Wireless Semiconductors – Radio Free Mobile

Intel – Anywhere to Go?

RICHARD WINDSOR — Mon, 17 Mar 2025 03:39:43 +0000

Mr Tan’s options already seem very limited.

The recruitment of Lip-Bu Tan as the new CEO of Intel should come with an expectation of a strategic shift that increases the probability of the company being broken up but as this cannot be done for some time, one wonders how the new CEO will do anything that changes Intel’s fortunes.
Lip-Bu Tan is a well-respected semiconductor executive best known for his excellent stewardship of Cadence and he also currently serves as the chairman of Walden International, a VC firm.
He also served on the board of Intel between 2022 and 2024 resigning in August 2024 reportedly due to differences in opinion with Pat Gelsinger over the right strategy for Intel.
I suspect that Mr Tan was not the board’s first choice it is fairly well known that he wanted the position, meaning that the board could have named him shortly after Gelsinger’s exit instead of waiting for over 3 months.
However, the fact that he is an outsider has played in his favour as the market has marked the shares up 15% in a weak market simply because he is not an insider.
The problem is that the market will now be expecting large and rapid changes, and it is not at all clear how these will be achieved.
Intel essentially faces two choices which are to either invest hard and become a leader again or split the company up creating a foundry and a fabless chipmaker.
Mr Gelsinger’s strategy was to invest heavily and execute to return Intel to the forefront of semiconductor manufacturing.
Intel’s board appears to have become impatient with this strategy even though the early signs are that Intel is beginning to close the gap to TSMC in some areas.
Unfortunately, this period of investment has coincided with a difficult period for Intel where its core business is under greater pressure than ever, meaning that there is much less space on the balance sheet for heavy investments than anticipated.
Consequently, I took Mr Gelsinger’s departure as a reaction to a change of heart by the board over Gelsinger’s plan which it agreed to when he was hired in 2021.
Lip-Bu Tan was not a board member of Intel at that time but must have given the plan tacit support when he joined in 2022.
Lip-Bu Tan’s disagreements with Gelsinger and his 2024 resignation from the board are signs that Intel will go in a new direction with Lip-Bu Tan in charge.
This strongly implies that the intention will now be to split the company into two pieces, but this will not be possible for some time.
This is because the vast majority of Intel’s fabs are 14nm, 10nm and 7nm all of which use Intel proprietary development tools and not the industry standard tools sold by Synopsis, Cadence and so on.
Until these fabs are migrated to using the industry standard tools that everyone else uses, these fabs will not be able to make anything for anyone other than Intel.
This is not likely to happen before 2027 and so spinning the fabs out now into a foundry makes very little sense.
If a spin-out into a foundry is off the cards and the board no longer likes Gelsinger’s invest-and-return-to-glory strategy, then one is left wondering exactly what strategy Lip-Bu Tan will be changing.
There are definitely places where improvements can be made such as the bloated headcount and bureaucratic culture, but Intel needs changes of a far more drastic nature.
Furthermore, Intel’s rivals are not standing still and Nvidia, AMD, Arm, Qualcomm and others are already taking share which will be very difficult to win back.
Consequently, other than some tidying around the edges, I am wondering what Mr Tan will be able to achieve over the next few years.
This leaves me sceptical with regard to the 15% rally we have just seen in the share price as I am far from certain what big changes Mr Tan is going to be able to make.
I have long been an advocate of the idea that when Intel hits 10x 12-month forward PER, it is an attractive opportunity, but for the last few years, the earnings have been falling even harder than the share price.
With uncertainty surrounding just what Mr Tan is going to be able to do, does not lead me to think that this opportunity is near.

