• Mediatek’s new chip for Chromebooks sets a new standard for the high end of the Chromebook market and looks to me to be merely a stone’s throw from Intel and AMD facing another newcomer in the PC Market.
• MediaTek’s Kompanio Ultra has an NPU that runs at 50 TOPS, 8 Big Arm v9.2 cores(Cortex-X925), 2 x 4K external monitor support and the promise of 20 hours of battery life.
• The chip is made on TSMC’s 3nm process node and is sporting the kind of specifications that would be acceptable in the middle of the laptop market.
• I expect that Qualcomm will take another step forward when it launches its X Elite 2 this year leaving MediaTek space to slot in right behind it just as it does in the smartphone market.
• The only real issue I see with MediaTek taking this chip and spinning it into the laptop market is software.
• Although both Qualcomm and MediaTek are using Arm, Qualcomm has its custom Oryon core whereas MediaTek is using a processor design that it buys from Arm.
• This is likely to mean that MediaTek will need to work with Microsoft to make its chips work as well with Windows as Qualcomm’s to ensure that its offering is competitive when it comes to market.
• This chip looks to me like an expression of MediaTek’s ambition to come to the PC market and put even more pressure on Intel and AMD.
• The real winner here is Arm, which will see its penetration of PCs accelerate and it will earn higher royalties from MediaTek which buys the processor design and licences the IP as opposed to Qualcomm which just pays for the IP.
• This is neutral for Qualcomm because competition in this space is healthy and will keep Qualcomm on its toes and push it to continue to lead when it comes to performance.
• Intel and AMD are the real losers here and it increasingly looks to me like the x86 architecture is increasingly obsolete.

Microsoft & OpenAI – The divorce

• Yet more delays in building data centres by Microsoft is not a sign that the AI freight train is about to hit the buffers but more an indication that it is slowly extracting itself from its relationship with OpenAI.
• The latest delay involves three data centres that were to be built in Ohio bringing the total delays and curtailments to 2GW in terms of capacity.
• The Ohio projects were announced in October when OpenAI was expected to continue sole sourcing its compute capacity from Microsoft.
• Microsoft is not relinquishing ownership of the real estate and so these look like delays as opposed to outright cancellations.
• Since that time, the relationship has soured substantially with OpenAI being reluctant to share key technological details of its models with its once closest partner and I think Microsoft has realised the substantial risk it is taking by being dependent on a company with weak corporate governance.
• Crucially, these delays have not resulted in Microsoft cutting its $80bn capex forecast for this fiscal year meaning that whatever it has saved is being used elsewhere given how strong demand remains.
• Hence, I don’t think that this has anything to do with end demand and when Microsoft, Amazon and Google report this quarter, we will continue to see strong demand for their services.
• I think that this is a wise move by Microsoft as I have long been concerned with Microsoft’s dependence on OpenAI and viewed it as the only real mistake made by an otherwise almost faultless Satya Nadella.
• Microsoft will now head off in another direction and source its models from multiple sources including in-house meaning that when OpenAI blows up or gets acquired, Microsoft will suffer no real collateral damage.
• Reduces risk and with the 2025 PER ratio now at a more reasonable 26.8x, leads me to think that there is a fairly easy 25% in the shares when market turmoil subsides.

