According to Semi.org, using data from its 2,400 members, 19 new semiconductor fabs will begin construction this year with an additional 10 beginning construction next year.  Of those 29 new fabs, 22 will be 300mm lines and 7 will range from 100mm to 200mm.  The cost of these new fabs will be ~$140b and will add up to 2.6m 200mm wpm.  China and Taiwan will see 8 new fabs each, while 6 will be built in the US, Europe with 3 and both Japan and South Korea having 2 each.  Semi also notes that there are 8 additional lower probability fab projects that could be added.
15 of the new fabs will be foundries with capacity ranging from 30,000 wpm to 200,000 wpm and 4 will be memory fabs with capacity ranging from 100,000 wpm to 400,000 (200mm equiv.)  As the time frame for new fab construction is two years from groundbreaking to equipment install, even those beginning construction this year will likely not be producing until 2023, and we expect each project will have developed its own justification (automotive, telecom, etc) for the expansion.  We show semiconductor fab equipment spending over the last 17 years in the chart below.

Did you ever dream of becoming a consumer electronics product designer?  Probably not, but you can become one in your spare time for a mere $959, and stun the world with a flexible display application that will make millions, or at least that is what flexible OLED display producer Royole (pvt) is hoping with its just released RoKit, a ‘developer friendly toolkit that the company is hawking along with its foldable Flex-Pai 2 Folding Smartphone ($1,082, down from $1,344), the RoWrite Notepad ($99), and of course, the flexible OLED Tee-Shirt or top hat, that allows you to stream content, making you a walking billboard.

The RoKit comes in a 14” x 10” Aluminum carrying case (with handle), which means you would be able to bring the development kit to work and continue your development project during your lunch break, and it contains a fully flexible 3rd generation 7.8” foldable OLED display with touch support, a 7.8” sensor module, an Android 10 development motherboard with Qualcomm (QCOM) Snapdragon SDA660 CPU, an HDMI adapter board, Wi-Fi antenna, microphone board, optical sensing board, and power converter, along with associated cables, so you can prototype your device quickly and show it to friends and colleagues anywhere. 

There are some caveats however, as the bending radius for your RoKit is 15mm (~0.6”), so don’t go overboard on folding your device and follow the advice given in the 7 page user manual, which shows how to connect the boards and ways in which you can fold the display without breaking it.  Once you’ve got it all together, you are on your way to developing the latest and greatest flexible/foldable device, which you can submit to Royole’s ‘Flexible Future’ innovation design competition, which has produced award-winning designs such as the ‘smart helmet with interactive display’ or the ‘rotating screen application’

​Understandably Royole wants folks to develop applications using their flexible/foldable displays, and while rarely do display producers offer ‘developer kits in a case’ to the public, they believe that offering such will encourage the CE development community to come up with applications that will increase demand for Royole’s particular display offerings.  While the company was the first to offer a foldable display to the public (really a developer kit of sorts), there has been little indication that its products have been selling well and in February the Shanghai Stock Exchange terminated Royole’s IPO application after review.  A detailed review of the prospectus by an outside company revealed many questions about customers and product returns, and while that report cannot be directly tied to the IPO termination, there were likely too many unanswered questions in the prospectus for the exchange to approve the deal, so its back to new product development for Royole and anyone with a RoKit.

​Corning (GLW) announced this morning that it will be instituting a moderate price increase for display glass substrates in the 3rd quarter.  According to the release, “the adjustment to price is a result of elevated costs in logistics, energy, raw materials, and other operational expenses during this extended period of glass shortage, as well as general inflationary pressures impacting the display industry.”  The company also mentioned that it expects glass supply to remain short to tight in upcoming quarters.  Typical ingredients for borosilicate glass are Silica (SiO2), Boric Oxide (B2O3), Soda Ash (NaCO3), and Alumina (Al2O3) and are combined at higher temperatures than typical soda-lime glass, raising the cost.  While pricing on the various components is difficult due to quality and gradation, the cost of electricity produced from natural gas has risen from $2.23 last March to $3.41 in March of this year, a 52.9% increase, and while that does not reflect actual cost to Corning’s display glass production business, it is certainly a factor that would push Corning to incorporate those types of cost increases in its pricing negotiations with customers.
While we cannot remember a time when Corning was able to raise glass prices, it might have happened, however typically, glass prices decline each quarter, although over the last few quarters Corning has been able to hold those declines to small amounts or no change.  Given that supply problems at competitive suppliers and continued demand for displays has kept the glass market tight, it is no surprise that Corning has established price increases for 3Q, when demand will likely be the highest.  As no data was given on the magnitude of the increases, we surmise that Corning Is able to cover the cost increases that it has absorbed over the last few months and likely a bit more.  However,  while the display division’s objective is to increase profits, expanding the company’s customer penetration is key to the longer-term goal of generating returns on the capacity investments Corning has made for its customers, and holding them for ransom with egregious price increases is likely not Corning’s goal in this case.

