The smartphone business is a tough one.  With hundreds of brands looking to catch a consumer’s eye and feature sets that grow longer with each new model, brands struggle to find something new and exciting each new model cycle.  Cameras were big a few years back, with phones having up to 6 cameras, but multiple cameras began to take up room needed for other components, so the fad moved from camera numbers to camera resolution, with current high-end phones in the 200MP resolution class.  To put that into perspective a 200MP camera has 16,384 x 12,288 pixels in a 1” x 1.2” package (18,369 Pixels per Inch).  But the challenge continues, as AI throws another variable into the smartphone mix.
It would seem that brands could be a bit wary about how AI will play out on mobile devices as it is quite difficult to tell whether the Ai flavor-of-the-month is pushing consumers to buy a new phone, so Samsung (005930.KS) has taken another track, and thin is in.  The just announced Samsung Galaxy S25 Edge is a mere 5.8mm thick, a 29.3% reduction from this year’s Samsung Galaxy S25 Ultra flagship smartphone. Here are just a few of the gushingly positive comments the tech press has made about the just announced Samsung Galaxy S25 Edge, their newly designed ‘super thin’ smartphone.

• "Thinner than a credit card..." - The Economic Times
• "Samsung's Galaxy S25 Edge is already turning heads in the tech world...for its sheer thinness." - The Economic Times
• "The S25 Edge is one of Samsung's thinnest flagships to date." - Techloy
• "The Shockingly Thin Samsung Galaxy S25 Edge Feels Great to Hold..." - PCMag

But while ‘thin’ might be the new rallying cry for smartphone designs, thin has been around for a long time, and the 5.8mm thick Galaxy S25 Edge is barely a contender for thinnest.  The earliest ‘thin’ phone we could find was the Samsung U100 that was released in February of 2007, 18 years ago, while this baby was a fat 5.9mm thick, it was certainly the first that met our criteria of being under 6mm.  That said, the real winners are the pack of contenders at the Wegovy-like 5.1mm, a staggering 12% thinner than the new Samsung S25 Edge!  In fact all of the smartphones that were 5.1mm (the smallest we can find) were released 10 years ago in 2015, a period in time when there was less smartphone hardware, less heat generation, and smaller batteries to take up space inside the case.  5.1mm is 0.2” or a bit less than ¼ of an inch.
We are sure that the ‘thin’ trend will have some sustainability, at least until something else comes along, as Apple (AAPL) (see below) is expected to be releasing the iPhone 17 Air, a substitute for the iPhone Plus model, with the iPhone 17 Air rumored to be only 5.5mm thick.  If both Samsung and Apple follow the trend, almost all others will join until one of the three largest components, the battery, the display, and the camera modules hits a limit.  Batteries can be made thinner by increasing the length and width, which, in theory, should make it easier for the battery to dissipate heat, but at the same time the thinner phone allows less room for the thermal interface that transfers the battery heat to the frame of the phone, so it’s a tradeoff.  The display itself is quite thin but the connector that attaches the display to the internal circuit board is also a limiting factor, and lastly the camera module needs a certain amount of depth to operate properly, also a gating factor.
All in, while thin is in currently, we expect thin as a smartphone feature to run its course relatively quickly unless primary component producers are able to work dimensional magic without compromising performance.  There is also the feeling of holding the phone in your hand when you use it, which changes with thickness and weight, but in a device that must fit into a pocket, less is more.

AI Writing tools are considered applications that have considerable benefits for content creators.  They can be used for creating blog posts, social media content, and generating presentations or e-mails, with some incorporating image creation tools to complement written content.  AI writing tools were among the earliest AI applications to be commercialized and have been popular since.  That said, as LLMs improve, the necessity for such specific applications will decline as these general purpose LLMs are able to perform essentially the same processes, albeit with a bit more prompting, and are increasingly part of desktops or phones, either embedded in applications or as standalone systems.
While we don’t expect a transition from such writing tools to be immediate, there seems to be a bit less momentum behind such writing tools and there is some recent data that seems to bear this out.  The chart below shows the traffic patterns for a number of popular AI writing tools and while there are certainly ups and downs only one managed to maintain a positive traffic pattern thus far in May.  It is very short-term data, but it could be a harbinger of how general purpose LLMs are capturing the ‘easy’ processes that more specific AI applications have been built for.  More to come…

