Great read, I think the vision for Mark in the last 5 years been precieved as bearish more than bullish. If we remember during the pandemic he started diverging into metaverse 3 which crypto fanatics say will be the new internet. He then proceeded to pour tons of FCF into this idea which he then changed his mind of as he could see the stock crashing based on investors fears that Zuck has gone mad. This made investors extremely bearish and caused the stock to crash tremendously but again he maganed to dampen shareholders fears and go back to his original framework of letting Facebook be Facebook and built different aspects of the business this investment turned out to be well rewarded as investors became bullish again and the price rises to ATH. To disclose I have brought the dip under 600$ and a few days later he said he was going to slash 30billion of his original commitment into AI spending. This gives us a better understanding into his personality he does things and sees how his investors reacts and then can retract just as fast. My thesis is this there are 3 billion people using Meta and this will continue to grow regardless of China which is 1.7 billion people that means out of the world there are another 2 billion people he can still capatilise on even if he does not net the full 2 billion and gets 1 extra billion on the platform I think he can absorb and justify the capex spending on AI. These are my thoughts any feedback would be great
I agree we shouldn't put too much value on Zuckerberg's initial comments, as he tends to retract and pull back, but I have to admit I am not a fan of this strategy. I want to be able to count on management's comments and commitments, and I can't here.
Realistically, I also don't think there are another 2 billion people that can adopt Meta. Global population growth is stabilizing, no longer growing aggressively + we have to consider minors + potential social media bans on youth coming in the future, similar to Australia.
1. The $600B is mostly a headline grabber and very likely includes opex, not just capex. Their run rate opex + capex in FY25 is roughly $200B so $600B cumulative spend by 2028 is not materially different. Moreover, aside from lease commitments, they can always back down from the biggest capex bucket, chips.
2. Probably unfair to say they have no visibility to returns on this spend. As you highlighted, they saw a 10% growth in time spent on Facebook and Instagram as a direct result of their investments. Clearly, there's upside and visibility. We don't know what the split between core ad capex and gen AI capex but management can pivot as they've said.
3. I agree I'd categorize this spend as defensive. I agree defensive probably leads to worse outcomes in more cases than not but META proved to be the exception just a few years ago when they were forced to invest in AI and short form video by Tik Tok and Apple. I held a similar opinion to you a few quarters ago that they don't need a leading model if they can't directly monetize it like the other CSPs but I've come around to the idea a leading AI model may be a key differentiator to serve better ads in an AI world. They may be forced to catch up or risk losing engagement/ad dollars to Gemini/ChatGPT over time. We'll have to see but it's certainly a higher probability outcome than before.
I generally enjoy and agree with your breakdowns and takes but wanted to offer some counters here
I am very glad with your counters, Jason! These discussions are always great, leading to new insights + I can always miss some things or misinterpret facts. So, thank you for taking the time. Let me respond to your points!
1. I think you are definitely at least partially right. To quote management, it is committing to $600 billion in U.S. investments in AI, consisting of infrastructure and jobs. So, I agree there is probably an asterisk on that number, since it includes CapEx and growth in operating expenses. However, I don't think we can take run-rate OpEx here, as I have to assume it includes AI-related expenses, primarily driven by growing headcount and operating costs for these data centers.
This leaves a lot of CapEx, which is tough to support solely through cash flows. The recent bond issuance speaks for itself.
And yes, management can back down from a large part of these commitments, which is my bull scenario indeed. However, Zuck has a history of commitment to its vision. Take reality labs.
2. "We don't know what the split between core and capex and gen AI capex is, but management can pivot as they've said." I think you have a bit more confidence in management than I do, haha. Are you saying Meta might have enough compute needs in its core business to justify this level of spend and expansion? There must be an alternative market here it aims for. And I can't see it being AI chatbots/assistants.
I would love to hear their monetization plans, because I can't see them. That is my problem here.
How do you view this? In what scenario, assuming it stays committed to current numbers, do you see a good ROI? What am I missing?
I am certainly not claiming I am right here 😁
3. So better ads is your bull thesis for this spend? Does it need this amount of infrastructure? Compete this hard for talent? Wouldn't it be a much cheaper, better option to use models from an external partner and remain a cash-generating machine?
