저자: Kevin · 외국인
요약 보기
연준 리사 쿡의 AI와 경제 관련 논문을 근거로, AI 관련 물가 상승은 공급 측 문제로 시간이 지나면 해소될 수 있어 추가 금리 인상은 실수가 될 수 있다고 주장. 메모리 가격 상승률이 2027년 하반기 정점 후 둔화되는 곡선을 제시. 이란 협상 타결 시 유가 하락으로 인상 필요성이 낮아진다는 시나리오를 제시하며 추가 금리 인상 없을 확률을 30~40%로 봄. 스타트업과 AI 활용 증가가 디플레이션 요인이라고 언급.
다듬은 글
Well, the Federal Reserve just published another paper on the risk that artificial intelligence poses to the economy. And that actually gives us a little idea of what the Federal Reserve's preferences might be for this rate hike cycle. Because now we already know that the markets are pricing in about five rate hikes over this cycle, which, you know, is starting to sound eerily similar to what we saw in the mid-90s, where we probably don't want to go in that direction.
We went from 3% rates to 6%, which would mean we would go from 3.5% to about 7% Fed rates. And, you know, the discount rate. So , the Federal Reserve, through Lisa Cook, just released this paper, an update on AI and the economy. And in it they warn about a potential error. Now, I find this very interesting.
So, I've already had the honor of just pointing out the good things. Okay, so let's get straight to the good stuff. So, AI is poised to become a significant part of our lives. Well, okay. Okay, let's begin. Over the past year, one of the factors driving inflation has been rising prices for AI-related goods, such as chips, computers, and software.
We literally just talked about this. If we move on to the prices of GPUs like the 5090, that is, consumer GPUs, then we are already clicking around. were like, " Okay, for $4,500 they're sold out." For $4,700, they are sold out. For $4,400, they are sold out. So how much should you pay for it? Want to buy one right now on Amazon?
It costs 7 thousand. $7,000 for a stupid 32GB 5090 that could run an open-source model. This is madness. So yes, there is real inflation there. And obviously, that puts some pressure on the Federal Reserve. So , do you know where the mistake is? This is a beautiful part. It will happen in a moment. Uh, I think some of these price spikes reflect a shift in demand toward AI-related sectors, rather than an increase in overall demand in the economy.
When a surge in demand is concentrated in one sector, the goods and services, which by the way are also called gas, in that sector can be pushed to the steep part of the supply curve. In English, when you try to push a lot through a small hole, the pressure increases. If this demand were evenly distributed throughout the economy, rather than concentrated, the overall inflation rate might not actually rise so much.
As supply chains adapt to this imbalance, then efficiency gains can begin to accrue , and then such pricing pressures should disappear without policy intervention. Wait, wait, wait, wait, wait. This is really important because it argues that if the inflation we're seeing now is related to artificial intelligence, then it should disappear on its own as you start getting more power in building artificial intelligence.
Where is an example of this? Well, an example of this would be high-speed memory or, to be honest , just memory. I mean, we also have Micron, so a lot of people were worried about, oh no, Micron's earnings. But I think what I'm interested in is what the curves are already showing for memory price projections.
For example, something that was quite popular and is now circulating is the price per gigabyte of high-speed memory. And you can see it peaking. Uh, and this view is located about here. in the second half of 2027 . And so you start to see these prices go to the point where you actually get a negative second derivative.
Uh, well, I mean, yeah, the second derivative is... I mean, the first and second go over... Wait a second. Uh, the growth rate is first, and then the rate of change of that growth rate goes ... That's right. Right. Good. In other words, it's a slowing of the first derivative here, where we go from 15% price growth to 11%6%6% eh, and then we go negative.
Okay, I'm just a little confused. The bottom line is that price growth is slowing and potentially becoming negative. This is consistent with Lisa Cook's argument. What's interesting is that her argument is that maybe we don't need to raise inflation to that level, which is actually the opposite of what we've heard from the Federal Reserve.
We have heard a lot from the Federal Reserve that we need to ensure that inflation expectations remain anchored. Basically, we need to put on our pants. We need to make sure the economy knows we are going to do something. In many ways, this is probably due, frankly, to the geopolitical issues of the war with Iran.
Lisa even goes so far as to say the following on behalf of the Fed, or at least from her perspective. Trying to combat inflation in individual sectors through monetary policy may be a mistake. Our tools are too straightforward to narrowly target these sectors and decide relative prices. Solving the problem of changes in relative prices is not our role.
