Job creation slows, dementia-addled chaos-monkey tariff risks continue, data-center construction booms with little ability to rapidly move resources into that part of the construction-investment...
If something cannot continue, then it will stop, and I wouldn't bet on the AI bubble stopping with a soft landing just because the underlying investors are well-heeled.
AI spending looks like the only growth sector in the economy. Real PCE is flat. The government sector is flat. Residential investment is falling. Non-residential structure investment is falling. It's asking a lot for AI spending to pull the rest of the economy. Even if it succeeds in statistical aggregate, for many industries, professions, and geographies, it will be a recession.
It could be a sectoral positive demand shock that calls for a dollop of over-target inflation. Of course, that is what we have and what TIPS expects, but modestly. Is that enough?
The Dallas Fed last July predicted that without net immigration, our population is set to start shrinking by 2040. 4 years of the Trump Administration will surely accelerate that population trend. The loss of new laborers/consumers will be a huge drag on GDP, and with AI construction sucking up the available construction labor by outbidding others, and because construction is an immigrant intensive industry, construction of much of everything else will be slowed down. With 2/3 of our economy being consumption, the rapid drop off in immigration this year has already almost stalled consumption growth. A tighter labor market, zero or negative net immigration, and TRUMP'S TERRIBLE TARIFF TANTRUM will give us stagflation and years of economic stagnation.
A large percentage of those job losses were in the federal government (or in state/local positions funded by federal transfer payments) so this, ironically, should be touted by Trump as a "WIN!!!"
I cannot understand what a Fed governor can mean by saying that inflation risks are (or are not) limited, when it is he and his colleagues on the FOMC that will decide on how to set policy instruments to get more or less inflation. Is it
a) I judge that an x bp change in the EFFR rate will have less effect on inflation in today's conditions that a some (implicitly understood) past time?
b) I judge that an x bp change in the EFFR rate will have less effect on inflation than my colleagues do?
And no, I do not think it is the Fed's "profesional" role to be more concerned with over-target inflation
The LLM biz, if you can call it that, is currently loss-making, with faith that it will be profitable with greater volume, somehow (divine intervention?). The big AI companies are spending vast sums of money to continue hyper-scaling, and are propped up by very high equity valuations.
The risks are high. A better way to get quality output from small, cheap models, undercuts the hyper-scaling approach. This might make the new "computronium" facilities obsolete, in turn dragging down valuations and with it the stock market. Then the recession? Add in the possibility to please 47, these facilities will be powered by fossil fuels rather than renewables, perhaps with the gas that the EU will not purchase. (Musk is already doing this for Grok.) On the other side of a meltdown, maybe we get to buy high-end GPUs for a song, like Aeron chairs after the DotCom bust.
Right now, LLM-based AI is relatively poorly functional, unable to deliver on the hucksters' promises, and AGI is most definitely not "around the corner". Which means there will not be vast armies of deployable humanoid robots to do all the jobs that humans do. That is good news, as few people have to act like Luddites once the hype deflates.
It should be clear that if you make inputs more expensive, both supply will be constrained, and even demand might fall from higher prices due to increased production costs, as well as trade partners exporting less to the USA and diverting supply to other nations with more favorable trade terms. If trade doesn't respond in the way 47 wants, what can he do? Keep doubling down on tariffs?
47 is also going to find himself in a very uncooperative world, trying to decouple itself from the US as painlessly as possible.
A very informative post, of which I have only contradictory thoughts. Mr Waller might be all that you said, but if the chaos monkey favors him, then his loyalty to anything but is subject to disbelief.
Without any direct knowledge, I believe there is more to the build-out of data centers than you have mentioned. If some in the AI community are correct, this investment could be similar to the breakthroughs in nuclear energy and our understanding of subatomic particles. I view these investments as defensive in nature, as the entire field of computing will be changed.
This has just occurred - the chaos monkey is firing Dr. Erika MaEntarfer as the commissioner of the Bureau of Labor Statistics. He is accusing her of manipulating the numbers to make him look bad. This only goes to prove my point that if you are not doing what he wants, you are gone. As he has said, he doesn't need experts, since he knows more than the experts. When the economy contracts, he will blame Biden or maybe Obama, but he will be blameless.