China vs. USA – Taiwan buys Insurance

RICHARD WINDSOR — Thu, 06 Mar 2025 06:51:26 +0000

A $165bn insurance policy

The more TSMC invests in the USA, the more secure Taiwan becomes because if Taiwan, for some reason, goes offline then the TSMC fabs in the USA will become the most expensive paperweights in history.
This is why I view the $100bn commitment by TSMC as being more about incentivising the USA to defend Taiwan against invasion than it is about avoiding tariffs.
TSMC has announced that it will invest a further $100bn on top of the $65bn already earmarked to build advanced semiconductor manufacturing in the USA.
This will mean a further 2-3 advanced fabs as well as facilities for back-end processing before the chips are shipped to end clients.
This is widely being seen as a move to placate the new US administration and avoid tariffs, but I think that there is more to this than meets the eye.
This is because the most complicated and proprietary parts of TSMC’s offering are the recipes (methods and machine settings) that it uses in its fabs to make chips for its clients at high yield and therefore, high margin.
These recipes are closely guarded secrets and are located in Taiwan meaning that all of the foreign fabs are effectively run out of Taiwan.
This is the equivalent of having a computer based in the USA that checks a server in Taiwan every day for the day’s updates and adjustments to firmware settings.
If for some reason the server in Taiwan goes offline, then the computer in Uthe SA will stop working and this is precisely what is likely to happen with TSMC’s fabs located in the USA.
The US administration is offering no further subsidies beyond the Chips Act which is unusual in a situation like this meaning that something else could be at play.
Even if the USA were to put tariffs on Taiwan, they are unlikely to affect TSMC very much because most of its chips are not imported into the US but are sent elsewhere for assembly into products which are then imported.
Consequently, there has to have been some other form of motivation, and I think some form of informal security guarantee is the most likely option.
This is not dissimilar to the materials deal that the USA is proposing in Ukraine as US jobs, manufacturing and ability to produce technology will be impacted should TSMC’s operations in Taiwan go offline for any reason.
Advanced semiconductor manufacturing is very sensitive to vibrations or disturbances meaning that if China were to invade Taiwan, the odds are very high that TSMC would go offline.
This would mean that the US Fabs stop working creating an incentive for the USA to ensure that this does not happen.
Consequently, the more operations that TSMC has in the USA or elsewhere, the greater the incentive is for those territories to ensure that TSMC in Taiwan continues to operate as normal.
This is of far more importance to Taiwan where the overwhelming majority does not want to become part of China even though it is by far its largest trading partner.
China has promised that it will seek reunification by any means which, with an intransigent population, can only mean a take-over by force.
The big question for China is whether the US would intervene to defend Taiwan and the more that TSMC is entwined with the US economy, the greater this probability is.
US involvement will be a big factor in China’s decision whether or not to seek reunification by force and so this looks to me like $100bn well spent.
These new fabs will be away from the Taiwan ecosystem and so the cost to make chips in the USA is likely to be higher than it is at home.
However, I expect that this extra cost will be spread evenly over all of the silicon that TSMC produces meaning that there is likely to be an increase in the price that TSMC charges to cover the extra cost of manufacturing overseas.
TSMC growing closer to the USA will cause China to increase its efforts to become self-sufficient and further accelerate the bifurcation RFM Research and Alavan Independent see coming in the technology industry.
Dual standards and multiple and incompatible networks are good for no one as the value creation from technology will be lower as a result.

iPhone 16e – Pugilistic Intent

RICHARD WINDSOR — Thu, 20 Feb 2025 04:44:16 +0000

New modem needs to succeed to re-open hostilities.

