I think the emulator is good enough. Apple has provided Microsoft with sufficient motivation. My 10% share assumption is for H2 2025 sell-in. The initial launch of Windows 12 (or whatever they call it) can be messy for 2-3 quarters and thesis still holds.
This is all well and good, but we seem to be forgetting a few things.
First, AMD with proven superior perf/W, battery life and perfect software compatibility hasn't been able to achieve better than 20% market share.
Second, if Dell thinks that they can price a laptop around the same price as existing higher end x86 laptops then I don't think they've analyzed it correctly. Nobody is going to pay extra for battery life when it means worse app compatibility and you will get about the same perf when running through the emulator (about 50% hit in the very best case).
Third, x86 emulation on Windows on ARM is never going to be as smooth and polished as Apple even if the perf is there (not given). E.g. See the various setup/install issues when x86 apps are installed instead of the native ARM ones. Let's not even get into the potential device driver/hardware peripheral issues. Apple is in in a completely different boat. They were committed to move, vested in making the emulation work (having a history of having done it twice before), control the hardware and have more sway on the software developers on their platform. Microsoft has none of those.
Finally, for about $1300, if you want to move away from regular Windows/x86, you can get a pretty decent MacBook Air with a much better experience overall.
I'm willing to bet that Qualcomm is quite unlikely to get to a 10% market share. We shall see.
No surprise, most of them are $1300 and up. Only Dell and HP make some effort with $1100/$1150 bulkier models. None of these hold a candle to $999 MacBook Air. Surprisingly only Microsoft is making an effort with $1000 PCs/Tablets, so what does that tell you about how much the PC OEMs want to push this? One has to wonder why there isn't there a new Surface Go laptop if Qualcomm/ARM is so much cheaper or better value for money.
The supposedly new Prism ARM to x86 emulation layer has zero technical details about the improvements. If it was a big deal, you'd expect them to talk about it a bit more.
In short, this is not a game changer. We shall see next year when nVidia and AMD get in on the game. Maybe it will make a difference if Intel doesn't get its act together by then.
QCOM claims that it has twice the perf/W of Intel, but now Intel can achieve the earlier performance at 0.6x the power with Skymont cores, which means that Intel has narrowed the gap quite a bit. Maybe QCOM will be slightly better but that will be complete wash for someone who doesn't want the software compatibility issues and is comfortable/familiar with x86/Intel.
Of course, Intel still needs to deliver these CPUs in Q3 and all this needs to be independently validated. But considering that TSMC is fabbing it and Intel has finally been backed into a corner to improve things, the chances are pretty good IMO.
Finally, we will get to see the best x86 and ARM (Apple M3) cores on the same TSMC process node for a real comparison! Exciting times.
What makes the PMICs "trash" in your own words? I don't think ball-pitch is a substantial enough justification, especially when logic board size in modern ARM-based laptops are getting smaller to make more room for increased battery, layer count is bound to go up. Costs should come down as ARM laptops proliferate. I suspect this is 3rd party VR vendors and/or OEMs not getting their cut on the power solution...
Charlie claims that "several OEMs were going to buy the PMICs and throw them away and put a cheap and more efficient off the shelf solution in it’s place". Dell leak clearly shows that the "carcass" cost has increased by $101 or ~22%. $18 of the increase is directly attributable to PCB cost. Presumably, a large portion of the remaining $83 is PMIC. What else could it be? Rest of carcass is the same.