If something cannot continue, then it will stop, and I wouldn't bet on the AI bubble stopping with a soft landing just because the underlying investors are well-heeled.
AI spending looks like the only growth sector in the economy. Real PCE is flat. The government sector is flat. Residential investment is falling. Non-residential structure investment is falling. It's asking a lot for AI spending to pull the rest of the economy. Even if it succeeds in statistical aggregate, for many industries, professions, and geographies, it will be a recession.
It could be a sectoral positive demand shock that calls for a dollop of over-target inflation. Of course, that is what we have and what TIPS expects, but modestly. Is that enough?
The Dallas Fed last July predicted that without net immigration, our population is set to start shrinking by 2040. 4 years of the Trump Administration will surely accelerate that population trend. The loss of new laborers/consumers will be a huge drag on GDP, and with AI construction sucking up the available construction labor by outbidding others, and because construction is an immigrant intensive industry, construction of much of everything else will be slowed down. With 2/3 of our economy being consumption, the rapid drop off in immigration this year has already almost stalled consumption growth. A tighter labor market, zero or negative net immigration, and TRUMP'S TERRIBLE TARIFF TANTRUM will give us stagflation and years of economic stagnation.
https://www.dallasfed.org/research/economics/2024/0702
A large percentage of those job losses were in the federal government (or in state/local positions funded by federal transfer payments) so this, ironically, should be touted by Trump as a "WIN!!!"
I cannot understand what a Fed governor can mean by saying that inflation risks are (or are not) limited, when it is he and his colleagues on the FOMC that will decide on how to set policy instruments to get more or less inflation. Is it
a) I judge that an x bp change in the EFFR rate will have less effect on inflation in today's conditions that a some (implicitly understood) past time?
b) I judge that an x bp change in the EFFR rate will have less effect on inflation than my colleagues do?
And no, I do not think it is the Fed's "profesional" role to be more concerned with over-target inflation
a) than politicians/public or
b) than with under-target inflation.
If anything, it should be less concerned.
The LLM biz, if you can call it that, is currently loss-making, with faith that it will be profitable with greater volume, somehow (divine intervention?). The big AI companies are spending vast sums of money to continue hyper-scaling, and are propped up by very high equity valuations.
The risks are high. A better way to get quality output from small, cheap models, undercuts the hyper-scaling approach. This might make the new "computronium" facilities obsolete, in turn dragging down valuations and with it the stock market. Then the recession? Add in the possibility to please 47, these facilities will be powered by fossil fuels rather than renewables, perhaps with the gas that the EU will not purchase. (Musk is already doing this for Grok.) On the other side of a meltdown, maybe we get to buy high-end GPUs for a song, like Aeron chairs after the DotCom bust.
Right now, LLM-based AI is relatively poorly functional, unable to deliver on the hucksters' promises, and AGI is most definitely not "around the corner". Which means there will not be vast armies of deployable humanoid robots to do all the jobs that humans do. That is good news, as few people have to act like Luddites once the hype deflates.
It should be clear that if you make inputs more expensive, both supply will be constrained, and even demand might fall from higher prices due to increased production costs, as well as trade partners exporting less to the USA and diverting supply to other nations with more favorable trade terms. If trade doesn't respond in the way 47 wants, what can he do? Keep doubling down on tariffs?
47 is also going to find himself in a very uncooperative world, trying to decouple itself from the US as painlessly as possible.
I should have said only two contradictory comments.
A very informative post, of which I have only contradictory thoughts. Mr Waller might be all that you said, but if the chaos monkey favors him, then his loyalty to anything but is subject to disbelief.
Without any direct knowledge, I believe there is more to the build-out of data centers than you have mentioned. If some in the AI community are correct, this investment could be similar to the breakthroughs in nuclear energy and our understanding of subatomic particles. I view these investments as defensive in nature, as the entire field of computing will be changed.
This has just occurred - the chaos monkey is firing Dr. Erika MaEntarfer as the commissioner of the Bureau of Labor Statistics. He is accusing her of manipulating the numbers to make him look bad. This only goes to prove my point that if you are not doing what he wants, you are gone. As he has said, he doesn't need experts, since he knows more than the experts. When the economy contracts, he will blame Biden or maybe Obama, but he will be blameless.
Well, that was an unsettling read.