I just can't see "better ads" as the main motivation for this spend. But, again, I might be wrong.
Thanks again for the thoughtful comment. Happy to discuss further!
1. I agree it'll probably be more capex but wording suggest some opex is included in that number. I believe this $600B number first came out of that dinner with Trump when he was placed on the spot. So I wouldn't spend too much time on this number was my point. The specifics around this number are purposefully vague from Meta for a reason.
2. Meta has a history of setting large opex targets for the year and lowering it throughout the year. Happens every year like clockwork, except for this year. You're right with Reality Labs - they've mostly maintained their spending as guided.
I'm not saying they can utilize all of the compute for the core ads - almost certainly they are wasting some money on speculative gen AI stuff that won't pan out. I have been skeptical AI chatbots (in business messaging) will work here but what they're saying about it working in Thailand and Vietnam jives with my personal experience in China (on WeChat). There's a chance it'll work but Meta AI needs to be better.
3. Correct. I was pretty concerned coming into the year about this year's step up in capex because I thought the low-hanging fruit from the ramp up of Reels was mostly harvested. Instead, impression growth accelerated this year. I had some concern this might be due to Meta jacking up ad loads but Q3 results suggest this is all organic. We're still quite early in Advantage+ and automation (video generation is clearly early). The improvements to GEM, Andromeda, etc. are also notable. Again, I definitely think a lot of the planned capex will be wasted. But we're seeing the fundamentals accelerate this year on higher capex. Full disclosure, I trimmed my META after Q2 but I think the narrative is starting to shift too far bearish. I see a lot of comments suggesting they have no way to monetize this capex but actually the opposite is true - the only profitable AI capex has been ads-related (META and GOOG). Zuck is just a bad spokesperson for ads - it's clear he just wants to talk about Metaverse, Gen AI, whatever captures his imagination. Susan Lee has done a great job replacing Sheryl on the calls but doesn't have the same command yet.
I don't think they can use an external model. Zuck wants to own his own platform and there are strategic arguments why it makes sense. The key differentiator between META and SNAP is they own their infrastructure. If they license Gemini, probably more value will accrue to GOOG, which will erode their margins and differentiation over time anyway. They have to do this.
2. Indeed, curious about their monetization ideas, but I am still afraid they are well overbuilding their own needs.
3. I am curious about how Meta will use this computing capacity in its ad business. Maybe I am underestimating the opportunity here. Time will tell.
I agree it makes sense to own the models in-house, but I am afraid Zuck is overly focused on owning the next platform and this kind of stuff in-house, up to the point where he will overspend on it. The same goes for Reality Labs, which is built on the same idea: owning the next-gen platform. It's very defensive spending, which then makes me doubt the future of the business.
Hi Daan, I just wish to add that you have a gift of distilling a mountain of information into a short cogent write-up that communicates the key points very clearly. And a free article at that!
I believe competing on LLMs is a bad idea for META. But AI itself is important for the core business in terms of personalized ads targeting. And if the Reality Labs business succeeds and smart glasses become mainstream, they will need AI for the user experience.
Agreed, AI is crucial to Meta's core business, but it doesn't need this kind of compute for that. Yes, if Reality Labs succeeds, these AI efforts could prove incremental and a massive opportunity.
However, these are too many "ifs" to pay this price for.
The bear case on Meta today in my view says that you, the bearish investor, have better insight into Meta's proper capital allocation, than Zuckerberg does. I'll stick with Zuckerberg ... and add on significant dips around excess investment. A clear buy in my book.
Great read, I think the vision for Mark in the last 5 years been precieved as bearish more than bullish. If we remember during the pandemic he started diverging into metaverse 3 which crypto fanatics say will be the new internet. He then proceeded to pour tons of FCF into this idea which he then changed his mind of as he could see the stock crashing based on investors fears that Zuck has gone mad. This made investors extremely bearish and caused the stock to crash tremendously but again he maganed to dampen shareholders fears and go back to his original framework of letting Facebook be Facebook and built different aspects of the business this investment turned out to be well rewarded as investors became bullish again and the price rises to ATH. To disclose I have brought the dip under 600$ and a few days later he said he was going to slash 30billion of his original commitment into AI spending. This gives us a better understanding into his personality he does things and sees how his investors reacts and then can retract just as fast. My thesis is this there are 3 billion people using Meta and this will continue to grow regardless of China which is 1.7 billion people that means out of the world there are another 2 billion people he can still capatilise on even if he does not net the full 2 billion and gets 1 extra billion on the platform I think he can absorb and justify the capex spending on AI. These are my thoughts any feedback would be great
I agree we shouldn't put too much value on Zuckerberg's initial comments, as he tends to retract and pull back, but I have to admit I am not a fan of this strategy. I want to be able to count on management's comments and commitments, and I can't here.