Apple has finally launched a phone that uses its own in-house 5G modem, but the real test of success will be whether anyone can tell the difference between the in-house version and the Qualcomm modem that is used in all of the other iPhone 16 models.
If the new modem works well, then the way will be open for Apple to restart its patent fight with Qualcomm but if Apple decides it needs 6G early, then it will be back to square one.
The new iPhone 16e is everything that one would expect making hardware compromises here and there to ensure that Apple can still earn an excellent gross margin on a $599 price point.
One camera, a smaller screen and no dynamic island are compensated for with stellar battery life, efficiency and the A18 chip which is present in its big brother.
This means that almost all of the software features of iOS18 will work just as well on the 16e including all the features of Apple Intelligence.
Hence, Apple is really pushing the battery life with a claim of 16 hours of video playback and up to 12 hours more battery life than older iPhones.
Here, the comparison is being made against iPhones 11, 12, SE (2^nd Gen), SE (3^rd Gen) and the 12 mini highlighting that Apple is mainly going after the users of older iPhones who are more cost-conscious.
It does not compare particularly favourably with Android as at this price point, users get a lot more hardware bang for the buck.
Instead, Apple is trying to make up the gap with software (which for older iPhone users it does) but against Android users, I think it will struggle.
The real test here is the in-house C1 modem which finally makes it to market after many years of trying.
This means that Apple thinks that it has worked out how to make a 5G modem that works just as well as Qualcomm or MediaTek when it comes to 5G, but no amount of band analysis, lab testing or specification comparison will determine whether it is good enough.
The acid test will be whether users notice any difference in connectivity compared to other devices that use Qualcomm or MediaTek modems because it is this that will determine whether the C1 and its successors can replace Qualcomm entirely.
The place where this will be noticed will be where there is poor coverage or network congestion as there is nothing more frustrating than seeing your device fail to operate when everyone else is continuing as normal.
In many ways, the iPhone16e is an experiment to see whether the C1 modem is ready to roll out across the rest of the range and if it is, then Apple will be free to restart legal hostilities over the patent royalties that it pays to Qualcomm via its contract manufacturer.
Apple gave up its fight against Qualcomm over patent royalties in 2019 because it needed a working 5G modem and at the time had nowhere else from which to source one (see here).
It also signed a relatively short patent agreement (usually 10 years) as it thought that it would have a working modem by the end of that agreement and be in a position to reopen patent hostilities.
However, the 5G modem proved much harder than expected with several false starts meaning that it is only now that Apple thinks that it is in a position to compete with Qualcomm on modem quality.
If this is successful, then I do not doubt that it will roll the C1 and successors quickly through its portfolio and refuse to sign a new patent licence when the current one expires in March 2027.
This will set the stage for yet another fight over royalties but by then 6G will be starting to make an appearance.
Crucial to the patent fight will be whether Apple decides that early access to 6G is important for its device portfolio.
I am pretty sure that when it comes to 6G, Qualcomm will be the first to produce a commercially viable modem (as it has in every generation since 3G) and so if Apple decides it needs 6G early, it will again have nowhere else to go.
Hence a decision to go with 6G early or the users noticing that the iPhone16e has inferior connectivity in difficult wireless environments could scupper Apple’s desire to have another go at paying lower royalty rates to Qualcomm.
Either way, it is still much too early to know how this will pan out, but Qualcomm is already planning for a fall in modem share from 100% today to just 20% of the models that are launched in 2026 and it has communicated this clearly to the market.
Hence, I think that Qualcomm is taking the sensible approach of planning for the worst but hoping for the best meaning that the balance of probability is that it does better than it has assumed.
However, the real question is whether Apple will be in a position to re-open its patent fight and it is the performance of the iPhone 16e that it is likely to decide the matter.
With relatively conservative long-term guidance and increasing penetration of new markets such as laptops and vehicles, I don’t think that this is going to trouble Qualcomm’s fundamentals too much.
Hence, I am quite happy to continue holding the shares.

Arm – Balancing Act

RICHARD WINDSOR — Fri, 14 Feb 2025 07:15:34 +0000

Arm will need a major charm offensive.

It turns out that Arm may be looking to make its own chips after all in a move that needs to be carefully balanced to ensure that it does not enrage its customers and accelerate the shift away from Arm to RISC-V.
The Financial Times (see here) and Reuters (see here) are citing unnamed sources (not very reliable) that claim that Arm intends to launch its chip offering, is recruiting staff from its customers and may have already won business from Meta that otherwise could have gone to Qualcomm.
This is not the first time I have heard this as Qualcomm made similar, but less specific, allegations as part of the court case that concluded at the end of last year.
As this possibility is coming from multiple sources and given that the FT is generally reliable, I think it is likely that Arm will launch a chip this year as the FT claims.
From Arm’s perspective, this strategy makes complete sense because the company has flown high on AI wings and now needs more revenues and profits to sustain this altitude.
Making its own chips will greatly increase the revenue per chip but it will bring margins down.
However, overall, I would expect to see a significant increase in long-term financial performance assuming that this strategy goes well.
For Arm, chipmaking is a high-risk strategy as:
- First, customers: where Arm will now be directly competing with its customers which will both anger them and compromise Arm’s independence in their eyes.
- According to the FT, Arm has secured Meta as a customer for a data centre CPU in what I would guess is a fully customised implementation based on Meta’s requirements.
- Qualcomm is likely to have offered Meta something similar but I suspect that the Arm offering could well have a greater level of customisation as it has no off-the-shelf designs to use and as the first customer, Meta probably got an excellent price.
- I suspect that Meta will be dual sourcing as Reuters sources claim that the talks with Qualcomm are still ongoing which makes sense given that this is Arm’s first foray into chipmaking.
- There is currently no love lost between Arm and Qualcomm at the moment but this will also concern all of Arm’s other customers with the exception perhaps of Apple who I don’t think will care very much one way or the other.
- Competition with clients could intensify further if SoftBank acquires Ampere which could greatly accelerate Arms migration to becoming a chipmaker.
- If Arm ends up offering only highly customised chipsets which are co-designed with customers, then the competitive issue will be significantly lessened but it is not clear at this stage which route Arm will be taking.
- Second, RISC-V: which will receive another boost if Arm begins selling complete chipsets rather than just IP and processor designs.
- This is because Arm’s customers will see a loss of independence and have concerns about whether they will be treated fairly given that there is an in-house competitor.
- Samsung has proven for years that it can keep its memory business independent from its handset business but convincing chip customers of this will not be easy especially when its relationship with one of its largest customers is in such a bad state.
- Hence, this is likely to push Qualcomm, Nvidia, Samsung, MediaTek and so on to look more closely at RISC-V which already has a lot of traction but only at the very low end of the market.
- RISC-V is great for the low end because chipmakers don’t have to add very much to get a working product, but at the high end, Arm’s economies of scale make it not cost-effective to do RISC-V.
- However, this is slowly changing because just as Arm is eating into the high-end market, RISC-V is eating into the bottom end of Arm’s addressable market.
- Google is already migrating wearables that use Android to RISC-V and a move like this will encourage chipmakers to accelerate the migration to RISC-V meaning that smartphones may migrate sooner than expected.
- This is not a big problem for Arm for as long as it can take share from x86 and proprietary processors in the vehicle, data centre and so on but at some point, it will run out of road.
- I am also pretty sure that Qualcomm is already working out what it would need to do to build an Oryon CPU on RISC-V and this move will only increase this incentive for all of Arm’s customers.
The net result is that if Arm is going to build its own chips (which seems likely given the number of sources) it needs to go on a charm offensive and make the case to its customers that its independence will be maintained.
Failure to succeed will see customers accelerate their RISC-V programs and give the overall RISC-V ecosystem far more air to breathe than it currently has.
If RISC-V hits critical mass and standards for higher chip functions begin to form and an ecosystem is created, this will greatly increase RISC-V’s ability to compete with Arm.
Arm currently has complete dominance in many verticals and is growing share but for this to continue with Arm as a chipmaker, it needs to strike a delicate balance between maximising its revenue opportunity, keeping its customers happy and ensuring that RISC-V stays in its box.
Making its own chips is a high-reward but high-risk proposition.

