This quote in your part 2 really spoke to me - “So Google isn’t just making up the AI demand and ROIC. They’re seeing it in their core business, and their customers are telling them about it and demanding more of it. Google is telling you that the return from investing in AI isn’t imaginary or speculative.” This is the whole ballgame to me…Investors are correct to be skeptical and do their due diligence, of course. Never take what management says at face value…HOWEVER, you are right - we can’t just ignore that Google is telling us the demand is real…they are not bluffing. For me, this gives me some downside protection because it think earnings will be strong, and AI fears will eventually subside.
Why do you think markets seems to disapprove of hyperscaler (GOOGL, MSFT, AMZN, META) capex spend despite management teams saying on conference calls that they are getting high ROICs, accelerating growth profiles in their CSP arms (and Meta's FOA), and high (but declining ex GCP) margins on accelerating capital bases?
These businesses strike me as examples of Buffett's 'lots of capital at high returns' ideal, yet other investors seem scared off due to their increased use of OCF for reinvestment.
I vividly remember a time when investors were discounting big tech because they didn't have any use for the OCF other than buybacks (and were mad at some of them for their net cash balances - they wanted a dividend recap). Yet, now that they've found a new place to funnel OCF at internal ROICs (>20%) (rather than externally via buybacks / dividends at market returns), investors are angry/scared.
The whole reaction to reinvestment has been strange to me. I try to imagine that market reacting poorly to Costco reinvesting all of their OCF to open new stores, and can't. Yet, despite hyperscaler ROICs and margins being pretty well understood (approximately 20% and >30% respectively) increasing reinvestment in some of the best business models available (CSP/hyperscalers) is met with skepticism and fear.
It's a great question; I'd sugest simple skepticism. I've heard plenty of oil and gas companies tell me they're investing at great ROIs as they dilute themselves into infinity
My weakly held thesis is that certain stocks (predicated on their specific business models) build their shareholder cohorts over the years into specific types of investor groups (low P/B classic value investors vs. growth investors vs. high FCF low reinvestment investors, etc.). Think of like how Berkshire has a certain type of shareholder vs. a Constellation Software shareholder vs. a Big Tech shareholder. They each have a type/flavor to their shareholder base.
Now there has been a fundamental shift in how the future business models of the hyperscalers (GOOGL, MSFT, AMZN, META) will look. Moving from a low reinvestment, buyback heavy, FCF heavy model into a high reinvestment, buyback light, FCF light business model; and this is causing a turnover in the shareholder base as the capital light shareholders are selling shares to the high reinvestment shareholders, and once the base has turned over, the hyperscalers will have a different shareholder cohort, which is more comfortable with reinvesting in high ROIC - long term opportunities at the expense of returning capital today.
I completely agree with your read on just how significant the signal is. It is the one signal I have been thinking most deeply about and has been most erroneously conflated with AI-capex fragility. People also don't seem to appreciate the fact that ~$30bn of the $40bn ATM is to meet 2026 employee equity-comp tax.
Your read '...they’re telling you that this capital raise (and the ROIC they’re projecting from it) is in large part informed by their customers’ backlog' also supports my argument that externally-forecasted 5y-capex-ROI (chart from FT that went viral) is futile. >> https://marketstack.substack.com/p/the-impossible-maths-of-the-impossible
The return side of a 5-year ROI is built almost entirely from variables that are (a) private and demand-driven (b) hostage to the single most-contested, least-visible assumption in the whole stack, GPU/TPU useful life and the depreciation schedule, where even insiders disagree and the real inputs are unobservable which is exactly why the financing decision is the only honest signal and this is the read that hasn't gone viral enough in my opinion because it's not a reductive doom chart.
I am looking forward to your bear case because I agree it has implications for fragility more broadly (not Alphabet). But lets not forget that Berkshire more than tripled its Alphabet stake in its Q1 2026 13F, it was both the largest add and the largest absolute single-name dollar shift in the filing.
