While there are still a few murky areas, Friday’s exemptions to the latest round of tariffs afford the CE space a bit of breathing room during the 90-day postponement of President Trump’s reciprocal tariff program.  At least as far as can be seen based on current White House statements, while the exemptions shown below remove the additional tariffs placed on specific Chinese goods entering the US, they are still subject to tariffs that were placed on said products previously, which we believe amount to ~50% on most semiconductor products.  Therefore exemptions only seem to remove the last round of additional tariffs, leaving those previously levied.
All of this said, the President and cabinet officials continue to insist that new semiconductor and similar tariffs are ‘under investigation’, but with no timeframe attached, only give very near-term relief to CE companies, doing nothing for mid-term or long-term planning.  CE companies can always make promises to shift more capacity to the US, but in most cases such commitments will take years to plan and execute.  For those that are willing to make such a commitment, we remind them that Foxconn (2354.TT) still has plenty of room at its Wisconsin “6th Wonder of the World” manufacturing center, even after Microsoft’s (MSFT) $3.3b data center construction plans for a portion of the site.  As of January (2025), Microsoft has put two of the three Foxconn site projects on hold[1]
While the Friday semiconductor exemptions give CE companies a bit of breathing room, the volatility remains, and while it will take some time for Trump-aligned CE companies to come up with satisfactory plan for production in the US, placating headlines can be created by leaking that some production has been shifted out of China, preferably to an existing production location with low potential reciprocal tariff liabilities.  Given how often the trade situation changes, we expect only those countries or even companies that are so desperate to maintain trade with the US will make more substantial concessions until more trade stability can be maintained.  Rumors stating that some of Apple’s (AAPL) production lines in China have been shuttered have been denied by suppliers, but we expect more of the same until a more stable policy can be established, 1,307 days left…
Semiconductor exemptions

• 8471:  Automatic data processing machines, magnetic or optical readers, machines  for processing such data, including (847330) parts & accessories.  While not specifically mentioned, it seems to cover all desktop and laptop computers.
• 8486: Tools for the manufacture of semiconductor boules or wafers, semiconductor devices, electronic integrated circuits or flat panel displays.
• 85171300: Smartphones.
• 85176200: Routers, modems, and network equipment  
• 85235100: Solid-state non-volatile storage devices.
• 8524: Flat panel display modules, with or without touch
• 85285200: Computer monitors.
• 85411000: Diodes, other than photosensitive or light-emitting diodes.
• 85412100: Transistors, with a dissipation rate of less than 1W.
• 85412900:  Transistors, with a dissipation rate greater than 1W.
• 85413000:  Thyristors, diacs and triacs, other than photosensitive devices.
• 85414910, 85414970, 85414980, 85414995:   Photosensitive semiconductor devices, meaning LEDs and other light-sensitive sensors.  
• 85415100: More specific to LEDs.
• 85415900: This covers pretty much any semiconductor that has not been covered previously, and parts (85419000) including piezoelectric crystals
• 8542: Broad coverage of Microprocessors (CPUs), Memory chips (RAM, ROM), Logic, Amplifiers, Processors and controllers.

We note also that ‘certain critical minerals’ were also excluded from the additional Chinese tariffs, and while not specified, we can guess at a few of them, particularly the ones for which the US relies almost exclusively on China.  Yttrium and Scandium fall under the 280530 code, with Yttrium used in red display phosphors, YAG lasers, microwave (radar) filters, automobile exhaust sensors, superconductors, and in fuel cells.  The US imported over 400,000 kg. of Yttrium in 2023, with China being the source for 99.2% and 94% for yttrium compounds.  Even a simpler material such as graphite, used to line furnaces, for lithium-ion batteries, as a lubricant, in electric motors, and, of course, in pencils, is not produced in the US, which buys 42% of its common graphite and 100% of its battery grade graphite from China.  China has put some limitations on where such materials may be exported, but they have yet to take a hardline when it comes to strategic materials tariffs.   are an area where the US


[1]  While no public details have been given as to why the postponement by Microsoft was made, the company has reduced data center capacity plans in a number of locations this year.