Meta (FB) has indicated that it is going to use personal data from Instagram and Facebook users in the EU to train it new AI systems as of May 27.  Under European GDPR an entity can use personal data if a user opts-in, essentially providing a “freely given, specific, informed an unambiguous” ”Yes” to processing their data.  However Meta is claiming “legitimate interest” in the data, a way of circumventing the opt-in/out rule and is even claiming that users don’t have the right to opt-out unless they do it before the training begins.
Meta tried to use the ‘legitimate interest’ clause to target users with advertising, but the European Court of Justice ruled against Meta, so a number of privacy advocate groups are filing cease and desist letters with Meta, triggering a return to the courts.  Should Meta lose the case, aside from not being able to use the personal information, they could be liable for damages, which are quite severe under GDPR, especially as once the personal data is included in Mata’s training data, they will also be unable to comply with other GDPR rules, such as those that require that user data be forgotten upon request or those that require user access to individual personal data and the ability to correct inaccurate data.  Because the Meta models are open-source, once the personal data is incorporated in the training material it cannot be called back and removed.
What makes this unusual is that there is a simple solution, just ask users for their consent, but Meta is making the case that any difficulties in using this data would remove the ‘local flavor’ of the training data and hinder the development of Ai in the EU, letting the US and China take the lead.  In fact if Meta made it understood that the data would be anonymous, they would most likely receive enough opt-ins to give significance to the local trends that they are trying to capture.  The necessity for absorbing all personal data from Facebook and Instagram users in the EU is ridiculous as other AI companies, such as OpenAI (pvt), the leader in the AI space, and France’s own Mistral (pvt) do not have any personal information from social media in their training data yet still perform in line with or better than those that do.
Meta’s ‘legitimate interest’ ploy is forcing a number of privacy groups to file injunctions before the proposed date when Meta will begin access to user personal information. This is being done in Ireland where Meta’s EU headquarters resides and in other jurisdictions.  If Meta were to lose in the EU, not only would it have to stop processing user personal information but it would have to delete any Ai systems that were trained with that data, and if EU data is mixed in with non-EU data during training, the entire model would have to be deleted.  The injunctions also stop the statute of limitations clock, and every day that Meta continued to use what would then be illegal data, would increase the potential damage claims by users or the equivalent of class action suits.  The privacy groups estimate that if the fine were only €500, with ~400 million monthly active Meta users in Europe, the aggregate fine would be over €200 billion,  It would be a lot easier to just ask users to opt-in.

​Sharp (6753.JP), once the shing star of Japan’s LCD TV display panel business is selling its Kameyama #2 LCD fab to its parent Foxconn (2354.TT).  The fab, which originally produced TV panels, has been producing small displays for laptops, tablets, and smartphones of late.  Foxconn has not indicated what they will do with the plant when the deal closes (expected by August 2026), but in the interim the fab will continue to operate under Sharp’s control.  Given that Foxconn is Sharp’s largest shareholder, this might be a case where the sale is being made to dress up Sharp’s profitability, with the losses from the fab easier to absorb at Foxconn.
When the sale is completed, Sharp will have two remaining Gen 6 fabs (Hakusan and Kameyama #1) and one Gen 4.5 fab (Mie #3).  The Mie #3 fab has been running at ~50% of its stated capacity, with Sharp expecting to more than double display revenue from that fab by fiscal 2026.  While it is difficult to pin down, given the rapidly changing CE market, we expect much of Sharp’s remaining LCD capacity will be oriented toward the automotive display and XR display markets, with larger panels being produced at Kameyama #1.  Sharp is targeting an increase in the production of panels over 13” at Kameyama #1 from ~30% now to 55% by fiscal 2027.
It would seem that with the sale of Kameyama #2, Sharp has trimmed its display business down to a more manageable level, although relying on continued growth  in automotive displays.  As Sharp is considered one of the most recognizable brands in the CE space, they continue to produce their own TV set branded product, which they can still produce, after the sale, maintaining in-house display production.  Unfortunately Sharp does not produce OLED displays commercially, so we expect Sharp’s OLED TV models are based on LG Display’s (LPL) WOLED panels not on Sharp’s own OLED production.  Hopefully a smaller, more profitable Sharp will be the result of the Kameyama #2 sale…