Thanks for the reply. I won't argue the BOM costs, that part is seemingly true, but I doubt it's wholly PMIC. You can see items like Camera, LCD, and Win11 costs went up. I think there's a difference in design philosophy at play here. Dell might be the perfect example. They come from an Intel/AMD background that mandates common footprint VRs. They think one rail with ~5% efficiency improvement is the answer to the problem. They don't have the Mobile background to understand Mobile design philosophy, where Snapdragon is from. The Snapdragon SoC has 15+ rails. They wake/sleep regularly, among other behaviors. There isn't a single VR vendor that will do what QC does from PMIC perspective, and if OEM's did, they'd implement it worse from a piece meal approach, all for the sake of pure efficiency gains 2 or 3 rails. QC SoC+PMIC pairing is the vertical integration that is needed to push various performance metrics, beyond just efficiency, much higher. It's only a matter of time when QC PMICs will reach parity on efficiency KPI, or reduce rail count. I'd put it another way, if Intel/AMD approach in laptops was so good from an efficiency perspective, would we be having this conversation...only time will tell how this all plays out.
The LCD and OS are not part of carcass line item. Camera is within carcass and that is a good point. Your point on # of power rails and design philosophy is very interesting, and I think quite valid. However, in terms of the rollout, this generation was probably still an unforced error.
Suppose there are three options. #1 3rd party PMIC piece meal approach. #2 current situation that bundles smartphone QC PMIC. #3 future QC PMIC designed for laptop-level power draws.
For the upcoming launch, option #1 was probably the best from cost, efficiency, and OEM relationship perspective. Orion V2 could be option #3 and that could be the best solution from everyone's perspective. However, choosing option #2 for Orion V1 still poor in my opinion. Charlie clearly has angry inside sources within Qualcomm and several OEMs who fed him the information. Not a good look.
Even with the incredible gross margins I estimate with Orion, vast majority of gross-profit is erased by the leaked $60 subsidy QC is allegedly providing.
Nvidia coming.
Yes. The Intel 18A chip for 2025. There is plenty of market share for them too. Feeding frenzy on Intel. :)
Great write up! still not sure about the 10% market share ~ all is dependent on Microsoft's windows on ARM emulator
I think the emulator is good enough. Apple has provided Microsoft with sufficient motivation. My 10% share assumption is for H2 2025 sell-in. The initial launch of Windows 12 (or whatever they call it) can be messy for 2-3 quarters and thesis still holds.
Thanks a lot for the write up. If you are bullish on QCOM, why are you using it as a funding short as of your Aug post? Thanks
Because Apple pulled forward their modem program by 1 year. https://irrationalanalysis.substack.com/p/maxlinear-implosion-apple-modem
The revenue QCOM is about to lose from Apple punching a hole into their income statement is much higher than PC revenue ramp.
This is all well and good, but we seem to be forgetting a few things.
First, AMD with proven superior perf/W, battery life and perfect software compatibility hasn't been able to achieve better than 20% market share.
Second, if Dell thinks that they can price a laptop around the same price as existing higher end x86 laptops then I don't think they've analyzed it correctly. Nobody is going to pay extra for battery life when it means worse app compatibility and you will get about the same perf when running through the emulator (about 50% hit in the very best case).
Third, x86 emulation on Windows on ARM is never going to be as smooth and polished as Apple even if the perf is there (not given). E.g. See the various setup/install issues when x86 apps are installed instead of the native ARM ones. Let's not even get into the potential device driver/hardware peripheral issues. Apple is in in a completely different boat. They were committed to move, vested in making the emulation work (having a history of having done it twice before), control the hardware and have more sway on the software developers on their platform. Microsoft has none of those.
Finally, for about $1300, if you want to move away from regular Windows/x86, you can get a pretty decent MacBook Air with a much better experience overall.
I'm willing to bet that Qualcomm is quite unlikely to get to a 10% market share. We shall see.
We finally have some actual shipping prices -
https://www.microsoft.com/en-us/windows/copilot-plus-pcs?icid=mscom_marcom_H1a_copilot-plus-pcs_FY24SpringSurface&r=1#shop
No surprise, most of them are $1300 and up. Only Dell and HP make some effort with $1100/$1150 bulkier models. None of these hold a candle to $999 MacBook Air. Surprisingly only Microsoft is making an effort with $1000 PCs/Tablets, so what does that tell you about how much the PC OEMs want to push this? One has to wonder why there isn't there a new Surface Go laptop if Qualcomm/ARM is so much cheaper or better value for money.