We have noted recently that Sony (SNE) has shown a prototype of a new Mini-LED backlight system that uses colored LEDs instead of the usual white (or blue) LEDs.  The system, which is expected to be commercialized sometime this year, should be able to deliver color purity that would rival OLED (See “Resurrection” (03/24/25) for more details).  However, Chinese TV set brand Hisense (600060.CH) also announced a similar RGB Mini-LED 116” set that would directly compete with Sony and others who are developing the same technology including Samsung (005930.KS), LG Electronics (066570.KS), and TCL (000100.CH).
While RGB backlight technology has the potential to enhance the quality of LCD based TV sets, once again extending the life of LCD as a display technology, there are two questions that must be answered when evaluating whether RGB backlight technology will be a gamechanger, or whether it will be just a marketing gimmick to bring shoppers over to the premium TV set section on the retail floor. 
The first is whether enough TV set buyers will be able to discern enough of a difference between Tv sets using RGB backlights and those using more standard white or blue backlights.  Unfortunately much of the feedback on RGB backlight systems comes from those at shows where such displays are being shown, and most of those viewers are there because they either work in the CE industry or report on it.  This leads to a more biased view of new display technologies and one that is a bit removed from that of the average consumer, so the true test will come when such sets are on the floor at Best Buy (BBY) or WalMart (WMT).
​

The second issue is price.  Most new display technologies, when initially released, carry a large price tag, as in many cases the sets are produced almost by hand until brands are able to develop cost-effective mass production processes.  A good example would be Samsung’s Micro-LED TVs (vastly different from Mini-LED TVs), which when released last June started at $110,000 for an 89” set and climbed to a jaw dropping $150,000 for the 114” set, putting them at the very top of the premium Tv market.  This does not lead to large sales volume but can be used as a banner for the entire premium Tv line.

However, things are different now.  The battle for supremacy in the display and TV markets is no longer a gentlemen’s game, where a civil meeting between rivals might have been held to discuss ways to avoid conflicts.  Now Chinese brands are willing to take that same discussion out to the back alley if they think they can gain share over incumbents, with those battles being waged primarily with product pricing.  It has become ever more difficult for brands that are known for their premium pricing, Sony in particular, to offer a new technology or improvement at such lofty prices, as Chinese CE companies are not only able to quickly match the technology, but are willing to offer it at a far more reasonable price.
​

One might expect Sony, if and when they commercialize their RGB Mini-LED TV, to offer it at a steep premium to other models, especially if it is offered in an extra-large size format. While that is certainly a strong possibility, Chinese brand Hisense has recently announced that it will be releasing its own RGB Mini-LED TV set this month.  What makes that unusual is that Hisense was the first to market with the technology AND at a low-enough initial price that both limits the premium that others might charge and also incorporates it into the largest Mini-LED TV available to retail customers.
We took a quick look at the XXL TV set segment (only sets 100” or larger) to see how the Hisense RGB set pricing compared to other generic XXL Mini-LED sets and here is what we found:

​The Hisense RGB Mini-LED TV is priced (in China) at ~$13,780 US, and we note that it is a 116” model, one inch larger than others on the diagonal (1.64% larger in area), but instead of being offered at a massive premium, like the Samsung Micro-LED 114” model, its price is more comparable to 100” generic Mini-LED backlight models, along with its value on a price- to-area basis.  This will make it more difficult for Sony, Samsung, and others  to introduce their versions at the lofty premiums they might normally use, especially if their sets are smaller than the Hisense model, essentially lowering the price bar before the market has even developed.
Of course, we take into consideration that RGB backlight technology will have a higher cost both in the fact that each backlight ‘pixel’ is comprised of three (RGB) LEDs instead of one, and the driving circuitry for each pixel is now far more complex and costly to produce.  That said, either Hisense has found ways to keep costs low, even during early production, or they are willing to forgo product profitability in order to maintain a share lead over rivals (TCL in China) and in South Korea and Japan.  Sony, Samsung, and LG will now have to find another feature to add to their initial RGB backlight Mini-LED sets when they are released, in order to garner a price premium over Hisense, likely something they were not planning for until recently.