Realistically, I also don't think there are another 2 billion people that can adopt Meta. Global population growth is stabilizing, no longer growing aggressively + we have to consider minors + potential social media bans on youth coming in the future, similar to Australia.
Just my quick thoughts!
Profound. Your analysis of Meta's AI investments highlights crucial finacial implications.
A few things I'd push back on
1. The $600B is mostly a headline grabber and very likely includes opex, not just capex. Their run rate opex + capex in FY25 is roughly $200B so $600B cumulative spend by 2028 is not materially different. Moreover, aside from lease commitments, they can always back down from the biggest capex bucket, chips.
2. Probably unfair to say they have no visibility to returns on this spend. As you highlighted, they saw a 10% growth in time spent on Facebook and Instagram as a direct result of their investments. Clearly, there's upside and visibility. We don't know what the split between core ad capex and gen AI capex but management can pivot as they've said.
3. I agree I'd categorize this spend as defensive. I agree defensive probably leads to worse outcomes in more cases than not but META proved to be the exception just a few years ago when they were forced to invest in AI and short form video by Tik Tok and Apple. I held a similar opinion to you a few quarters ago that they don't need a leading model if they can't directly monetize it like the other CSPs but I've come around to the idea a leading AI model may be a key differentiator to serve better ads in an AI world. They may be forced to catch up or risk losing engagement/ad dollars to Gemini/ChatGPT over time. We'll have to see but it's certainly a higher probability outcome than before.
I generally enjoy and agree with your breakdowns and takes but wanted to offer some counters here
I am very glad with your counters, Jason! These discussions are always great, leading to new insights + I can always miss some things or misinterpret facts. So, thank you for taking the time. Let me respond to your points!
1. I think you are definitely at least partially right. To quote management, it is committing to $600 billion in U.S. investments in AI, consisting of infrastructure and jobs. So, I agree there is probably an asterisk on that number, since it includes CapEx and growth in operating expenses. However, I don't think we can take run-rate OpEx here, as I have to assume it includes AI-related expenses, primarily driven by growing headcount and operating costs for these data centers.
This leaves a lot of CapEx, which is tough to support solely through cash flows. The recent bond issuance speaks for itself.
And yes, management can back down from a large part of these commitments, which is my bull scenario indeed. However, Zuck has a history of commitment to its vision. Take reality labs.
2. "We don't know what the split between core and capex and gen AI capex is, but management can pivot as they've said." I think you have a bit more confidence in management than I do, haha. Are you saying Meta might have enough compute needs in its core business to justify this level of spend and expansion? There must be an alternative market here it aims for. And I can't see it being AI chatbots/assistants.
I would love to hear their monetization plans, because I can't see them. That is my problem here.
How do you view this? In what scenario, assuming it stays committed to current numbers, do you see a good ROI? What am I missing?
I am certainly not claiming I am right here 😁
3. So better ads is your bull thesis for this spend? Does it need this amount of infrastructure? Compete this hard for talent? Wouldn't it be a much cheaper, better option to use models from an external partner and remain a cash-generating machine?
I just can't see "better ads" as the main motivation for this spend. But, again, I might be wrong.
Thanks again for the thoughtful comment. Happy to discuss further!
1. I agree it'll probably be more capex but wording suggest some opex is included in that number. I believe this $600B number first came out of that dinner with Trump when he was placed on the spot. So I wouldn't spend too much time on this number was my point. The specifics around this number are purposefully vague from Meta for a reason.
2. Meta has a history of setting large opex targets for the year and lowering it throughout the year. Happens every year like clockwork, except for this year. You're right with Reality Labs - they've mostly maintained their spending as guided.