This quote in your part 2 really spoke to me - “So Google isn’t just making up the AI demand and ROIC. They’re seeing it in their core business, and their customers are telling them about it and demanding more of it. Google is telling you that the return from investing in AI isn’t imaginary or speculative.” This is the whole ballgame to me…Investors are correct to be skeptical and do their due diligence, of course. Never take what management says at face value…HOWEVER, you are right - we can’t just ignore that Google is telling us the demand is real…they are not bluffing. For me, this gives me some downside protection because it think earnings will be strong, and AI fears will eventually subside.
Why do you think markets seems to disapprove of hyperscaler (GOOGL, MSFT, AMZN, META) capex spend despite management teams saying on conference calls that they are getting high ROICs, accelerating growth profiles in their CSP arms (and Meta's FOA), and high (but declining ex GCP) margins on accelerating capital bases?
These businesses strike me as examples of Buffett's 'lots of capital at high returns' ideal, yet other investors seem scared off due to their increased use of OCF for reinvestment.
I vividly remember a time when investors were discounting big tech because they didn't have any use for the OCF other than buybacks (and were mad at some of them for their net cash balances - they wanted a dividend recap). Yet, now that they've found a new place to funnel OCF at internal ROICs (>20%) (rather than externally via buybacks / dividends at market returns), investors are angry/scared.
The whole reaction to reinvestment has been strange to me. I try to imagine that market reacting poorly to Costco reinvesting all of their OCF to open new stores, and can't. Yet, despite hyperscaler ROICs and margins being pretty well understood (approximately 20% and >30% respectively) increasing reinvestment in some of the best business models available (CSP/hyperscalers) is met with skepticism and fear.
It's a great question; I'd sugest simple skepticism. I've heard plenty of oil and gas companies tell me they're investing at great ROIs as they dilute themselves into infinity
My weakly held thesis is that certain stocks (predicated on their specific business models) build their shareholder cohorts over the years into specific types of investor groups (low P/B classic value investors vs. growth investors vs. high FCF low reinvestment investors, etc.). Think of like how Berkshire has a certain type of shareholder vs. a Constellation Software shareholder vs. a Big Tech shareholder. They each have a type/flavor to their shareholder base.
Now there has been a fundamental shift in how the future business models of the hyperscalers (GOOGL, MSFT, AMZN, META) will look. Moving from a low reinvestment, buyback heavy, FCF heavy model into a high reinvestment, buyback light, FCF light business model; and this is causing a turnover in the shareholder base as the capital light shareholders are selling shares to the high reinvestment shareholders, and once the base has turned over, the hyperscalers will have a different shareholder cohort, which is more comfortable with reinvesting in high ROIC - long term opportunities at the expense of returning capital today.
I would love to hear any pushback on this idea.
because of fear and overly-reductive frames going viral / conflating with dot.com
I completely agree with your read on just how significant the signal is. It is the one signal I have been thinking most deeply about and has been most erroneously conflated with AI-capex fragility. People also don't seem to appreciate the fact that ~$30bn of the $40bn ATM is to meet 2026 employee equity-comp tax.
Your read '...they’re telling you that this capital raise (and the ROIC they’re projecting from it) is in large part informed by their customers’ backlog' also supports my argument that externally-forecasted 5y-capex-ROI (chart from FT that went viral) is futile. >> https://marketstack.substack.com/p/the-impossible-maths-of-the-impossible
The return side of a 5-year ROI is built almost entirely from variables that are (a) private and demand-driven (b) hostage to the single most-contested, least-visible assumption in the whole stack, GPU/TPU useful life and the depreciation schedule, where even insiders disagree and the real inputs are unobservable which is exactly why the financing decision is the only honest signal and this is the read that hasn't gone viral enough in my opinion because it's not a reductive doom chart.
I am looking forward to your bear case because I agree it has implications for fragility more broadly (not Alphabet). But lets not forget that Berkshire more than tripled its Alphabet stake in its Q1 2026 13F, it was both the largest add and the largest absolute single-name dollar shift in the filing.