Qualcomm vs. Arm – Short but Sharp pt. III

RICHARD WINDSOR — Fri, 17 Jan 2025 01:44:41 +0000

Qualcomm looks to be off the hook until 2033.

As details emerge following the end of the trial in December, documents confirm that Qualcomm does have a license for V9 meaning that it is in a stronger position than I had previously anticipated.
The jury in the trial decided that Qualcomm is not in breach of the Nuvia contract with Arm and that Qualcomm’s Orion processors are properly licenced under Qualcomm’s existing architecture licence with Arm.
However, the jury was deadlocked on the question of whether Nuvia breached its licence with Arm, and it is on this basis that Arm could seek a retrial.
Arm has said that it is seeking a retrial but it is not clear whether it is seeking to relitigate all of the jury findings (i.e. appeal) or just the one where the jury was unable to come to a verdict.
My view at the time (see here) was that this was a victory for Qualcomm but because it did not have an architecture licence for the latest instruction set (v9), it would have to go back to Arm before the expiry of its current agreement which for all practical purposes is in 2033.
The contract expires in 2028 but Qualcomm has the option to extend this until 2033 which as things stand today, I am certain that it will exercise.
Documents that have subsequently been made available are pretty clear stating that Qualcomm and Arm added an architecture licence for v9 on June 23, 2020.
Assuming that the original two verdicts are not overturned on appeal, this means that Qualcomm can migrate its Orion custom CPUs to v9 without having to enter into any further negotiations or agreement with Arm.
Hence, Qualcomm should be able to continue to ship all of its arm-based products without having to enter into any negotiations with Arm for another 8 years.
However, there is the issue of v10 which is the next version of the Arm instruction set which I expect will be released by Arm sometime around 2028 or 2029 based on Arm’s history.
If there are new features in v10 that necessitate a quick move to v10 in order to remain competitive, then Qualcomm will have to take a v10 license before 2033.
I do not think that this will be the case as there are areas of overlap between the last versions of one instruction set and the early versions of the next one.
This is how I expect Qualcomm has been able to come up with very competitive products using v8 even though v9 has been around for more than 5 years.
If Qualcomm does need v10 before 2033 then it will be in a difficult position because Arm’s design is not an official standard and so it does not have to adhere to the fair, reasonable and non-discriminatory (FRAND) principle that governs the licensing of standard essential patents.
Hence, Arm can be as unreasonable as it wants when it negotiates with Qualcomm which will put Qualcomm in a difficult position in 2033 or earlier if it turns out it needs v10 before that time.
Arm says it intends to seek another trial, but I continue to think that this is not in the best interest of the Arm ecosystem.
Arm and its customers should be working together to expand its reach into PCs, vehicles, servers, AI and so on and not wasting time and resources fighting each other.
Eight years is a long time and so I am hopeful that there will be a negotiation at some point where everyone gets something, the two companies kiss and make up and get back to the business of driving the Arm ecosystem.
I continue to think that another 8 years of frozen dialogue and no cooperation will only benefit Arm’s rivals such as RISC-V and x86.
This is crucial as the standards for AI are beginning to form and with Intel on its knees, the time to push the Arm ecosystem hard is now.