I'm not saying they can utilize all of the compute for the core ads - almost certainly they are wasting some money on speculative gen AI stuff that won't pan out. I have been skeptical AI chatbots (in business messaging) will work here but what they're saying about it working in Thailand and Vietnam jives with my personal experience in China (on WeChat). There's a chance it'll work but Meta AI needs to be better.
3. Correct. I was pretty concerned coming into the year about this year's step up in capex because I thought the low-hanging fruit from the ramp up of Reels was mostly harvested. Instead, impression growth accelerated this year. I had some concern this might be due to Meta jacking up ad loads but Q3 results suggest this is all organic. We're still quite early in Advantage+ and automation (video generation is clearly early). The improvements to GEM, Andromeda, etc. are also notable. Again, I definitely think a lot of the planned capex will be wasted. But we're seeing the fundamentals accelerate this year on higher capex. Full disclosure, I trimmed my META after Q2 but I think the narrative is starting to shift too far bearish. I see a lot of comments suggesting they have no way to monetize this capex but actually the opposite is true - the only profitable AI capex has been ads-related (META and GOOG). Zuck is just a bad spokesperson for ads - it's clear he just wants to talk about Metaverse, Gen AI, whatever captures his imagination. Susan Lee has done a great job replacing Sheryl on the calls but doesn't have the same command yet.
I don't think they can use an external model. Zuck wants to own his own platform and there are strategic arguments why it makes sense. The key differentiator between META and SNAP is they own their infrastructure. If they license Gemini, probably more value will accrue to GOOG, which will erode their margins and differentiation over time anyway. They have to do this.
Sorry for the delayed response, Jason!
1. Fair enough! Will keep it in mind.
2. Indeed, curious about their monetization ideas, but I am still afraid they are well overbuilding their own needs.
3. I am curious about how Meta will use this computing capacity in its ad business. Maybe I am underestimating the opportunity here. Time will tell.
I agree it makes sense to own the models in-house, but I am afraid Zuck is overly focused on owning the next platform and this kind of stuff in-house, up to the point where he will overspend on it. The same goes for Reality Labs, which is built on the same idea: owning the next-gen platform. It's very defensive spending, which then makes me doubt the future of the business.
Hi Daan, I just wish to add that you have a gift of distilling a mountain of information into a short cogent write-up that communicates the key points very clearly. And a free article at that!
Thanks a lot, Daryl!! I appreciate your kind words.
I believe competing on LLMs is a bad idea for META. But AI itself is important for the core business in terms of personalized ads targeting. And if the Reality Labs business succeeds and smart glasses become mainstream, they will need AI for the user experience.
Agreed, AI is crucial to Meta's core business, but it doesn't need this kind of compute for that. Yes, if Reality Labs succeeds, these AI efforts could prove incremental and a massive opportunity.
However, these are too many "ifs" to pay this price for.
The bear case on Meta today in my view says that you, the bearish investor, have better insight into Meta's proper capital allocation, than Zuckerberg does. I'll stick with Zuckerberg ... and add on significant dips around excess investment. A clear buy in my book.
I surely do not believe that I know Meta and its finances better than Zuck himself.
Instead, I am more afraid of Zuck's AI vision and his commitment being wrong, driven by the wrong reasons.
If you are confident in Zuck and his ability to execute on his vision, then it makes sense to stay bullish!
Thanks a lot for the comment! I hope your bullishness works out!
So…if you own the stock, it’s waiting and praying they find a way to monetize AI and execute perfectly.
Selling now would suggest I believe/agree META will drop even more (you calculated 25%..).
I’ll continue praying that institutions will keep buying at the current support level. Worst time ever to not check your portfolio for 2 weeks 😵💫
Indeed, I see much more downside than upside. I don't own any Meta shares, but I would see no reason to hold onto them if I still had any...
Isnt that a bit too bearish? They will still pay a div and do buybacks.
I sure hope it is. But I am afraid it's realistic.
I also wouldn't count on many buybacks in the years ahead, and a 0.35% dividend isn't enough to make a difference.
I do see your points though.
Not sure why but I never get your articles in my ‘inbox’…
Your email? Or within the Substack app?
Maybe you can send me a PM about this so we can figure it out!