​While the quarter is not quite over, it looks like US TV sales are heading down and the channel fill that helped to boost March sales does not look like it will materialize this quarter.  Expectations for June are for ~2.3m units, which would put US TV sales for 2Q at just under 7.8m units, down 24.2% q/q   Channel purchases are expected to be ~8.7m units for the quarter while sales are expected to be 7.8m, leaving near-term inventory of .9m units.  Last quarter channel purchases were ~9.3m units, while sales were 10.3m, leading to what we estimate was between .3m and .3m in incremental channel purchases in April.
That said, if our expectations for June are correct, the Q monthly sales average has declined to 2.59m from 3.42m units/month in 1Q.   Some of the strength in 1Q can be attributed to stimulus checks but we believe the US vaccination program has allows consumers to return to at least a semblance of what was a normal lifestyle, which includes being out of the house and away from a television, lessening the need to replace, upgrade, or enlarge existing installations.  Should the next two months continue that lower trend, if would signal a return to pre-pandemic levels, and while 3Q and 4Q are typically the seasonally better quarters, this holiday season could prove a bit difficult given the rapidly rising prices of TV display panels and components.
TV set manufacturers and retailer seem to be becoming a bit more realistic about their expectations for the full year, while IT product producers are still looking for the rocket ride to continue.  There are some signs that notebook demand is slowing, but it is hard to discern whether it is demand driven or supply limited due to component shortages, but as we head into the summer months and a bit of normalcy returns, at least in the US, we should get a better understanding of the split between the two.  Other regions that do not have the luxury of the high availability of current COVID-19 vaccines might take longer to return to patterns seen pre-COVID, but once the data confirms the trends in the US, it will likely set the tone for other regions.
 

On March 19 in the N3 building of the Renesas (6723.JP) Naka factory caught on fire.  The fire broke out on the company’s 300mm production line and was later extinguished by the local fire department who determined that the fire was caused by and electrical problem with one of the plating tools on the line that overheated and burst into flames.  The damaged area was ~600m2, or ~5% of the total cleanroom area, and while originally 11 units were thought to have been damaged, the total was upped to 23 at a later date, which was ~4.2% of the equipment in the N3 building.  The company obviously halted production at N3 but was able to continue the 200mm line in the N2 building and the wafer testing operation.  At the time, the company estimated that the impact on the company’s production would be ~17b yen ($156m US) per month.
One month late Renesas management set expectations for the reopening of the N3 line at ~50% capacity by the end of April and full production by the end of May with full shipping volume about a month later.  The Japanese government also pushed tool manufacturers to accelerate delivery schedules so the fab could reopen quickly as the Naka plant produced ~33% of automotive semiconductors, particularly MCUs, which already were in short supply before the fire.  It seems that things could take a bit longer with Renesas management indicating that 88% of the equipment to be replaced has been installed and is in operation as of mid-June and that full production levels will be reached by the end of this month, although questions remain as to when the impact will be felt by its customers. 
We give credit to Renesas management in that they came close to their original target of 100 days, but we expect real full scale MCU or similar product production is still in the preliminary stages, with some process steps available but full line availability still developing.  Once the line is in full operation, which we would expect in early July, it will take between two and three weeks, depending on the product, to begin volume shipping to customers, which would mean early August would be when customers would begin to receive product, other than small volumes.  Much will depend on yield, as tools will need to be tuned and glitches removed, so the positive impact on supply could still lag a bit in August according to supplier sources.
 