To say the current administration is pro-business is a gross understatement.  In dozens of instances government regulations, good or bad, are being discarded in favor of a no-holds-barred free-for-all among producers, putting aside any caution over environmental or other potential consequences.  We’ve been here before, with some administrations favoring a more restrictive stance while others a more ‘self-regulating’ focus.  But when it comes to most of the industries where the government has intervened on behalf of the environment or on behalf of industry, the general outcomes of both regulation and de-regulation are well known.
Not so with AI, a technology that is evolving at an extremely rapid rate, and as we have noted, not always in a good direction.  Much of the negative side of AI comes from its use by humans’ so in those cases we cannot blame AI, but as the technology gets more sophisticated, we seem to notice that AIs themselves have some ‘issues’. There are many in academia and in the Ai community itself that are cautious about the rapid pace of AI development and the fact that even those who develop the infrastructure and architecture that Ais run on, do not always understand how they work and why they have ‘issues’.  Letters have been written, cautious tales have been told, and even a few examples have been given as to how models can cause problems, and we have barely put them into service.
Once AI systems are embedded in our infrastructure, they will become even more of both a tool and a potential hazard to our everyday lives.  Think of what happens to your kids when the internet is down because of a poorly written slice of code in an update and then think of that same poorly written code being inserted into the Ai system that controls stoplight timing in a small city.  Taking it a step further, what if that code snippet was put into the model that runs the power grid across a state or region?.  But, according to a bit of legislation inserted into the current budget reconciliation bill, all Ai regulation at both the federal and state level will be halted for 10 years to let the industry grow.
That means that developers will not have to report new models to the government, nor meet specific testing rules, and, in its most pure form, would allow AI developers to claim almost anything with little or no backchecks.  The business of Ai loves this idea and promotes the potential change as opening the door to Ai innovation (aka ‘beating China’) and the administration sees it as dollar signs.  But what about the poorly tested health oriented Ai that is giving bad advice to users, or the AI that is making value judgements about potential job candidates with some obvious biases?  There are many scenarios where an unregulated Ai industry, tasked with self-regulation will do what most self-regulating industries do, de-regulate until a major incident and then complain when the regulations return.
We are not saying that the Ai industry should be highly regulated and relegated to single digit 10 year CAGR, but to ban all regulation for AI decision systems is a bit of an extreme in the other direction, a direction where we have yet to travel.  If this was happening in the oil and gas industry, the potential outcome would likely be a huge oil spill or contamination, as such things have occurred, but with Ai we have no idea what the consequences might be, so perhaps just a bit of caution might be a good idea…We know who is going to speak for the industry, who is going to speak for us?  It could be too late once the s$&t hits the fan…

The volatile trade situation between the US and China has actually existed for years, however more recent trade issues have been based on agreements set by both countries as early as 2016 and 2017.  But in 2020 the US and China signed the Economic & Trade Agreement, known as “Phase One”, which was to level the trade playing field between the two  countries.  As this agreement is the basis for many current trade issues, we took a look at the agreement itself and how the current situation developed.
Economic & Trade Agreement (“Phase One”) Agreement) between the US Government and the government of the People’s Republic of China.

• BASICS: To Improve access in the agriculture and financial services sectors, refrain from problematic policies and practices related to IP and technology transfer and increase its (China) purchases of certain U.S. goods and services.

Under its 2015 10-year plan the Chinese government targeted 10 sectors where it wanted to raise industrial productivity: Advanced information technology, automated machine tools and robotics, aviation and aerospace equipment, maritime engineering equipment and high-technology ships, advanced rail transit equipment, new energy vehicles (NEVs), power equipment, agricultural machinery, new materials, biopharmaceuticals, and advanced medical device products.