In other words, their job is to focus on inflation across the economy. Not sitting around and trying to say, “Hey, people are trying to squeeze a lot of demand out of this one sector, or there’s a lot of demand that’s squeezing this one sector. Therefore, the supply is contracting. Therefore, prices are being squeezed.
Oh, we're going to respond to the AI boom by raising rates." She says, "No, don't do that." Because if you do, you might make a mistake. The word " mistake" is very useful because everyone remembers Allen Greenspan and says that he raised rates too much, and his political mistake led to the bursting of the dot-com bubble when he last raised rates in May 2000 .
And then you had your little " oohs." She then goes on to say that there could be some pressure on the economy from the demand for AI fuel, mainly, I mean, if you have workers making more money and that expansion, uh, through construction, energy, related labor, water prices, electricity prices, etc., you could have broader inflation that you have to watch carefully and be vigilant about.
She says a productivity boom could counteract such growing pressures. However, if demand does not expand to meet the additional supply , you can also eventually expect that as supply increases, for example in the memory market, you should expect prices to fall on their own. In other words, it states that you expect prices to fall as supply catches up in the future.
This is exactly what is happening with memory prices. I currently expect productivity growth to provide a moderate reduction in inflation over the next few years. Okay, great. However, I do not expect these effects to emerge in time to offset broader inflationary pressures later this year. Interesting.
So you have this balancing act from Lisa Cook where she essentially says, " Look, everyone is buying so much AI right now that we're seeing inflation everywhere." So we got a raise in the near term , but after that we expect an increase in supply and an increase in productivity that says, "Hey, maybe after this one rate hike that we did, maybe another one, we could just watch." And that's what we're watching because these pricing issues can change on their own , which is good .
So I expect relief to be limited when it comes to the productivity boom that provides inflation relief. In other words, we are really counting on the offer becoming active. Okay . She then goes on to say that scenarios where we get more inflation relief mean that demand will be negatively affected , and that is significantly less than the economy's capacity.
This can happen. So, for example, how can you get strong deflation? There are several different methods here. First, the appearance of a proposal is a kind of base case. But then some more problems may arise. One of them is the high concentration of productivity growth among high- income earners, who tend to spend less.
Productivity gains are not passed on to employees. So you don't have that kind of wage increase, or you have some form of painful job reallocation that leads to higher unemployment, which increases household savings. Okay, that's a lot to take in, but I think it's actually worth breaking it down because it really does give you some warnings.
First, she says that if you raise prices solely because of supply-side problems lik grows. Okay, that would be a mistake. However, you need to be careful with this. Okay , then it will be vigilance, and it will be these three things. This will be the number one productivity concentration. Concentration of product performance.
That's it. Okay, that's number one. So it's a concentration among wealthier people who aren't really going to spend that extra money. Secondly, wages are not falling. Wages aren't rising, are they? They don't get a share of that productivity. Uh, and then, number three, it's job redistribution , uh, job redistribution .
So those are your three big risks that could actually lead to significant deflation. So, we'll make a little note here. All these forces are here, deflation. Each of them indicates deflation. Artificial intelligence supply points to deflation. Concentration of productivity among the rich, deflation.
Wages are not growing, deflation. Redistribution of jobs is not good because people need time to retrain, right? Deflation. These are all deflationary forces. Note that it doesn't specifically talk about geopolitics because it is inflationary. Good. So let's continue here. Okay . Good. So, if market concentration leads to less competition and higher markups, you can get inflation.
In other words, if you don't have enough competition in the chip market, but I think we have a lot of competition in the chip market. So I think that's one less argument, uh, less compelling, if you will. I think there is a lot of market competition for A6, GPUs and other products. Ultimately, the timing of any disinflationary payoff will depend on how quickly and widely businesses adopt AI tools .
We see widespread concern about progress in times of any new technological innovation. So she's basically saying that whenever you have something like a printing press or a loom or something, whenever these revolutions happen, it's perfectly normal for people to be excited about new technologies. What will it give?
Will my children have jobs? What will the future look like ? This creates a lot of uncertainty. So she acknowledges it to some extent. She also says that at the moment, while AI is delivering significant results, there is limited evidence that AI is delivering significant changes in the job market. The data underscores this.
Unemployment and layoffs remain low , which is good. However, some workers have really been affected, and I understand the hardship that unemployment creates for families. Ultimately, artificial intelligence could reduce the demand for labor. Recent college graduates are facing increasing difficulty finding their first job, and there is evidence that in some sectors, including coding and software, labor demand may be declining.