​If you are thinking about becoming a coder after you have finished school, you likely already know about LeetCode, a software application designed to help you study coding, practice, and improve your coding ability.  It contains a large database of coding problems and exercises that help users to improve their skills, along with sets of tools (editors, debuggers, forums, etc.).  However a portion of LeetCode’s content is also designed to help one prepare for a ‘technical interview’, such as one might encounter if applying for a coding job at a large corporation, with common interview questions, practice algorithms, and data structures that can be used to simulate what might be asked in an interview.  This makes LeetCode an ideal tool for interview preparation, and also provides interviewers with lots of potential coding questions that can be ‘asked’ during an interview.
As is human nature, even the smartest of us has pangs of inferiority (except Elon Musk), as did a sophomore in Columbia’s Department of Computer Science, as he used a self-developed ‘cheating tool’ that improved his interview performance enough to receive employment offers from Amazon (AMZN) , Meta (FB) TikTok (pvt) and Capital One (COF).  The cheating tool (named ‘Interview Coder’) acts as an invisible plug-in that cannot be detected by software used for remote test monitoring and is able to create perfect answers to questions with a few minor flaws.  That said, after it was discovered that the sophomore had cheated on his interviews, he was expelled. 
Interviewers look for those that are constantly glancing at the right or left during interviews, or those that are just a bit slower to answer a verbal question (waiting for the answer to come up on another computer), but the ‘Interview Coder’ is much smarter than that and was specifically designed to deal with LeetCode content.  Interviewers favor LeetCode content because it is easier to pull questions from LeetCode than it is to write hundreds of original ones. However ‘Interview Coder’s’ ability to be a ‘hidden’ window and to use system permissions  to bypass the browser’s recording processes, while it remains an open widow on the cheaters computer that can be queried without any cursor movement or keyboard tracking.  It can even break down complex or ambiguous questions into sections to make sure the cheater can answer them with a complete level of understanding and can copy them directly to the test, with a few spelling or grammatical errors thrown in.
Looking to capitalize on his new application after his fall from grace at Columbia, the former sophomore put the app on Github, where it immediately gained large scale recognition.  At a cost of $3,000/month for server support, a $60/month user price, and a 99% profit margin, the app took in $228,500 in its first month, and less than 2 months after its launch it had an annual recurring revenue of $2.2m.  He marketed the application through social media, posting pictures of himself cheating on exams..
LeetCode questions have been long criticized as ~90% having little or nothing to do with actual day-to-day coding work.  Despite that fact, Google (GOOG) uses it for interviews that can take up to 300 hours of answering coding questions.  As the pass rate for those exams is less than 2% it is not surprising that the application is so popular, especially when the typical competitive ratio for a single position garners over 200 interview candidates, making the $60/month fee against a $100,000+ job easy to justify.

​We would like to take a moment to introduce a new member of the SCMR staff.  Previously, when we needed to illustrate a concept of create a visual for a story, we spent hours searching for ‘just the right’ image.  Unfortunately, many of such images are only available for license, leaving us to hope that ShutterStock or similar organization would not notice.  We were happy to credit companies or organizations that supplied us with product images, and we continue to create technical illustrations ourselves, but in order to shorten our image search time, we let our illustration search team go, and hired Imagen 3, a cutting-edge text-to-image model developed by Google DeepMind. 
Imagen 3 is extremely easy to use and as you might have noticed, is able to create images that illustrate a point with only a few sentences of text and perhaps a single redo.  However Imagen 3’s most important feature is that it is free and is already built into Gemini.  So unless it’s a technical illustration or a product photo, most of the images seen recently (and going forward) have ben created by this collaboration between SCMR LLC and Imagen 3.  
Of course this is a remote position for Imagen 3 so he/she/it is unable to say hello, but while we are cautious about the reliability of AI in general, Imagen 3 has been doing exemplary work over the last few weeks!  Please note that any images not attributed are therefore either developed directly by us or in conjunction with Imagen 3.  While Imagen 3 could not be with us right now, we asked him/her/it to create an image of how it would like to be visualized (see below)