• The goals of the “Made in China 2025” program (a subset of the 10 year plan), are to develop ‘indigenous technologies, IP and know-how, substitute domestic technologies, products, and services for foreign ones, and to capture a larger share of worldwide markets.  The US government sees these goals as problematic in that they are supported by more than $500 billion in governmental financial funding and would create excess capacity and cause ‘long-lasting damage to US interests’, although it is hard to understand why such incentives are different from those offered by the US government.
• Along with the subsidy issue, back in 2017 the US government began a Section 301 investigation into China’s IP practices.  Specifically, the use of a variety of tools to require or pressure the transfer of technologies and IP to Chinese companies.   This is viewed as a way of depriving U.S. companies of the ability to set market-based terms in technology licensing negotiations with Chinese companies, a valid complaint.  The US took it further citing Chinese intervention in markets by directing or unfairly facilitating the acquisition of U.S. companies and assets by Chinese companies to obtain cutting-edge technologies and IP; and conducting or supporting cyber-enabled theft and unauthorized intrusions into U.S. commercial computer networks for commercial gains.  Since then the US reviews all acquisitions of US companies to determine if the technology is sensitive or crucial to the US military or key industries.  They have nixed many such deals.
• ISSUES: Tariffs have been a method for protecting industries from aggressive incursions by those who have  a financial advantage and both the US and China use them.  China’s trade tariff for MFNs was 7.5% in 2023, with an average of 14% for agricultural products and 6.4% for non-agricultural products, but China bound 100% of its lines to the WTO, which means the rates they set are the maximum rates they can collect.  This while not perfect, gives a sense of stability to the tariff system, something the US (under th current administration) has decided not to do.
  • The US government has been complaining that  as of the end of 2024  China has not even met its import tariff rate quotas for most products, which are agreed upon levels that cannot be exceeded without triggering even higher tariffs.   That said, those levels have been exceeded for corn and wheat imports from the US.  The Phase One agreement also required that China make improvements in its administration of TRQs for wheat, corn, and rice, which the US government says that China has not done so. 
  • The US government has also been complaining about other trade barriers it believes China has set up to prevent US companies from developing share on the Chinese mainland, particularly regulations around standards for agriculture, cosmetics, and meats and poultry, along with the difficulties China presents to those external suppliers that wish to participate in the country’s Government procurement programs.
  • Since the 2020 agreement, either individually or as part of the WTO the US and China have had innumerable meetings and discussions concerning the abovementioned issues and other policy complaints, particularly covering IP, trade secrets, trademark ‘squatting’, online piracy (movies, books, audio, etc.), counterfeit goods, and what the US government calls ‘indigenous Innovation’, essentially China’a favored approach to IP developed in China. 