The supposedly new Prism ARM to x86 emulation layer has zero technical details about the improvements. If it was a big deal, you'd expect them to talk about it a bit more.
In short, this is not a game changer. We shall see next year when nVidia and AMD get in on the game. Maybe it will make a difference if Intel doesn't get its act together by then.
FYI Intel just revealed how it plans to neuter the QCOM/ARM threat -
https://www.techpowerup.com/review/intel-lunar-lake-technical-deep-dive/4.html
QCOM claims that it has twice the perf/W of Intel, but now Intel can achieve the earlier performance at 0.6x the power with Skymont cores, which means that Intel has narrowed the gap quite a bit. Maybe QCOM will be slightly better but that will be complete wash for someone who doesn't want the software compatibility issues and is comfortable/familiar with x86/Intel.
Of course, Intel still needs to deliver these CPUs in Q3 and all this needs to be independently validated. But considering that TSMC is fabbing it and Intel has finally been backed into a corner to improve things, the chances are pretty good IMO.
Finally, we will get to see the best x86 and ARM (Apple M3) cores on the same TSMC process node for a real comparison! Exciting times.
What makes the PMICs "trash" in your own words? I don't think ball-pitch is a substantial enough justification, especially when logic board size in modern ARM-based laptops are getting smaller to make more room for increased battery, layer count is bound to go up. Costs should come down as ARM laptops proliferate. I suspect this is 3rd party VR vendors and/or OEMs not getting their cut on the power solution...
Charlie claims that "several OEMs were going to buy the PMICs and throw them away and put a cheap and more efficient off the shelf solution in it’s place". Dell leak clearly shows that the "carcass" cost has increased by $101 or ~22%. $18 of the increase is directly attributable to PCB cost. Presumably, a large portion of the remaining $83 is PMIC. What else could it be? Rest of carcass is the same.
Thanks for the reply. I won't argue the BOM costs, that part is seemingly true, but I doubt it's wholly PMIC. You can see items like Camera, LCD, and Win11 costs went up. I think there's a difference in design philosophy at play here. Dell might be the perfect example. They come from an Intel/AMD background that mandates common footprint VRs. They think one rail with ~5% efficiency improvement is the answer to the problem. They don't have the Mobile background to understand Mobile design philosophy, where Snapdragon is from. The Snapdragon SoC has 15+ rails. They wake/sleep regularly, among other behaviors. There isn't a single VR vendor that will do what QC does from PMIC perspective, and if OEM's did, they'd implement it worse from a piece meal approach, all for the sake of pure efficiency gains 2 or 3 rails. QC SoC+PMIC pairing is the vertical integration that is needed to push various performance metrics, beyond just efficiency, much higher. It's only a matter of time when QC PMICs will reach parity on efficiency KPI, or reduce rail count. I'd put it another way, if Intel/AMD approach in laptops was so good from an efficiency perspective, would we be having this conversation...only time will tell how this all plays out.
The LCD and OS are not part of carcass line item. Camera is within carcass and that is a good point. Your point on # of power rails and design philosophy is very interesting, and I think quite valid. However, in terms of the rollout, this generation was probably still an unforced error.
Suppose there are three options. #1 3rd party PMIC piece meal approach. #2 current situation that bundles smartphone QC PMIC. #3 future QC PMIC designed for laptop-level power draws.
For the upcoming launch, option #1 was probably the best from cost, efficiency, and OEM relationship perspective. Orion V2 could be option #3 and that could be the best solution from everyone's perspective. However, choosing option #2 for Orion V1 still poor in my opinion. Charlie clearly has angry inside sources within Qualcomm and several OEMs who fed him the information. Not a good look.
Even with the incredible gross margins I estimate with Orion, vast majority of gross-profit is erased by the leaked $60 subsidy QC is allegedly providing.