Unless you are very adept at using smartphone OS and browser applications, every time you use your phone or open a browser, lots of information gets collected.  Most folks don’t know or care that information is being harvested about where they go, what they do, and what they do it on, and that gives those who know how (or pay to) collect that data, the ability to analyze all sorts of trends.  In most cases, the data is relatively benign and (hopefully) anonymous, so it can be used to come up with thousands of metrics that are primarily used for marketing.
An example might be browser statistics, essentially the number of times a particular browser type is opened in a given period, or how often a website is visited and how long users remain on the site.  These are routine statistics that are readily available, so in the fierce battle for Ai supremacy, such data is as important or more important than the benchmarks that are used to compare Ai model performance.  Depending on which metric is used, AI models can be ranked according to the number of times they are accessed and can even be segmented by country, phone OS (Android, iOS, etc.) and a variety of other categories.  However, when it comes down to who is using what AI and how often, such user data is a good start toward trying to calculate a model’s potential for profitability.
The big problem is the fact that there are typically a number of AI model pricing tiers , with the far most popular being ‘free’, and the data that is collected by your phone’s OS or by your browser does not know if you are paying for the model’s service or if it is free.  This means the value of such user data is more general than might be hoped for, but we take what we can if it helps to put together the puzzle.

In the Table above we show US daily, weekly, and monthly engagements with each of the 5 most popular AI models, with the obvious leader being ChatGPT (pvt) at almost 6 times the closest competitor on a daily basis.  That ratio drops considerably when calculated on a weekly basis, and further on a monthly basis but is still far and away the leader.  GROK (pvt) engagements are quite close to Gemini’s (GOOG) on a daily basis, however weekly and monthly ratios are lower.   DeepSeek (pvt), the newest model, is certainly a contender in the US even though it has been bad-mouthed by the US government as a tool of the Communist Party.
We note that ‘stickiness’, as it is used here is a ratio between daily users and either weekly (Ud/Uw) or monthly users (Ud/Um), with the higher number indicating that users are more likely to return to the model on a regular basis.  Again, ChatGPT is the obvious ‘stickiness’ leadeer  while surprisingly Gemini the weakest.  We would have expected Gemini, with its ability to access some of the Google search index, to have been stickier.

At the global level, while ChatGPT is the leader by far, DeepSeek takes the number 2 spot at more than 4x GROK’s position, and makes a stab at ChatGPT’s weekly stickiness ratio, likely due to its origins in China.  We take things a bit further in the table below where we track the global user base for each model against the percentage of population represented by the US.  As the population of the US currently represents 4.22% of the global population, it can be seen that while the absolute user base for ChatGPT is considerably larger than any of the others, it is a global phenomenon, as is DeepSeek, while Gemini, GROK, and Claude (pvt) are much more US centric, especially Gemini, with US users representing an average of 44% of its total user base.

One unusual datapoint we came across was the average session time spent on each model.  While there was a relatively small difference between the leaders, ChatGPT (3:39) and DeepSeek (3:18) and GROK (2:47) and Claude (2:28), again the surprise came from Gemini with a meager 0:17 average user time.  We expect this short stay is a function of Gemini’s use as a ‘summarizer’ where search results are boiled down to a paragraph or two, although given the model’s large US user base, it could also be due to the very short attention span of US users (joke!).  We will dig further.
As noted above, the data that shows what percentage of each model’s users are ‘paid’ users, is not publicly available, as it would open each model to a level of transparency that would likely make management uncomfortable, especially as growth slows, so right now we have to rely on more typical ratios for subscription models, especially as AI models are a new category.  Worst case would be a 2% to 5% ‘paid’ ratio, and best case between 10% and 25%, the average for ‘freemium’ subscriptions that offer a free trial period that converts to paid if the user finds value.  The calculations get very complex even if we know the exact paid ratio for each model as the models are priced based on the number of tokens processed rather than a monthly fee.  How each user breaks down as to token use is another ‘internal’ metric that will take time to discover.  Until then estimates are just supposition, making future growth and profitability unusually difficult to pinpoint, at least for now, although we expect AI investors might disagree..