 

​While not the first time such a deal has been mentioned, the Korea Times has posted a story suggesting that Samsung Electronics (005930.KS) is close to ordering between 3m and 4m OLED display panels from LG Display (LPL), although a Samsung executive is quoted as saying nothing has been decided at this moment, and after a report that the two had met a few months back to discuss such an agreement, Samsung has denied that the meeting took place.   Samsung’s display affiliate, Samsung Display (pvt) has been developing a quantum dot/OLED display production process, for which it has been sampling what are likely pre-production level panels to both its parent and other potential customers, but no date has been given as to when SDC will go into full scale production and such panels will appear.
That said, Samsung Display is converting its L7-2 LCD fab into a Gen 6 OLED line and is planning the conversion of its L8-1 LCD line into a Gen 8.5 OLED line, which would be used for monitor and notebook OLED panels, and would be the only Samsung Gen 8.5 OED production line.  Gen 8.5/8.6 lines however are not particularly efficient above 55” so the odds that Samsung would choose a Gen 8.5 fab configuration and produce large OLED panels (>55”) seems odd, and producing monitor and notebook panels on such a line makes more sense.
But the real question is why Samsung would choose to add OLED to its mid-priced TV line, using WOLED technology (White OLED with Color Filter) when it has repeatedly trashed the concept?  While the Samsung Electronics marketing organization can surely find a way to company-speak around that problem, the only reason would be to capture premium where such premium does not exist, and Samsung seems to already have staked out quantum dots and a combination of QDs and mini-LEDs for its high-end lines.  This leaves mid-priced TV sets, which would be straight LCD or QD/LCD as the most competitive point in the Samsung TV line, and Samsung having to buy much of its LCD display product from other producers does not give it any pricing leverage.
Adding a mid-range OLED line would give the company a bit of potential margin improvement, but again they would have a cost disadvantage to LG Electronics (066570.KS), the parent of LG Display and if such a deal were to occur, the delivery timeline would be spread out over a number of quarters, unless LG Display was willing to dedicate some of its large panel capacity to Samsung, knowing that they could abandon the WOLED concept if SDC is successful in developing the QD/OLED product.  It all seems a bit implausible, but stranger things have happened in the display space.

5G gets a lot of press.  If we are not hearing about new 5G smartphones, we are certainly bombarded with maps and statistics about how this carrier or that carrier has the best  5G network, but the question remains, where are we with 5G on a global basis, and where will we be in 5 years?  While the average guy/gal in the street probably still does not care about 5G, as they probably have a Samsung (005930.KS) Galaxy S8 (released 4/21/2017) or an Apple (AAPL) iPhone 7 (released 9/12/16 here in the US, or a Huawei (pvt) Mate 20 Pro (released 10/16/2017) in China, but they will over the next few years when they finally get frustrated with apps that run slow, dropped connections, battery deterioration, or they are just embarrassed to be using a phone that is almost 5 years old.  So the question remains, where are we now and where will we be with 5G in five years?
Ericsson (ERIC) puts out a summary of facts and figures concerning 5G and broadband generally each year and while it certainly has an optimistic point of view, not surprisingly so give Ericsson is among the top five 5G equipment vendors, but it does have extensive 5G data and some forecasts as to where the company expects 5G to be in a number of years, some of which we summarize here:
5G subscriptions have been growing at a very high rate, although we note that prior to 2019 there was no real 5G service available so 2020 growth is going to reflect the first and second real year for 5G subscriptions, and for most areas, just the beginning of real deployment.  On a global basis, worldwide mobile subscriptions in 2019 were 7.93b, increasing by only 0.25% in 2020 to 7.95b, but are expected to grow by 10.4% during the period between 2020 and 2026 to 8.778b at a 2% CAGR.  Here’s how it breaks down by region, including China and the Gulf Cooperative Council, which consists of Bahrain, Kuwait, Oman, Qatar. Saudi Arabia, and the United Arab Emirates.  China and the GCC are also broken out separately:

Looking at subscription data by technology shows how each technology fares in terms of subscription growth over the same 2020 – 2026 time period.  As we are using rounded figures for calculating absolute growth and CAGR, our numbers differ slightly from those in the report, however the only significant difference of note is with 5G, where we calculate a 72.9% CAGR on rounded numbers and Ericsson calculates a 59% CAGR on rounded numbers.  We note that while 2G and more recently 3G networks are being eliminated in the US, together they still represented 38.4% of subscriptions globally last year, and even in 2026 are still expected to represent 15% of global subscriptions, so while the technology might not be visible here in 2026, it will still be part of the global network.
5G itself is expected to grow faster than 4G, reaching 1b subscriptions roughly two years earlier than 4G, which began its rollout in 2009, although it will not see a greater share of global subscriptions than 4G (LTE) until 2027 or 2028.  That said, even using Ericsson’s 59% CAGR through 2026, 5G will see very substantial growth in subscriptions.  Breaking down the 5G subscription growth by region, while Latin America will see the most rapid subscription growth during the period, North East Asia, primarily China will see it share decline from 88.0% in 2020 to 40.6% by 2026.  The 5G subscription share in 2020 and 2026 are depicted in Fig.1 and Fig. 2 below.