• Not only did President Xi turn up but he also made comments which indicate that, while he may not like the private sector very much, he realises that he needs it if China is going to come close to meeting any of his ambitions.
• All that was missing was a picture of President Xi smiling and shaking Jack Ma’s hand which would have been the icing on the cake.
• President Xi led a meeting of the top entrepreneurs in China with many from the private sector including Jack Ma at which he gave the most supportive commentary in over 5 years.
• President Xi expressed in clear terms that it was necessary to remove the obstacles that have been holding up the private sector and promised a level playing field and affordable financing for private companies.
• The commentariat is unanimous in the view that this is a clear sign of a major change of heart at the top of the Chinese government which, at the end of the day, is all that really matters.
• Almost immediately, the Chinese parliament has said that it will review laws that disadvantage the private sector in another signal that President Xi’s apparent change of heart is already having an effect.
• However, this is currently all talk and it won’t be until we see real action that we will know whether this is a genuine change of heart or merely a ploy to restore confidence in China both locally and abroad.
• Given the dire state of domestic consumption, the demographic issues and the real estate crisis, I think it is most likely that this is a genuine signal that China intends to allow its private sector to thrive once again.
• It will be many years before anyone in the Chinese technology sector forgets the 5 years it spent in the wilderness at the hands of a disapproving state and so I think what emerges will be a much more disciplined environment as opposed to the craziness that existed between 2015 and 2019.
• No one is going to forget who really is in charge in China and as long as the private sector can get prosperity moving in the right direction once again, there are better times ahead.
• This is precisely what China needs if it is to come close to fulfilling its ambitions to be a leading technological and geopolitical power and now that President Xi appears to have come around, the market may begin to reflect this once again.
• However, this will do nothing for the decoupling or the ideological struggle that is being fought between China and the West and a stronger China will only make the rivalry more intense.
• Hence, RFM Research and Alavan Independent see no let-up in the Balkanisation of the Internet with global standards becoming a thing of the past.
• This means future technologies like AI, 6G, robotics, autonomous driving and so on will have one standard for China and another standard for the West with everyone in the middle being forced to choose one or the other.
• This means two smaller networks as opposed to one global network meaning lower growth and value creation for everybody in the long-term.
• However, in the meantime, the Chinese technology sector remains extremely cheap and if the Chinese state means what it says, there is a very long way to go in terms of upside in the shares of the technology companies.
• I have held Alibaba for a long time and while I am still not yet in the black, my losses are greatly reduced and I am increasingly confident that I might actually make some money.

• Both Qualcomm and Arm reported good results, but the weaknesses seen at Microsoft Google and AMD made investors skittish although those weaknesses have no bearing on either Arm or Qualcomm.
• Arm FQ3 25 revenues / Adj-EPS of $983m / $0.39 ahead of estimates of $947m / $0.34 and it guided fairly well for FQ4 25.
• Here, revenues / EPS will be $1.175bn – $1.275bn ($1.23bn) / $0.48 – $0.56 ($0.52) which is in line with the current consensus estimate.
• Meanwhile, Qualcomm FQ1 revenues / Adj-EPS were $11.7bn / $3.41 ahead of estimates of $10.9bn / $2.97.
• Qualcomm’s efforts to diversify the business away from smartphones continue to develop with automotive and IoT (everything else) growing by 61% and 41% YoY respectively while handsets grew by just 13% YoY.
• That being said, this is a slow process as handsets still make up 75% of revenues meaning that the fortunes of the business will be hostage to the fickle smartphone market for a little while yet.
• Qualcomm also guided well for the coming quarter with revenues / Adj-EPS of $10.2bn – $11.0bn ($10.6bn) / $2.70 – $2.90 ($2.80) which was ahead of forecasts of $10.4bn / $2.69.
• The medium-term outlook for both Arm and Qualcomm looks pretty good.
• In Arm’s favour is the transition from v8 to v9 which carries much higher royalties as well as the market share that Arm stands to gain over x86 and other processor architectures in PC, cloud, automotive, robotics and so on.
• In many of these verticals, customers do not make their own processors but take the processor design from Arm allowing Arm to earn more revenue per chip shipped.
• One can see this in the license revenues which grew by 14% YoY where new licenses represent a growing opportunity for Arm to earn royalties as customers move from design into volume production.
• In Qualcomm’s favour is its increasingly successful penetration of markets outside of handsets like laptops and automotive and other nascent opportunities such as The Metaverse and robotics.
• However, other results that went before spoilt sentiment and both Qualcomm and Arm’s failure to completely blow the roof off expectations led to the fickle market selling down their shares in after-hours trading.
• This is on the back of worries around demand for AI as both Microsoft and Google’s cloud divisions performed more weakly than expected.
• However, anyone who took the time to look at the transcripts will realise that this was not due to a lack of demand but an inability to build the infrastructure fast enough to keep up with it.
• This is a very different proposition and is an indication that the AI freight train is not slowing down but may actually be speeding up.
• This is why you are seeing the cloud providers increase their capex plans for 2025 with numbers now reaching dizzying levels.
• As suppliers into these markets, this is good news for both Qualcomm and Arm but the market failed to look beyond the surface.
• AMD’s miss was also not due to demand in my opinion but its (predictable) inability to take market share from Nvidia and so has no bearing at all on overall demand for AI or demand for Arm or Qualcomm products.
• Consequently, I think that demand should remain robust in the short-term meaning that both Qualcomm and Arm have a good shot at meeting or even exceeding the expectations that they have set.
• Hence, I continue to think that Arm can grow into its valuation while Qualcomm has the ability to expand the valuation that the market affords it through continued diversification.
• This means that there is more short-term upside in Qualcomm which is why I continue to hold it in my portfolio.

