effect on its adoption across the industry.  If that comes this year, we expect estimates for $270m in mini-LED backlight sales will be low, if not, 2022 will see even greater mini-LED expansion, as it gives well-established LCD technology a bit more sustainability against OLED and other potential display technologies.  Given the massive LCD infrastructure that the industry has developed, LCD panel producers are usually want to adopt anything that will add to the longevity of their investment, especially one that uses much of the existing LED supply chain infrastructure, which mini-LED does.

​Early last month the highest court in France ruled that the provisions in the government’s 2019 “anti-Huawei (pvt) Law” were constitutional and held to the law’s aim of ‘safeguarding the interests of national defense and security and protecting radio networks from the risk of espionage, piracy, and sabotage.”  The law, which limited the number of licenses issued by governmental agencies for operating equipment produced by Huawei.   The ruling was the result of a challenge by two French telecom companies, SFR (pvt) and Bouygues Telecom (EN.FP) that faced replacing over 3,000 mobile base stations made by Huawei by 2028, with the government also indicating that it would not compensate carrier for losses or the cost of the changes, despite the mandate.
Under the law, carriers must dismantle Huawei equipment in densely populated areas during any network upgrades and demolition and replacement work has been initiated in a number of French cities, where Huawei 4G equipment is being removed. SFR is replacing the 4G equipment with 5G base stations and software from Nokia (NOK) equipment, while Bouygues is replacing with Ericsson (ERIC).  This all comes despite the early 2020 pledge by the French Foreign Minister that Huawei would not be prohibited from investing in France, with the government now stating that it hopes to find a ‘balance between sanctions and operations’ that would allow Huawei to act as a supplier while preventing the company from participating in a ‘more important part of France’s wireless infrastructure.’  Good luck with that!  Huawei set up an R&D center in Paris last October across the street from the Basilica of Sainte-Clotilde which we guess might become available real estate soon.

​As we noted previously, Chinese semiconductor supplier SMIC (688981.CH), a member of the infamous US ‘entities’ list, noted that some of its suppliers had been granted a license by the US Commerce Department, allowing them to sell components or tools that were made using US technology.  Among the most important was ASML (ASML), one of two global suppliers of lithography equipment necessary for production of all semiconductor products.  SMIC, the largest semiconductor producer in China, has been tasked with the goal of reducing or eliminating China’s dependence on US semiconductor process equipment and product, and to that end it has both capacity expansion plans and node size reductions as part of its timeline.
When the exemptions were given out, some made the assumption that SMIC would then be allowed to purchase any equipment made by ASML, including EUV tools that would allow SMIC to move to the 7nm and 5nm nodes to compete with Samsung and Taiwan Semi (TSM)., and the extension of SMIC’s contract with ASML through the end of this year, furthered that speculation.  That said, ASML has released clarification as to the extent of the license that enabled the SMIC contract extension, which indicated that the contract extension under the license, was for DUV lithography tools only.  By making that statement, it indicates that ASML and SMIC do not have an agreement on the more sophisticated EUV equipment needed for the 7nm/5nm nodes, with most now making the assumption that the US license specifically indicated the limitation on EUV tools.
SMIC has been moving its advanced production from 14nm to what it calls the N+1 7nm node, which it says is comparable to TSM’s 7nm node and does not require EUV, but the technology is still being developed and no word on whether it can be transferred to the 5nm has been revealed.  SMIC has little choice at this point, at least at the 7nm node and will likely continue to push its hone-grown (non-EUV) technology for as long as possible, while ASML is happy to supply DUV equipment to Chinese fabs, with 7nm production expected about 30 months out.  Until then SMIC stays at 14nm, which keeps them far from being competitive with Samsung and TSM, who are working toward less than 5nm nodes in roughly the same timeframe.