We have mentioned how the rise in bitcoin prices has caused the price of graphic cards to increase, as the GPUs in these cards are well suited form bitcoin mining, however a recent bitcoin mining crackdown in China, specifically the provinces of Inner Mongolia, Xinjiang, and most recently Sichuan, has not only caused the price of bitcoin to drop from its April 13 peak of 65,503 to its current level of 32,410, but has also caused the price of graphic cards to decline.  While the price of bitcoin is still 10.3% higher than on 1/1/21, it is almost 49% lower than the peak and the notification from two (Energy & Science) bureaus to crypto-miners in Ya’an City in Sichuan that they must cease operations for inspections, indicated a change in policy for the region, the 3rd to do so.  While power in Ya’an is primarily hydroelectric, making it less costly and environmentally damaging than fossil fuel, bitcoin mining uses about 10% of the power generated in the region.  The shutdown begins on June 25 and does not have an end date as yet.
The effect of these crackdowns, which have been made under the title of the government’s plan to reduce the risk of loss to investors and prevent money laundering, but also reflects China’s inability to come to terms with cryptocurrency generally given its unregulated environment, are expected to shift bitcoin mining away from China to areas where both low-cost power is available and regulation is at a minimum.  China’s state-run TV has stated that bitcoin is ‘no longer an investment to avoid risk.  Rather, it is a speculative instrument’, and added that it is a ‘lightly-regulated asset often used in black market trade, money-laundering, arms smuggling, gambling and drug dealings,’ leaving little room for discussion, at least for now.
We took a sample of graphic card processors on Amazon (AMZN), and while there are only 8 items in this sample, it seems to be quite representative of the impact of the crackdowns in China.  While the cards are produced by various manufacturers, they are based on Nvidia (NVDA) and AMD (AMD) GPUs with little price difference between cards base on the same processor.  The average change from the peak is 21.6% and the average spread is 84.2%, and while the price reductions make it seem that graphic cards are a bargain today, the average current price in our sample is still 30.7% above the average low price.

​As we noted last week smartphone shipments in China were weak during the Labor Day holiday and generally in May, and if June sales are up ~10%, which would be the average of the previous two months, 2Q in China would see shipments fall 22.7% from the 1st quarter and be down 30.6% y/y. While we do expect smartphone sales to be up m/m in June, Chinese smartphone brands have remained overly optimistic this year, and after weaker than expected holiday sales, even with June being up over 10% m/m, most Chinese smartphone brands will likely see sales decline at least 10% to 15% and possibly more if June does not live up to expectations.  We expect Chinese smartphone brands to lower full year targets again in July if that is the case.
As Chinese smartphone brands face a weaker 1H, they have less of a need for components, many of which have been in short supply, and while this could help to reduce some of the shortages, demand related issues reducing shortages are far less positive for the industry than if supply was increased.  Unfortunately capacity increases are more of a long-term process, and with some component suppliers seeing less demand from China in 2Q, they might be a bit more cautious as to how fast and by how much to expand, especially given the time it takes to build out capacity.  To illustrate the point, Taiyo Yuden (6976.JP), a major producer of MLCCs (Multi-layer Ceramic Capacitors), a component that has been in short supply, recently announced that it is committing $45m US (construction only) to build a new MLCC plant, with construction set to start in September.  While this will add to capacity the new plant will not be ready for production until the end of 2023, doing little to alleviate near-term shortages.  MLCC demand has been growing as they are necessary for electric vehicles, but demand has also been spurred by their increased use in 5G smartphones, and that demand is quite volatile. .
That said, the same suppliers that are seeing lower component orders from Chinese smartphone brands are also expecting production for the iPhone 13 to begin in July, hopefully offsetting any slowdowns, so there has been little change in component pricing thus far.  Given the lead time for new component capacity and the even longer lead time for semiconductor capacity expansion, it will likely take more than a slowdown in Chinese smartphone sales to loosen the component shortage deadlock and slow upward pricing momentum or actually bring down component prices, but in the CE space that theory is only good until the day when buyers postpone orders expecting a lower price the next day.. 
One material where we have seen some relaxation of pricing momentum is copper, and while the commodity is still up over 35% this year, it is down over 13% from its high about a month ago.  While there are many factors that drive copper prices, it is a key material for electronics with its inclusion in everything from wire to PCB boards, and almost any material related to electronics, particularly those that have automotive applications, pushed prices higher up until recently.  While copper for wire is the material’s biggest application, which is likely more related to housing and construction than electronics, we expect overall demand to flatten a bit and then pick up at the end of the summer as new products for the holidays begin production, and while off its peak, we expect the raw material to remain in demand, at least from the electronics sector. 