Apple FQ1 25 – Still no growth.

• Apple reported good results despite weakness in China but credited Apple Intelligence for the interest in the iPhone 16 which I suspect is misplaced.
• FQ1 25 revenues / net income were $124.3bn / $36.3bn slightly ahead of forecasts of $124.1bn / $35.5bn.
• iPhone sales were slightly weaker than expected at $69.1bn compared to forecasts of $71bn, but this was mainly due to China where local brands continue to make inroads and where revenues overall fell by 11% to $18.5bn
• iPad, Mac and services took up the slack and registered good growth.
• Apple noted that interest in the iPhone 16 was greater in markets where Apple Intelligence has already been launched and consequently attributed that to interest in Apple Intelligence.
• This looks like a tenuous assumption to me as features such as the improved camera, computational photography and processor have generated much more consumer interest.
• The almost complete absence of generative AI in new products at the Consumer Electronics Show earlier this month is evidence that generative AI is not ready for the consumer.
• Consequently, I think that this fiscal year and next will be years where Apple struggles to find real growth making its 2025 PER of 32.4x look pretty stretched.
• I continue to look for better value elsewhere.

Samsung Q4 24 – All about memory.

• Samsung reported a difficult set of results where the AI -memory problem continued to plague the numbers, but I remain confident that Samsung will fix the problem and regain its crown as the king of memory.
• Q4 24 revenues / operating profit were KRW75.8tn / KRW6.5tn which was broadly in line with expectations, but semiconductor underperformance was helped by other areas of the business.
• The semiconductor business recorded operating income of KRW2.9tn which was much better than Q4 23’s loss but below what it managed in Q3 24.
• Results in semiconductors remain depressed because Samsung has effectively ditched the current generation of high bandwidth memory (HBM) and is focusing all of its efforts on getting the next generation right.
• The outlook for 2025 for semiconductors remains difficult which is why the shares continue to languish at very low levels.
• There was better news elsewhere in the portfolio with an improving performance from networks but there remains space for improvements in many areas.
• This is the second time in 10 years that Samsung has faced an existential challenge in one of its businesses and in smartphones it managed to turn the situation around despite my misgivings.
• Hence, I am confident that Samsung will return to form with the next generation of HBM which will in turn allow the shares to enjoy a multiple rerating.
• The catalyst will be when it qualifies with Nvidia as a supplier of HBM which I hope to see at some point this year.
• The current forecast for 2025 is for EPS to be broadly flat compared to 2024 which is definitely below the guidance the company is giving where it is expecting some improvements in 2025.
• Even with consensus, Samsung is trading on 2025 PER of 10.9x and 2026 PER of 8.5x where I suspect there could be meaningful upside to the 2026 EPS forecast.
• I own Samsung where I am down about 8% from where I bought it and if this continues, I suspect I will buy some more.