All of these issues, along with similar issues in the financial services sector, and the alleged Chinese government’s inability to follow through on mutually agreed on changes, are the basis for the current administration’s trade policies with China, at least on the surface, while China’s government subsidy programs are also a bone of contention. Labor practices, preferences for state-owned companies, and VAT taxes are also on the list of complaints.
However we note that a quick check of US government subsidies shows that this year the US is estimated to be spending $42.4 billion in direct government farm payments and is expected to spend $181.8 billion for programs like crop insurance, risk coverage, and price loss over the next 10 years.  Fossil fuel direct subsidies and tax breaks typically range from $10 billion to $52 billion each year, but if you count indirect subsidies that benefit the industry, the number climbs to ~$760 billion, and those are only two of many categories, making US complaints about Chinese subsidies a bit moot..
When it comes down to the current state of trade between the US and China, the true motivating factor is political.  Yes, China has the largest trade deficit of any nation with the US, but it is also the largest market in terms of purchasing power parity and the 2nd largest in terms of GDP and is a heavily manufacturing oriented country with a relatively low consumer consumption rate.  As the US consumer is less savings oriented than most other countries with high GDP’s, US consumers are going to want to buy more goods from manufacturing oriented countries like China than China buys from us[1].  Trying to balance trade with China by bringing production of googly eyes or sandbox toys into the US is a waste of time as most small business owners will confirm as it is quite difficult to compete in labor intensive markets.  When threatened, as they are now, China and other manufacturing countries will agree to terms, make token gestures and continue on a similar path as they have been.
Trying to convince US consumers to buy only US made products, even if they are more expensive than those made in other countries only works in time of prosperity, and in most cases US consumers don’t examine everything they buy as to its country of origin.  So expecting all goods that are manufactured in China to shift production to the US is a losing game.  That said, the ratio of US services exported to China against goods imported from China has increased from 3.4% 20 years ago to 10.9% last year.  In creasing that ratio will work toward reducing the US/Chinese deficit more realistically over the long-term, without forcing the US consumer to change habits, buy less from China, or pay more for goods manufactured in the US.  We should certainly try to retore some manufacturing to the US, but it needs to be manufacturing things that are based exclusively on US technology, not generic products that have labor as their biggest cost.
For example, Semiconductors make sense, but not high volume mature node products that are relatively easy to manufacture, just leading edge products and the tools needed to produce them.  When it comes to CE products, how can we compete with rows upon rows of factory workers who get paid (average factory worker) between $8,184[2] and $14,435/year to a US factory worker whose average salary is $53,416[3].  Even when adjusting for purchasing power parity, the US factory worker makes between 3.1x and 1.7x more than the Chinese worker, and that is only the labor cost.  Tariffs are a way of trying to force long-term change in a short period and when it comes to responding China has shown that it can provide lip service for as long as necessary to prevent ‘stick damage’.  They will likely do it again, so perhaps another method should be tried.
 
 


[1] We note that US consumption to gross GDP is 62%; China’s is 39%.

[2] National Bureau of Statistics of China, CEIC

[3] US Bureau of Labor Statistics
 

​As we noted yesterday Samsung Electronics (005930.KS) was expected to release what might turn out to be the replacement for the long-lived Samsung Galaxy S Plus flagship smartphone that has been a part of the company’s flagship line-up on an off since 2011.  While Samsung and many other brands have strict rules about suppliers and employees leaking information about a product before its release, leaking’ information about upcoming smartphones has become both a marketing tool and a game of cat and mouse. As can be seen by our comparison of yesterday’s (pre-release) feature set and today’s actual feature set, much of the key information about the Galaxy S25 Edge was already circulating and turned out to be largely correct.
Below we show our feature set list from yesterday and the actual feature list from the release.  We highlight those features that were not correct in yellow.  As can be seen, we were off by 3/100ths of an inch on the length and 1/100th of an inch on the width.  The Edge can handle both Sim types, not just one as we expected, the display has an LTPO (Low temperature Poly-Oxide) backplane (we were unsure if it would be LTPS or LTPO), and the initial price will be at the low end of our expectations at $1,100 for both memory configurations.  Other than those errors, the device features seem to have been well telegraphed.

The feature set is mostly as expected and even a bit better, the price is exactly between the price of the S25 and the S25 Ultra and the same price as the S25 Plus (512GB).  Of course lower would have been better, but in the case of the S25 Edge it seems Samsung is trying to update design rather than create a new price category.  The Galaxy FE (Fan Edition), which was released in September of last year, creates a lower-cost bridge between the Galaxy S25 series and Samsung’s mid and lower price tier lines.