To say that 1Q was not a typical quarter for the display space is an understatement, and we expect something similar in 2Q as the current US administration seems to change course on trade on a daily basis.  Inventory stocking and pre-tariff buying have altered typical buying patterns as can be seen in the table below, which shows the monthly sales (NT$) for AU Optronics (2409.TT), Taiwan’s largest panel producer.  Our expectations, prior to yesterday’s tariff course adjustment, was that large tariff related price increases would begin to show as previous landed inventory begins to wind down (end of April/early May), with panel producers seeing a more accelerated slowdown as brands hold back or reduced new orders as prices rise.
The fact that there was a date when the new tariffs were to be enacted was at least an anchor point where brands could make decisions about how to deal with new tariffs, but with yesterday’s 90-day hold, that anchor date has moved again and leaves brands in a quandary about how consumers will react to the postponement, in order to plan production and shipments for 2Q and beyond.  This goal-post movement will elongate the ‘tariff cycle’ that has already changed the typical production patterns for the display space, particularly for those producers that feed major brands in the US, so whatever we thought might happen soon is now spread across a wider window.. 

Given the circumstances, we would still expect brands to pause or slow production and shipments, but for a relatively short period of time, while they try to make sense of the administration’s tariff volatility, but the real question is how consumers will react.  There has certainly been signs that some consumers had been buying CE products in anticipation of expected price increases in March, and the table above, albeit specific to a single large panel producer, illustrates that pull-in buying, or at least production that could facilitate that buying, has helped 1Q sales results.  However, the increased Chinese tariffs and the remaining 10% across-the-board tariffs still overhang much of the CE space and the more general anxiety over the potential for a tariff-related economic slowdown still remains.

​The Indian government has indicated that recently (March to early April) Apple (AAPL) has been air shipping iPhones from their production sites on the sub-continent to the US in order to beat the latest round of tariffs.  According to sources in India, six cargo jets have been used, each with a capacity of 100 tons, to move the phones into the US before the April 5 deadline, with the last one earlier this week.  This capacity implies, using a package weight of 12.4oz, that the total shipments included ~1.5 million iPhones (not our calculations), that had been part of an emergency assembly program that included a Sunday production addition at Foxconn’s (2354.TT) largest assembly plant, and an accelerated push through customs at the Chennai airport that cut customs throughput time from the usual 30 hours to ~6 hours.  For reference Foxconn typically ships between $110m to $331m (value) from India to the US monthly, but saw shipment value increase to $770m in January and $643m in February.  It was said that the Indian government put in a request with customs officials to ‘support’ Apple’s expedited clearance, which the company had been planning since last June as tariff rhetoric increased.