• Meta’s Reality Labs has had a good 2024 from a sales perspective but this is because Meta has returned to seeding the nascent Metaverse market as losses in 2024 are likely to be the biggest ever.
• According to an internal memo, Reality Labs has beaten its sales targets and grown 40% YoY most likely spurred by strong demand during the holiday season.
• This is because the companion app needed to set up a Quest device was the No.1 free app on the Apple App Store on Christmas day, a crown it has subsequently lost.
• It is also worth noting that 2023 was the year of efficiency where Meta’s subsidisation of Quest devices was curtailed as a cost-cutting measure and sales weakened as a result.
• Consequently, all we are seeing is a reversion to the mean where subsidisation of the hardware has resumed, meaning that the devices have become more affordable.
• This is not a sign of a take-off of the Metaverse, but it is a sign that Meta is absolutely determined that it will not be a hostage to Apple and Google like it is in smartphones and that it will have its own platform upon which to launch its ecosystem.
• This strategy is working as Meta Horizon OS is the go-to place to develop an app or service for the Metaverse but the flywheel effect is very far from being in motion.
• This is the goal of all ecosystems and is the time when more users trigger more developers to write apps which in turn triggers more users to join and so on.
• What is happening now is that Meta is artificially growing the user count so that developers will continue to develop for the platform with the hope being that at some point, it can take its thumb off the scale.
• The problem is that I think that this is going to take a while as the barriers to the take off the Metaverse have not moved much in the last 2 ½ years.
• These remain mostly technical in terms of size, weight and comfort but most importantly, everyone is still insisting on doing their own platform.
• This creates what RFM refers to as “a surfeit of silos” (see here) meaning that the promise of a single digital environment where all users can communicate with and interact with everyone else remains unfulfilled.
• Unless there is complete interoperability between the different players, the Metaverse is not going to take off and has no chance of replacing the smartphone as the place where users live their digital lives.
• The last few years have seen some progress on the technical side in terms of visual quality and integration of the virtual and real worlds but in terms of interoperability, nothing has changed.
• Hence, the Metaverse is still on a trajectory to be a niche digital destination as opposed to the next digital revolution as its proponents suggest.
• I can see Meta making Horizons OS available to open source like Android but whether anyone is willing to become dependent on Meta for a platform remains to be seen.
• This is why the games engines (Unity and Unreal Engine) are worth watching as almost everyone uses one or the other to render the 3D worlds that they are creating and so they represent a possibility to bridge the gap between the different players.
• However, this plays out, it is likely to take some time as progress has been slower than expected and the production of a device with a decent user experience remains elusive.
• I continue not to expect this before 2026 or 2027 at the earliest meaning that mass market takeoff would not be expected before the early 2030s at the earliest.
• Hence, this remains a waiting game and a theme that remains fairly uninvestible as many of the players are already publicly listed but their performance is dominated by other themes.

• Samsung produced one of the dullest hardware updates with the s25, but the real action was on the inside with new silicon and Google AI that relegates Bixby to a taker and executor of orders.
• Samsung has released its now annual update to its flagship lineup with the s25 range with the Ultra being the highest-end edition.
• For anyone who does not have intimate knowledge of their s24, the s25 is exactly the same as the s24 with such tiny upgrades that no one is going to notice.
• However, it is on the inside where the difference can be felt where three areas are worth highlighting.
  • First, System on Chip (SoC): where the s25 is now sporting a custom version of the Qualcomm Snapdragon 8 Elite which is the first smartphone chip to use the custom Oryon CPU core.
  • This delivers a stepwise change in both performance and power efficiency which many reviewers have noticed without attributing it to the new silicon.
  • The user experience feels much faster and more fluid while the battery life improves further on what was offered in the s24.
  • The on-device AI functions (see below) also work faster and given the upgrades that are being offered in terms of functionality, it looks very likely that the computational load has had to increase to deliver that.
  • The Snapdragon 8 Elite also sports an upgraded NPU which looks to be delivering a good portion of what Qualcomm promised at launch.
  • Second, AI: which remains almost entirely Google with Bixby lurking in the background to offer orchestration with the Samsung family of apps.
  • Front and centre of the offering is the Gemini assistant which runs on the device to ensure a decent user experience but will also refer to the cloud when needed.
  • The s25 is all Google and only summons the lowly Bixby when it needs the grunt work of accessing the Samsung apps.
  • This is part of the messy plumbing that is needed when an AI assistant is supposed to be able to execute tasks on the user’s behalf like booking events and travel and adding stuff to calendars.
  • For regular Google apps and services, it will be Gemini that does the orchestration and seeing as hardly anyone uses the Samsung apps, this is just another sign of just how irrelevant Samsung’s AI is.
  • However, Samsung is doing precisely the right thing in getting closer with Google as this gives it the opportunity to offer some of the best AI on its devices before its competitors bring it to market.
  • Third, One UI 7, which is a new coat of paint to the user experience along with a few tweaks, bells and whistles.
  • While this will not be unique to the s25, it will be on the device first and it will work best there given the silicon upgrade.
  • This will help the s25 look new and different at the point of sale.
• This launch is a further sign of Google and Samsung working more closely together where Samsung pays less attention to the software and where Google offers its best software to Samsung rather than reserving it for Pixel devices which almost no one buys.
• The ideal situation for both companies is one where Google saves its shareholders billions of dollars by stopping consumer hardware entirely and one where Samsung stops spending money on software that no one uses and improves its device margins.
• This is still a long way away but, very slowly, Google and Samsung are seeing the light.
• As a Samsung shareholder, this is icing on the cake as the investment case remains predicated on a semiconductor division turnaround.