USA vs. China – Carrot and Stick

RICHARD WINDSOR — Tue, 14 Jan 2025 04:08:48 +0000

I expect the new rule to stand.

The USA has released new export rules that make good on the intentions released a few months ago that make life harder for China and incentivise other countries to join the struggle.
I think it unlikely that the new administration will seek to overturn this new rule but will instead build upon it over the next four years.
I also think that this has very little to do with military access and everything to do with limiting China’s rise as a technological and political power which has been vocalised multiple times by the incoming administration.
The new rule creates a three-tier system to determine what licenses are required for the export of advanced AI semiconductors.
The first tier consists of 18 countries that are all allied with the USA consisting of Western Europe, the UK, Australia, New Zealand, South Korea and Japan which will face no restrictions.
The second tier is the countries whose relationship with tier 3 countries is less clear and exports to these countries will be capped and will need a licence from the Department of Commerce to go over that cap.
Tier 3 consists of the now familiar group of countries considered to be hostile to the USA such as China, Russia, Iran and North Korea to which exports cannot be made.
I see two main impacts of this rule:
- First, loopholes: which have been exploited to get around the previous restrictions.
- The main loophole is re-export to blocked countries from countries that are not on the blacklist which has been significant in recent years.
- This is why there is a volume cap (estimated in-country demand) with a license needed to go over that cap (i.e. explain why you need more than we think you do).
- This rule should make some headway in closing this loophole.
- Second, the carrot: which was spelled out by the Department of Commerce in September of last year (see here).
- While the USA has been able to unilaterally control exports of advanced semiconductors, it has had to negotiate with Japan and The Netherlands to extend the restrictions for 20nm – 10nm technology.
- With quantum computing and other new technologies, more countries will be needed to keep exports to China at a minimum and so there has been a switch of strategy from the stick to the carrot.
- The USA appears to looking to incentivise cooperation rather than threatening with the stick of the Foreign Direct Product Rule (FDPR see here).
- The idea here is that those who join with the USA in posing harsher restrictions on China will find it easier to source technology and talent from the USA as well as presumably trade with the USA.
- I suspect that this is what the list of 18 countries is all about and these are the ones that are expected to co-operate with the USA in aiding the prosecution of its China policy.
- By swaying some of the less-willing participants, the strategy to limit China’s rise as a technological and economic power will be much more effective.
I suspect that over time, the scope of what is covered by the rule will expand and countries will be moved up and down the list based on co-operation with the USA.
The timing of this rule is poor as it looks like the old administration is trying to make things hard for the new one, but I don’t think that this is the case here.
The one thing that the two sides of the aisle agree on is China and the Biden Administration has been more hawkish on China when it comes to execution of policy than the Trump administration that preceded it.
In the short-term this could hurt demand for AMD and Nvidia’s chips and with half of the EU excluded from the list one can understand the concern expressed in the joint release from Nvidia and the EU.
However, I don’t think this will hurt medium or long-term demand as the datacentres will simply be built elsewhere assuming the appetite remains as insatiable as it is today.
Hence, I don’t think this harms the investment case for Nvidia or any of the others, but it does mean more bureaucracy to make sales that to date have been easier.
The only real losers here are the tier 3 countries who will find it even harder to access advanced AI chips meaning that their AI development will fall further behind.
This does set them up to fare better when the endless money being pumped into AI dries up as they are already being forced to do more with less but for the moment, they are falling behind.
This will deepen the decoupling which is now well underway and divide the technology sector into two distinct and incompatible pieces.
The Balkanisation of the global network inevitably means less growth for the technology sector in the long term.

Qualcomm vs. Arm – Short but Sharp

RICHARD WINDSOR — Wed, 18 Dec 2024 07:01:25 +0000

I still think the dispute will settle.