Back in January we noted that President Trump announced , on his first day in office, the Stargate Project, a $500 billion dollar project to build 500,000 ft2 data centers across the US.  The President outlined ‘Big beautiful buildings’ that would create over 100,000 jobs in partnership with Japan’s Softbank (9434.JP), Open AI (pvt) and Oracle (ORCL), the first of which actually started construction in mid-2024.   The initial $100 billion was said to be deployed ‘…very, very quickly’.   As the President’s speech warmed up, he noted that on the first day of his presidency his administration had already lined up $3 trillion in new US investments and expected to have a total of $6 - $7 trillion by the end of the first week.
The Stargate project, which bears an eerie resemblance (see our note “Stargate – Different This Time?”) to the Foxconn (2354.TT) ‘6th Wonder of the World’ display project in Wisconsin that the President touted during his first term, a project that is still looking for a purpose and has barely produced anything, although much of the Stargate Project is to be funded by private funds, rather than the public funds used for the Wisconsin project. Softbank is expected to directly fund 10% to 20% of the project with capital from its Vision Fund, with the rest of the capital coming from debt.
While Softbank met with a number of banks and PE firms in January. Those negotiations did not prove meaningful, and it seems that Softbank’s team of 20 to 30 has yet to complete the project financial plan and is now working on  smaller project by project financing.  In the interim a number of changes have occurred that are making it difficult to find investors willing to fund such a project.  Those changes are:

• China shocked the world with DeepSeek (pvt), a model that promises a much lower cost of training.
• Trump’s own tariffs are estimated to potentially add between 5% and 15% to the cost of data center construction.
• Fears of industry overcapacity as Microsoft (MSFT) cut back its data center plans and Amazon Cloud Services (AMZN) saw slower than expected growth.

These issues have caused investors to become far more cautious than in 2024 and many are waiting to see how the trade situation plays out before making any large scale investments in the data center market.  This leaves Softbank with less of a story to tell investors and a potentially more difficult task to explain how the project will maintain a high utilization level, necessary for the return that Softbank expects. 
While none of this is really surprising given the fact that the current administration is big on promises and light on details, it would mark another time when US politics overpromised and underdelivered.  We expect more building will be built in Abilene (They were being built even before Project Stargate) but ‘very, very quickly’ already seems to be fading away.  It’s a good thing that the public has a short attention span for such things or they might notice a pattern.

AI chatbots are not new, despite the recent flood of conversational LLMs.  ELIZA, developed by Joseph Weizenbaum at MIT AI lab in 1964 is considered the ‘first’ AI chatbot as it used pattern matching as the basis for its responses, but over the years much of what were known as chatbots were based on pre-written scripts and rule based systems such as the “Ask XYZ” systems that have been available on many websites for years.  The current crop of AI LLM based chatbots are far more sophisticated and based on large, high-performance models.  More recently chatbots features have been extended to included internet search capabilities and early predictions were that search enabled chatbots would spell the end to standard search systems.
While there certainly has been impact from Chatbot search, the numbers reflect only a small incursion on regular internet search visits.  A recent survey[1] indicated that over the period between April 2023 and March 2025, there were 1.863 trillion search engine queries, down 0.51% y/y.  The obvious dominant search engine was Google (GOOG) Search with an 87.6% share, and a1.41% reduction in search volume.  Bing (MSFT) was a big gainer in search volume, up 27.77% during the period, but even as the number two search engine, Bing only has a 3.2% share of the search market.


[1] Onelittleweb.com
 

The chatbot search top 10 is similar in that ChatGPT (OpenAI) is the dominant player, however when one looks at the number of searches generated by chatbots against generic search engines, the search engine traffic is over 33 times greater, so as of today, chatbot generated searches are only 2.96% of what is generated by generic search engines.  However chatbots offering search capabilities are new, as can be seen by the ‘Search Added’ and the ‘Current Period Change’ columns, where newer chatbot show extremely large y/y increases but small shares.  That said, most chatbots that have been around for more than a year have more reasonable search growth rates.