• The advent of LLMs potentially creates a dislocation in the market that has many start-ups and every beaten-down and written-off competitor of Apple and Google in the digital ecosystem wondering whether the time for revenge has finally come.
• Apple and Google have had the digital ecosystem (ex-China) market pretty much to themselves for nearly 10 years, and from this, they have been able to extract trillions of dollars of value.
• This has been based on the fact that for 10 years, the combination of an operating system and an app store has allowed them to create digital environments where users choose to live their digital lives.
• They have reached such scale that the general view is that while users live their digital lives on smartphones and tablets, nothing will knock them off their perch.
• All may not be as it seems as there are tiny signs that change is coming and if the stars line up in the right way both Apple and Google could be in trouble although there is more that Apple can do to push back (regulators permitting).
  • First, natural language which has the potential to change the way users interact with their devices.
  • In 2017 the use of voice as the man-machine interface was widely hyped by the technology industry but because the technology didn’t work properly, AI faded into relative obscurity for a few years.
  • However, when LLM-powered services like Gemini, Bing, ChatGPT, Claude and so on came onto the scene it rapidly became clear that LLMs had solved the natural language problem.
  • This means that the industry is finally in a position to keep the promises that it made in 2016 and 2017, and we may be able to use natural language as a good and engaging man-machine interface.
  • At the moment, this is being done mostly with text but there is no reason why this could not migrate to voice as the really difficult technical problem has been solved.
  • This would mean that users either talk or type to an agent which then goes off and executes the user’s request using one or more apps and returning the result via natural language.
  • The agent will be LLM-based meaning that the model weights could be adjusted by usage meaning that the understanding of the user accrues to the agent and not necessarily to the app or apps being used.
  • This is dangerous for the incumbents because if it is not their agent that is being used, then they risk losing relevance with the user.
  • This is how empires have fallen many times in the technology industry meaning that even if this notion is completely wrong, it is worth following given that it could start an earthquake.
  • Second, Search: where Google is facing the first credible challenge to its dominance and its business model in many years.
  • Google’s differentiation in search for many years has been its ability to search the long tail of the Internet as well as work out what the user is searching for even when he or she is vague or types incorrectly.
  • LLMs and the AI that they power, are visibly closing this gap and Google is no longer the only obvious place to start an Internet search.
  • This could be extremely dangerous because, search is the backbone of Google’s revenue and its market position and if we really are seeing the first credible threat, then there is cause for great concern.
  • There is no sign of this whatsoever in revenues, market share or internet traffic and I don’t think we will see any in the short-term, but there are ripples everywhere.
  • Furthermore, most of the credible competition is so small that they would have to double their user base every year for 7 years or more just to equal what Google currently has.
• These are very early signs of trouble and given Google’s panic over the last 18 months, it is clear that it is doing everything it can to ensure that upstarts are beaten with improving the quality of products and that it is the Gemini agent that will be used with Google ecosystem products and services.
• The problem is that the open-source nature of Android will make it very hard for Google to stop competing agents from working well although it should be able to block them from its own services.
• If this results in its services being used less, then it will be in real trouble.
• Apple is in a much better position because its vertical integration gives it the ability to prevent any competition for Apple Intelligence or Siri from getting a foothold on its products.
• However, the current litigious climate in the USA and especially in Europe may have something to say about these sorts of practices.
• The net result is that nothing is going to happen now or even in the short term.
• Furthermore, using natural language as a user interface either by text or voice is not a given by any stretch of the imagination.
• It will have to improve the user experience substantially to encourage large numbers of people to switch, but the ramifications of this, should it come to pass, are so large that it is worth keeping an eye on it.
• I still don’t like the shares of either Apple or Google, and this possibility only serves to increase my dislike.