Since we have been speaking about postponements, once again the President signed an Executive Order giving TikTok (pvt) another 75 day extension to finalize a deal to remove itself from its Chinese owners.  One might remember that the first 75 day extension came on Trump’s inauguration day when the original ban (Biden) and Congressional actions were to go into effect.  The President recently stated, “The Deal requires more work to ensure all necessary approvals are signed, which is why I am signing an Executive Order to keep TikTok up and running for an additional 75 days.”, as his initial extension was to expire on April 5.
That said, it seems that while the WH hints that at least four groups are currently under consideration to purchase the app, which is used by over 170m Americans, there has been little about who might be high in the running.  At one point last year it was thought that Oracle (ORCL) might be close to a deal, but the Chinese government indicated it would not approve the Ellison deal at that time.  Since then Frank McCourt, former owner of the LA Dodgers, has been mentioned, as well as Perplexity AI (pvt), a variety of VCs and private funders, and more recently Amazon (AMZN), who has been trying to build a media presence (Bought Goodreads (pvt) in 2013 and Twitch (pvt) in 2014), but recently closed their short-form platform ‘Inspire” after ~2 years.
While the most recent postponement, which is again in violation of the congressional law (Trump promised not to prosecute those involved), is delaying the finality of a deal, there still seem to be a few loose (very loose) ends that are keeping things up in the air.  Trump fired David Feith, the Senior Director for Technology & National Security at the National Security Council, who was a key negotiator in the TikTok deal, which set back the conversations with the Chinese government, and even more recently the additional tariffs placed on Chinese imports cooled negotiations even further.  That said, even more recently another bidder seems to have appeared, and while the US government might not take this new bidder seriously, he certainly has credentials.
The new bidder is Only Fans (pvt) porn site founder Tim Stokely, who sold a 75% stake in the site in 2018 (sales of $6.6b in 2023).  Along with the HBAR Foundation (pvt), who runs the Hedera block-chain network, Stokely is looking to create another profit-sharing model like Only Fans, where those willing to post explicit videos are able to rake in thousands a day for a popular pay channel.  Hopefully the administration will not take the Stokely bid seriously, as the potential for such a model to destroy the public value of TikTok is quite high, but there is still another stumbling block that we see as an even greater risk to any potential TikTok deal.
As the Chinese government has to approve the sale, the current atmosphere between the Trump administration and the Chinese government is a bit tense, and we expect the Chinese government will try to use TikTok to gain some leverage in the tariff debacle.  Trump can easily hold out with this most recent extension, but as such a high visibility deal, it will become embarrassing to  extend it again, especially if existing tariffs continue to put excessive pressure on the US economy. 
We expect the Chinese government will use the TikTok deal as a negotiating tool, but as China seems to be the easiest target for Trump’s ‘Big Stick’ agenda,  we don’t believe the deal leverage has much chance of making an appreciable difference in trade negotiations.  That means either no TikTok deal and a shutdown, which would be an embarrassment to the administration, or some concession on the US side, likely something easily buried in the publicity rhetoric that would surround a deal.  (“…we saved TikTok for the good of America”.  It has become a messy situation that has gotten even messier with the Stokely bid, but there are bigger things to worry about than TikTok at the moment, at least until the middle of June when the deal is slated to be completed.  As long as Stokely’s bid is not accepted, TikTok lives on, despite the efforts of the government to ‘protect’ our TikTok videos from being seen by the Chinese government…
https://www.tiktok.com/@creamycanoli/video/7350849230010928430?is_from_webapp=1&sender_device=pc
https://www.tiktok.com/@biogenesis__/video/7478422860545969430?is_from_webapp=1&sender_device=pc
https://www.tiktok.com/@shiverchain/video/7468920864952552726?is_from_webapp=1&sender_device=pcc
https://www.tiktok.com/@wufudufu/video/7253693814928379178?is_from_webapp=1&sender_device=pc
These are all real TikTok videos…don’t let anyone in the Chinese government see them!
​

Yesterday it was reported that as the 32% tariff in Taiwanese goods entering the US was about to begin, a number of key notebook brands were notifying their supply chain to suspend all notebook and related component shipments to the US for the next two weeks.  Among the brands thought to have participated in the suspension are Dell (DELL), HP (HPE), Lenovo (992.HK), Asus (2357.TT), and Acer (2353.TT), although official statements have yet to be verified.  Over the last 8 quarters, those 5 laptop brands represented between 72% and 76% of all laptop shipments.
While laptop brands have reacted quickly (and after today are likely regretting the idea), component manufacturers are not seeing a radical change in demand and see shipments as relatively normal thus far.  Some hardware oriented component manufacturers have seen a few postponements, although discussions are ongoing about the potential for tariff sharing should they remain in force.  That said, with the tariff front changing daily or hourly, the biggest issue is not the tariffs themselves but the inability of businesses, large and small, to plan on a forward basis, especially as trade talks between individual countries and the US will take weeks or months to resolve.  We don’t need a war or pandemic to create economic instability, Trump is able to do that himself…  +With only 1,306 days until the next election, a good CEO could promise to move production to the US but get bogged down in the contract details for a mere 186 weeks…