The dispute has gone to trial with both sides scoring points, but the outcome remains pretty uncertain where the outcome of this trial is likely to form the basis of a settlement.
Much of the detail of this legal fight is fairly confusing but it essentially boils down to a fight over royalty rates.
However, even this is confusing because Arm has not asked for a remedy that involves money but is only requesting that Qualcomm be forced to destroy products that are based on the technology it acquired from Nuvia.
In Arm’s opinion, this means everything that uses the Orion custom processor, but seeing as the X series of laptop processors, Cockpit Elite, Ride Elite and 8 Elite processors were all developed at Qualcomm after the acquisition had closed, it follows that these chip designs should not be covered under such an order.
It is these disagreements over what the contract between the two companies means is what this case is supposed to resolve.
Once that is determined, there will be further legal wrangling over how the finding should be implemented.
The case will be short as each side has a very limited amount of court time to put their case and so it is likely that the jury will retire before the end of the week with a verdict shortly after.
An exchange that characterises the nature of this case occurred when Arm’s lawyers questioned Qualcomm’s chip designer Gerard Williams.
Mr. Williams repeatedly stated that the contract did not mean that all of Nuvia’s work was a derivative or modification of Arm’s technology and that only around 1% of the IP in Nuvia’s chips met that definition.
Arm’s lawyer immediately countered with “Maybe you wouldn’t say that, but that’s what the contract says”, to which Mr Williams responded, “I wouldn’t say that, but I am not a legal expert”.
It was at this point that Arm’s lawyer wrapped up her questions.
It is perfectly clear to any reasonable observer that only a proportion of a product built using Arm’s instruction set is Arm IP, but the problem appears to be that it is not hard to infer from the contract that the opposite is true.
If this were the case, then one would expect that all of the products made using Arm would be fairly similar (like AMD’s version of x86) rather than the huge range of specifications and differences observable in the Arm ecosystem today.
It is this that the court will decide and here, the outcome is very uncertain.
Whichever way this goes, the immediate result is likely to be an appeal, but I am hopeful that the verdict will establish a position from which a settlement can be reached.
This is in the best interest of the Arm ecosystem as Qualcomm is one of its biggest participants and I think that both companies would be best served by working together to take share in new markets such as PCs, Automotive and Cloud rather than fighting each other.
Qualcomm is clearly quite confident of victory as this is not the first time that it has been to court to fight over IP and so it has far more experience in judging legal cases of this nature than Arm does.
However, at the same time, this is an unusual case for Qualcomm, so despite its large experience advantage, I am far from certain which way this will go.
I think it extremely unlikely that Arm will be able to force Qualcomm to destroy its IP and it is not in Arm’s long-term interest to do so, meaning that this remedy looks to me like a negotiating tactic to extract higher royalties and set a precedent for the rest of the industry.
The most likely outcome is one where Qualcomm pays higher rates for its Orion processors to Arm but less than it would have done under the Nuvia licence.
Either way, Qualcomm’s processors are here to stay meaning that the diversification strategy outside of smartphones remains intact and with it, the investment case.
If Qualcomm’s shares crash after a loss, this would be a good opportunity to have another look at the stock for those who missed it when it was around $100 per share.

Apple vs. Qualcomm – Black Magic pt. III

RICHARD WINDSOR — Mon, 09 Dec 2024 06:51:13 +0000

This fight is becoming less important.

Apple is having another go at removing Qualcomm entirely from its products, but I suspect that this will continue to be much harder than expected meaning more revenues and profits for Qualcomm but by the time Apple does finally succeed, it will matter far less than it does today.
The Apple and Qualcomm dispute goes back many years and is almost identical to the fight that Nokia and Qualcomm had over patent royalties nearly twenty years ago.
While I have always thought that Nokia got the better of Qualcomm the first time around, this time it was clear that Qualcomm won a crushing victory against Apple in 2019.
This was because Apple decided that the iPhone needed a type of 5G that only Qualcomm could make and as the time for launch came around, Apple was forced to settle to get access to the modems that it needed.
It seems that no one can bear a grudge quite like Apple and ever since that day it has been determined to rid itself of Qualcomm silicon which would give it the ability to reignite legal hostilities over royalties without putting the iPhone business at risk.
Apple subsequently acquired the modem business from Intel and has been trying to make its own 5G modem over since even though it probably makes almost no economic sense to do so.
The original plan was to have working silicon to remove all of Qualcomm’s chips by 2024, but Apple is a long way short of this target.
The problem is relatively simple in that the cellular radios are very difficult to make and take years of practice to perfect which Qualcomm has pretty much been doing since 1985 when it was founded as Quality Communications.
The general assumption on the street is that because Apple is arguably the best processor maker in the world, it should have no trouble replicating the chips that control cellular communication.
Anyone who has been around for a while understands that when it comes to cellular radios simply being good at designing chips is not nearly enough.
Dealing with analogue radio systems, moving standards and a dizzying array of different radio frequencies is fiendishly difficult where often experience is more important than raw engineering talent.
This is what Apple failed to understand when it acquired the second-rate radio modem business from Intel in 2019 which originally came from Infineon and had been floundering for years.
Solving these kinds of problems is often referred to as a black art and Apple has struggled with this part ever since.
However, it looks like Apple will finally launch an entry-level iPhone (economically insignificant) with its own silicon in 2025 and if it works well, the plan will be to slowly replace Qualcomm like it replaced Intel (see here)
This assumes that Apple suddenly gets good at wireless communications, but the issue here is that this is a moving target as the standards continually evolve and Apple is not a meaningful contributor to any of the committees that select the IP that is used in the standards.
This means that it has to follow where everyone else goes which will make it very difficult for it to be a leader.
Hence, I think that for as long as the standards continually migrate and change (6G is on the horizon), Apple will not be able to replace Qualcomm in its highest-end products which will need to have the latest radios with the best performance.
Furthermore, even if I am wrong, by the time Apple can match Qualcomm, the non-smartphone businesses will probably have become big enough that the loss of Apple’s modem business is a small blip that doesn’t meaningfully move the needle.
The problem is that the market is obsessed with Qualcomm’s Apple business even as it becomes less and less significant meaning that every time stories like this emerge, the shares take a hit.
Wobbles like these look like opportunities to add or initiate a position in the shares I have already done and remain happy to sit on.