There has been much talk, ours included, about TV set prices, especially since the beginning of the year, as large panel prices have been rising steadily since last June and have now exceeded peak yearly TV panel prices by between 26.9% and 56.0% going back to 2016.  That said, talk is cheap and consumers face a harder reality when they walk into a store or go on-line to purchase a TV.  As we get closer to one of the US’ ‘shopping’ holidays, Prime Day, we ran through our list of TVs to see how pricing has fared since we did our original survey (11/16/20), the last survey (4/10/21) and current prices.  As we have noted in the past, we started our survey by detailing all of the 68 4K TVs that were available on the Best Buy (BBY) website, although we noted then that 18 of the 68 (26.5%) were actually not available, despite their listing, and over time we noted that a number of those listed units never returned, although some remain on the site to this day.
Here’s what we found currently…
Of the original 68 models that appeared on the website, of which 58 (85.3%) were available at the time, only 34, or exactly half, are still available on the site.  Given that new models appear and older model lots are sold out, this is not surprising considering 215 days have passed, however what is surprising is that of the 34 models that remain from the original group, 16 or 47.1% are not available locally.  This means that of the original 68 models listed on the website only 18 (26.5%) would be available to a customer wishing to purchase one of those units. 
The site now shows that there are 105 4K TVs available of which 24 (22.9%) are OLED TVs, which is far above the 6 (8.8%) OLED TVs that were available in the initial study..  While 18 (75%) of the 24 OLED sets available currently are under the LG (066570.KS) brand, Sony (SNE) represents the remaining 25%, while initially only LG had OLED sets available on the site.  The increase in OLED TV representation can also be attributed to LG Display’s (LPL) increased large panel OLED capacity, particularly the increasing production at their Guangzhou OLED fab, which has allowed the brand to expand its size offerings and offer more OLED TV panels to outside brands.  With LGD’s expectations of being able to ship between 7m and 8m, up from ~4.4m last year, we would expect OLED representation to continue to rise this year
TV pricing, looking at the 34 sets that were still available from the original study, has changed both from the original survey and the last time we checked, with 19 (55.9%) of the 34 higher, 10 (29.4%) lower, and 5 (14.7%) the same against original (November ’20) pricing.  When making the same comparison against the last time we checked pricing (66 days ago) there were 20 models (59.8%) higher, 8 (23.5%) lower, and 6 (17.6%) the same, indicating that from a practical standpoint consumers would see a slight decrease in aggregate TV prices (the sample) of 1.9%, but an increase of 8.9% since mid-November of last year.  But wait, there’s more…
Looking at pricing of the 10 most expensive sets of the 34, 6 (60%) are higher than original pricing, 3 (30%) are lower, and 1 (10%) is the same, however of the three that saw a lower price today than last November, all were OLED TVs, reflecting LG’s push to lower OLED set prices to expand unit volume now that production has been expanded.  This follows true with the comparison against pricing done in April, with all of the OLED sets lower in price and the other sets in the top tier the same or higher.  In the lowest priced group of 10, 5 of the 10 sets were higher in price against last November and 5 lower, but when comparing against the April data, 9 out of 10 in the low-priced tier were higher than in April.  Since there was an obvious push by LG to lower OLED TV prices, if those sets are removed from the data, the aggregate price of the TVs in the survey increased by 16.4% since last November, and by 3.4% since April.
As we approach Prime day, we do expect to see some TV set price discounts, but we note that in many cases they will be coming from a higher point than the last holiday season and even last April.  OLED TVs are certainly the exception, and indicative of LG’s higher capacity and lower cash cost, along with the need to generate unit volumes close to this year’s target.  LCD brands however, are facing the ongoing problem of higher material and component costs, and while they might be willing to absorb some of those increases given their high utilization rates, they have been and will continue to be forced to pass those increases on to customers.  Whether the consumer is able to ‘see’ the increases that have already been instituted or whether they show as less overall holiday discounting would depend on both panel price increases going forward and any effect component shortages would have on LCD fab utilization, but at least at this point in the year, it seems obvious that the TV set space has been feeling a bit of pain.