​Based on the current data it would seem that chatbot searches have had little effect on overall generic searches, but it is far to early in the cycle to make a long-term judgement.  In fact we expect at least Gemini, Google’s LLM/chatbot, chatbot related searches will be incremental to generic search traffic.  Likely, although to a lesser degree, Co-Pilot will have the same effect on Bing.  On a longer-term basis, as AI chatbots become more embedded in operating systems, it would seem logical that unless you were looking for an answer to a question that you specifically request includes an internet search, the AI would try to answer the question using its most recent training or update data.  If, and that is a big if, the data corpus of the AI is broad enough, the answer might not require the chatbot to conduct an internet search, and in that way could weigh on internet search growth. 
Much in that scenario depends on how ‘transparent’ the Chatbot is.  If it is always available on a home page and looks and acts like a search bar, users will gravitate toward the chatbot/AI.  If it has to be chosen, is slow to answer, or comes up with skimpy answers, users will remain generic search fans.  But there are other factors that come into play. 
It is understandable from the standpoint of the AI owner that they keep the compute time as low as possible, so the default would likely be a query first run without an internet search.  However as time goes by, we expect most chatbots will default to  include internet search results in query answers, but it is not quite that easy because those searches are not always free.
Currently, some (Perplexity (pvt), Co-Pilot, You.com (pvt), Komo (pvt), Andi (pvt), and Brave Search (pvt)) check the internet on every search, while others either have their own decision mechanism (Gemini, Claude (Anthropic), ChatGPT, GROK (xAI), META (FB)) or users can request an internet search.  But going forward things get more complicated.  Some CE companies use their own AI infrastructure to run embedded chatbots, while others are based on ChatGPT or other existing infrastructure.  Apple (AAPL) runs Apple intelligence on its own servers, it does not ‘pay’ for searches, although the all-in cost of each search is amortized over the infrastructure cost.  Others, who might use ChatGPT  or Gemini as their AI infrastructure providers that support their chatbot, would have to pay for each search, as both ChatGPT and Google charge on a per search basis, so the advantage goes to those who can support their own chatbot/search infrastructure.
All of this comes down to the value that consumers see in ‘free’ chatbots.  If results from trained data are enough for most, chatbots will have a modest impact on search and a modest cost to chatbot owners.  If chatbot results are not sufficient for the average user, the cost to maintain ‘free’ chatbots will rise as search fees climb, unless the chatbot owner can convert users to paid plans. 
Search has been around for a while and chatbots are relatively new so we expect the impact from chatbots will be small over the next year, but as the general public gets more used to using chatbots, the competition between them will increase and we expect search will be an integral part of all chatbots, especially as the average consumer becomes more aware of chatbot data sources. 
We use eight different chatbots, typically querying at least two each time, especially if the query has a need for very current data.  We know which chatbots are search oriented and which are not, and which will cite sources, either trained or search related, but the average person will likely use what is easiest, and that is usually whatever is built in to the OS or specific applications, so it is up to the brand to decide what level of search is sufficient for their users.  As Ai is such a high-level feature now, the competition will continue to increase and that means more ‘free’ features  which is certainly  good for consumers

Samsung (005930.KS) has been producing a ‘plus’ model of its flagship Galaxy S line on and off since 2011, substituting Galaxy Edge, Galaxy Lite, Galaxy Active and even the Galaxy FE in the years when the company did not release a Galaxy S+ version, yet since 2017 the Galaxy Plus model has appeared each year along with the standard Galaxy S and Galaxy Ultra models of its flagship line.  Tomorrow, Samsung will announce the Galaxy Edge, a new and improved Edge that has the opportunity to replace the Galaxy S Plus going forward.
The Galaxy S Plus has been the underperformer of the three flagship models, and this year Samsung is looking to see if a new Galaxy Edge model might be a better value for consumers than the S Plus going forward.  There are lots of Galaxy Edge features that have been leaked or speculated on, so we put together a comparison of the current Galaxy S25+ and the upcoming S25 Edge.  We note that the Edge feature list could change upon its release.  We note also that the concept behind the Edge is a thinner, lighter, sleeker’ phone with most of the features of the Plus, with less cameras but higher resolution ones.

According to suppliers, of the 37.7m Galaxy S25 series phones to be produced this year, the standard Galaxy S25 will represent 36.1%, the Ultra, 46.2% and the Plus only 17.89% based on expected demand, so the question will be whether the new Galaxy 25 Edge will be able to outsell the Plus.  If the Edge catches on with consumers, we expect that the Galaxy Plus will join the Samsung Neos, the Samsung Minis, and the Samsung Actives, all of whom have been sent to that farm upstate where all of the disabled and sickly hamsters are happily living out their retirement days.