Samsung Electronics – Efficient Market

RICHARD WINDSOR — Wed, 27 Nov 2024 08:09:52 +0000

Priced for a Nokia.

The case for Samsung Electronics is purely based on recovery in semiconductors because if things stay the way they are, there is little to no upside in the shares.
Following Samsung’s second horrendous mistake in 8 years, the shares have given up 36% of their value but as the problem now resides in its core Semiconductor division, the market has treated the shares unfairly.
Samsung was the king of both memory and storage but unfortunately, it made a mess of high-bandwidth memory (HBM) which has allowed Micron and SK Hynix to win a rare victory.
The size of today’s AI models and the amount of data being used means that the speed of the memory and the transport of information between GPUs is a crucial factor in training AI cost-effectively.
This is why memory has become so important in AI explaining the explosion of demand for memory that operates at very high speed.
However, getting this type of memory right is no easy task and Samsung is still struggling with heat and power consumption problems with its silicon meaning that it has yet to qualify as a supplier to Nvidia.
Hence, Samsung is unable to address 85% of the fastest-growing market in memory and its results in 2024 have underperformed both SK Hynix and Micron as a result.
Samsung has apologised to investors, shuffled management and most recently has promoted the new head of Samsung Semiconductor (Jun Young-hyun) to Co-CEO.
This is a signal that it is deadly serious about a turnaround and Mr. Jun now has executive authority over other parts of Samsung to push through a number of changes.
Some of these changes involve cost reductions to boost profitability and also changes to its corporate culture to ensure that blunders like the one currently being experienced do not reoccur.
This is the second major blunder in 8 years with the first blunder being with its Note 7 smartphone where the batteries were found to be defective and many of them caught fire forcing Samsung to recall all of the Note 7s that it made.
At the time, I looked at Samsung and compared it to Nokia, Motorola, BlackBerry and so on and I concluded that the handset business was at dire risk of the death spiral that killed off so many of its rivals.
However, contrary to my expectations and against great adversity, Samsung fixed the problem, halted the market share slide and returned to being the major Android handset vendor with reasonable profitability despite brutal competition from Chinese rivals.
I see the current situation with HBM Memory being very similar to what happened in 2016 and given its history, I think that there is a very good chance that Samsung fixes this problem, qualifies for Nvidia and takes a good slug of market share at Nvidia and other AI customers.
Crucially, none of this is priced into Samsung’s current share price at KRW56,000 per share.
If I assume that the situation remains unresolved and Samsung fails to win any business in HBM, then Semiconductor margins will remain in the low 20s which is low for a company of this calibre.
If the problem is fixed and Samsung can regain its crown as the king of memory, then I can see semiconductor margins coming back to the mid-30s.
I think that this is fair as Samsung Semiconductor has achieved much higher than this in the past and so a lower margin but over the cycle looks like a reasonable estimate to make.
If semiconductor margins remain in the low to mid 20’s then a DCF calculation with a 10% WACC gives a value of around KRW60,000 per share but mid-30s at the same cost of capital gives KRW88,500, 58% above where it is today.
Essentially, the market is pricing in no recovery at all meaning that the risk-reward ratio on Samsung Electronics is very favourable.
The shares are also very cheap based on PER with 2024, 2025, and 2026 at 10.5x, 8.5x and 6.9x respectively.
However, I suspect that should the recovery fail, then Samsung shares will not be re-rated leaving them as a classic value trap.
I am inclined to take a position in Samsung via the GDR as I see little downside but have not done so yet.

