This thread by Guido Appenzeller from a16z's AI team is a strong rebuttal of the Sequoia article.[1] I personally found Sequoia's argument a bit strange, since it not only misunderstands the unit economics of the industry at a basic level, but further misunderstands the ways that AI will be "just another component" of software stacks — not unlike databases, high-level programming languages, network infrastructure, etc.
Seriously, author just throws out “for every $1 spent on a GPU, roughly $1 needs to be spent on energy costs to run the GPU in a data center” without bothering to share even a semblance of assumptions. Then we get a casual “let’s assume they need to earn a 50% margin.” What? Why?! Also, at these scales and timeframes, the time value of money becomes significant.
This is not margin question. This is question, of extreme energy ineffectiveness of current AI, so even most energy-effective GPUs spent their cost to energy in less than year (if used 100% of time).
FPGAs and ASICs are better on energy effectiveness, for 3-4 times, but cost per installed FLOP also larger for 3-4 times.
And money are expensive now, so we cannot consider, $200B investments will be returned with low rates for example 20 years, now much more real scenario, interests about 20%, so need to have much higher margin, to pay debt within 4-5 years maximum.
If only there was a non-X-thread version of this that was readable (as the threads are now hidden for people not logged in).
Of course, the author is free to choose to publish his views in any way they see fit. Except this way it gets a "meh" from me at most (and even that only because of the intro in this post here), since it's functionally hidden, as good as if it didn't exist.
> What are you going to use all this infrastructure to do? How is it going to change people’s lives?
there are some questionable linkages in that article, in the rush to discuss "the big numbers"
specifically, core AI "foundation" models are (supposedly) built once, while tuning and interaction follow, with less energy for each activity. If the infrastructure arms-race settles down (not there yet) then what does this technology do .. in markets? to society? for investors?
It is difficult to project actual social concern onto SequoiaCap here.. if they had it for a moment I am sure it would pass quickly.. But sure, what is the effect of the use of this money?
When people realize that their jobs are being replaced, that suddenly ads for lotteries and junk food are a lot more plentiful, that ads are auto-generated and placed in every conceivable device while human needs are visibly shortchanged (housing). Yes, actually, there might be consequences to all that.. is that the implied question, too?
"The important question to be asking is: How much of this CapEx build out is linked to true end-customer demand, and how much of it is being built in anticipation of future end-customer demand? This is the $200B question."
This is an amusingly incorrect question. It presupposes AI as if it was a technology like any other, which it isn't. It's deeply transformative. Money actually doesn't matter here.
AI is not for making money. Money is for making AI.
Unfortunately we are not in state of constant economy grow, now economy in recession and have great deficit, so money are expensive, so investments must be returned within short time interval, so need extremely high margins.
Question is, if people invent lot of things, totally worth about $1000B in just ONE YEAR, and now clearly seen less then 1/10 of this number.
Examples could be fully autonomous car autopilot, robot capable to replace 90% of building industry workers. They are possible, but not seen in ONE YEAR time frame.
ML is a technology very much like computers, the internet, smartphones and electricity. All highly transformative across all industries and society as a whole.
Author said, people enter BIG race, to buy huge number of GPUs, but at the moment not seen, from where their cost will be paid.
I mean, if $200B should been invested, this must consider at least 4 x $200B of returns (including salaries, fees, taxes, etc), so at least $1000B must be somewhere on horizon.
If industry will not found source of $1000B, will be just next bubble.
Yes, it is possible, to happen some wonder, for example, government could consider this very important, for example for defense, raise taxes and pay these money from budget.
But any possible money must appear nearest months.
From this list, only Global Wireless Telecommunications could easy withstand to spend 1000B, all others will have big troubles if spend so much, as this is very large share of their revenue.
- Only government could just rise taxes (or just print money) to gather additional 100B, just because they need money. But very few businesses could do such things in so short term and without very serious foundations.
This struck me as if it was written by an investment banker, so I looked up the author, and confirmed he was indeed an investment banker prior to joining Sequoia.
[1]: https://x.com/appenz/status/1704915400096649696