RE: PYPL
A bit above the $6B as it was $20.8B in 2024....by about 3.46x or ~247% that estimate; bit loose since that $6B is actually €6B.
Most of their debt is low interest which is pretty nice and a good chunk is termed out too which is also nice. I personally am not against debt use for the purposes you mentioned. If a company can use leverage and get it cheap and long duration then not a bad financial move. I just think the risk here is if they can't move the stuff it changes the valuation and the risk of $PYPL.
OT & rant: As for AI, while there are benefits no doubt, I am strongly tilting my view into there is a valuation bubble. I wrote my views a bit on the Bear! thread on SI but the more I look the more interesting things I find.
The GPU's being bought do not create more revenue once deployed so the point isn't that they "last longer" but do they produce more over that period and that answer is likely no. A GPU is not creating more profit in years 6 than it did in years 2 so extending useful life because they last longer isn't a valid adjustment. If you account for the fact that current generation GPU's will be obsolete in a few years then current gen do not have a useful life of 5-6 years as they are extended into. Meanwhile you have reported competition in China that states they are improving their models that require less GPU's than what is being used in the U.S.; $BABA claims it slashed $NVDA GPU use by 82%.
The above assumes that these centers are profitable at all too. With aging chips, new advancements, and future maintenance CAPEX, I find this to be interesting. There's situations unfolding too in a few states where the power consumption is possibly going to be pushed onto the data centers which further eat profitability. If unresolved the average consumer is being tasked with absorbing this utility cost.
Throw in the fact you have some inflation in here driven by the data center demand and backlog, that if that were to change the salvage value changes and will eventually lead to impairments charges.
Not the same situation but accounting is similar is in 1992-1997 where Waste Management adjusted the useful life of their trucks from two years to four years. Did the trucks suddenly become more valuable or did earnings need a boost? What we have today is everyone in the GPU space is extending useful life. And the issue I see is it's happening multiple times in short periods too. Like Palpatine somehow returning, somehow more value is created.
$NVDA did this in 2023 and got a net income boost of $114M and EPS of $0.05:
$ORCL did the same, increasing useful life in 2023 and again in 2025 giving them a net income boost of $573M or a EPS boost of $0.21 in 2025:
SBC the past few years is ~20% of ORCL CFO and CAPEX/Revenue use to be ~6% and in 2023 & 2022 it was ~15% with last year being 37%.
$META did this in 2021 which gave a net income boost of $516M or $0.18 EPS:
And then again with $META in January 2025 which increased net income by $1,960M or $0.76 EPS:
Railroads, dot-coms, AI, tale as old as time:
Then you have the real interesting ones like $CRWV who also extended useful life heading into their IPO:
But then is even going the route of creating a Special Purpose Vehicle with OpenAI in which the SPV will incur debt to finance obligations:
Also, how does OpenAI, a 990 tax-exempt company, drive such a backlog with Oracle?
Sparkline painted a prisoners dilemma with the optimal choice for all is to moderate AI spend and maintain their status but instead over investment is chosen:
Supported by Mark Zuckerberg who stated "If we end up misspending a couple of hundred billion dollars, I think that that is going to be very unfortunate, obviously, but what I'd say is I actually think the risk is higher on the other side.". He also later stated in the very worst case Meta ends with excess capacity that will result in loss and depreciation.
In my opinion earnings today are being driven up by accounting adjustments, revenue trading between a handful of companies, creation of SPV & joint ventures to continue the growth off the books, and in the end no one can tell us how close are they to really solving AGI? At what point does this growth CAPEX turn into maintenance CAPEX and really start eating in cash flows? Does one really assume we won't see asset impairments in coming years?
These adjustments changes reinvestment rates which eventually impact cash flow projections and then valuations. Given market concentration into MAG7 and AI if these valuations change what does that do? Do money flows reverse course?
<<Of course with any technological innovation there will be lots of entrants and few survivors but that does not make it a bubble>>
I think the above makes it a bubble the more I pay attention.
Happy investing,
-Sean |