Qualcomm – Confident Outing

RICHARD WINDSOR — Wed, 20 Nov 2024 07:05:11 +0000

Low-balled numbers still give good growth.

Qualcomm used its 2024 investor day to focus on the areas outside of smartphones where the diversification strategy that was laid out in 2021 is starting to bear fruit.
It also gave some targets for its businesses outside of smartphones which, in some cases, look pretty conservative to me and could be handsomely beaten.
The strategy is simple in that the company identifies a market vertical which is going through a dislocation such that connectivity and processing are becoming important, and then adapts its silicon and builds software to help clients deal with the dislocation.
Qualcomm also attacks markets where vertical players use the technology barrier they have constructed to keep competition at bay and brings those barriers down with its horizontal approach.
This is exactly what it did in smartphones and, combined with MediaTek, it has enabled the Chinese smartphone makers to dominate the Android market.
This is currently being repeated in ADAS and I suspect that in time we will see it in The Metaverse if the segment ever takes off.
This has worked extremely well in automotive and The Metaverse and is now being applied to both PCs and Industry.
- First, Automotive, where Snapdragon Digital Cockpit has become an industry standard, and Snapdragon Ride is rapidly taking share from Mobileye.
- These chipsets directly address the migration of vehicles from being hardware-defined to software-defined as well as the slow progression towards being able to drive themselves.
- Here, Qualcomm is targeting $8bn in revenues in 2029 up from $4bn in 2026 and its $45bn pipeline covers 80% of this target.
- Second, PCs where Windows-on-Arm finally delivers on the promise of excellent performance and great battery life rather than having to make a compromise.
- Qualcomm now has just over 50 design wins in this segment up from around 20 when it launched which is expected to grow to 100 design wins by 2026.
- Qualcomm estimates that 90% of laptops will be AI PCs (i.e. have an NPU) by 2029 of which 30% to 50% will be non-x86 products.
- This gives a $4bn revenue target in 2029 which at the mid-point of these numbers and 75% share of non-x86 PCs, gives an ASP of $60 per chip which looks pretty low to me.
- Hence, I think that there is a lot of margin for error in this estimate which is not unexpected given how uncertain the speed of change can be.
- I see space for a big outperformance of this forecast.
- Third, Industrial which has had some difficulty recently, but where Qualcomm now intends to perform a rinse and repeat of Automotive.
- This will not be easy, given how fragmented and different all of the verticals are, but having Saudi Aramco as a cornerstone client is a big vote of confidence.
- Both Honeywell and Deloitte chimed in to support Qualcomm’s Industrial offering.
- The target for 2029 revenues is $4bn which again leaves quite a lot of wiggle room as business very rarely progresses as one expects.
- Fourth, The Metaverse which is where I think the real risk to the forecasts lies.
- Qualcomm contends that the ability to use natural language as the man-machine interface greatly reduces the shortcomings of using a head-mounted display which will help The Metaverse to win adoption.
- RFM Research generally agrees with this view, but there are a lot more barriers that need to be overcome for take-off of The Metaverse and a 30m annual unit shipment target in 2029 is a punchy estimate on that basis.
- Apple’s struggles in this market and the heavy subsidisation that is required to sell units are indications of just how immature it is and that the technology is far from ready for the market.
- However, if it does take off, then Qualcomm is in pole position as it is pretty much the only supplier to this market and it has 100% share at the leading supplier, Meta Platforms.
When one adds all of this up one gets $22bn in revenues in 2029 up from $8.3bn in FY2024 which assuming mid-single digit growth in handsets gives 41% of revenues in 2029 coming from non-handset sources.
This means that the 50% non-handset revenue in 2030 is unlikely to be met, but this is an aspiration rather than a hard target and it looks like the company will get there shortly after 2030.
I continue to think that most of these numbers have not been incorporated into financial forecasts and so I think it likely that long-term FY2027 and beyond estimates will now start to rise.
This means that growth of 10% or better for the next 5 years looks reasonably safe and with operating expenses stable as a percentage of revenues and cash being returned via buybacks, EPS should grow somewhat faster.
Even without further multiple expansion and upside surprises, the shares should deliver a steady return over the next 5 years leaving me very happy to sit on my position.