These types of shifts may expand as AI adoption deepens and AI technologies improve . In other words, the more we implement artificial intelligence, the more pressure there is likely to be on the job displacement aspect . Therefore, she says , I am very closely following the scenario where artificial intelligence leads to at least a temporary increase in unemployment.
In this scenario, the rise in unemployment among employed people may not necessarily reflect a shortage of demand, but rather a mismatch between workers' skills and available jobs. This is your, uh, retraining problem that I was talking about. In this case, we at the Fed will have limited tools. In other words, good luck, bro.
We could lower the Fed's interest rate in an attempt to lower unemployment, but that could run the risk of fueling inflation . Good. So, we have the spread of artificial intelligence between firms. I see evidence that smaller firms are showing resourcefulness and adaptability when thinking about how to leverage their advantage in AI, even if they lag behind larger firms in their implementation .
Now, essentially, this is where she starts saying, "Hey, there are a lot of companies that are using AI." The good news is that this doesn't just apply to the big ones. We know that big companies are using AI and spending money on it like crazy. But there is also good news. Small early-stage investors and founders who have AI and use AI tools have made it easier to start a business, which may be a factor in the historic boom we’ve seen in new business formation recently .
Okay, that's interesting. Let's try to discuss this here, because this is actually a very important point that I think applies to startups . First, she argues that there are reasons why AI could create significant deflation. We have n't even considered the real productivity gains that could result from inflation, but some of them are listed in her three main questions.
All of this suggests that we can't do anything about AIEL inflation, and we shouldn't. If the geopolitical problem goes away, then we need to worry more about these deflationary forces than inflationary ones, which is generally good for the economy as a whole if the geopolitical problems are resolved.
So, if the Iran deal is resolved, we should focus on deflation, which is crazy to talk about right now when yields are skyrocketing, right? But that's exactly why I find it a pretty interesting little bet. This is on Kalsha, their channel sponsor. But in any case, I find it very interesting. According to Cauchy, there is only a 10% chance that there will be only one rate hike this year.
Right now, I think the volume is a little bit understated, so it's less than ideal, but I find it very interesting because I would say there 's probably closer to a 30 or 40% chance that there won't be any more rate hikes. Just listen to me for a minute about this scenario, okay ? So how can you potentially get no more rate increases?
It seems incredible now to assume that we won't have any more rate hikes. And by the way, getting $25 free when you sign up with Cali and make your first trade. Okay, sponsored affiliate link, paid advertising. Okay, we understand. So, if we get a deal with Iran in the next, say, six, five weeks, right before the election, right?
So, we'll get a deal with Iran between now and then, and then all of a sudden the price of oil jumps from $105 to $80 or $70, right? Saudi East- West pipelines have already reopened. Maybe we'll even get a temporary ceasefire, uh, for grain and fuel exports in Ukraine and Russia. It's a little more like dreaming about reality.
But then the Federal Reserve might be tempted to say, "Hey, why don't we wait for more data?" Then they stay in October. Then you get data in November and December that shows , wow, oil prices are down. Overall inflation is decreasing. Some of this will start to be reflected in services inflation. And therefore, inflation in the services sector will to some extent reduce the potential need for a rate hike in December.
So I personally think that the 10% probability of just one rate hike, which we've already had, is actually low. It should probably be closer to 30% , assuming we make a deal, right? This is my opinion. You know, this isn't personalized betting advice, predictions, or investment advice, whatever. I just have that thought.
I think there is a way after the Iran deal. You already know that I'm very vocal about this deal with Iran because I think it's going to happen. Uh, there is a real way to ensure that we don't get any rate increases. Given the latest Purchasing Managers' Index (PMI) we received, it does n't look good, but is it 90%?
No, it's not that bad. Uh, but anyway, it's about that and the formation of new businesses, because if we don't get more rate hikes, you're going to see more and more of these smaller startups taking on funding and business development during the AI boom , which then could continue to run for at least a while longer.
I don't think we're at the end of the boom yet. This is good because artificial intelligence actually makes humans smarter for competition. I think that's great. I mean , I look at what we're doing with our startup, and I mean , we were a small real estate company in 2022 . It's kind of like how Elon, he was like, "Oh, it's all about Starlink, and we're a trillion dollar company or a $300 billion company or something," and then they add artificial intelligence, and then they're like, "Well, you know , what's a one and a half to two trillion dollar company, right?" So this is kind of how you can grow.
Us too. And so when I read this about startups, I'm really excited because I look at it and think, "Okay, the last two times we filed financial statements, we were already operating at an operating profit," which is great , because that means that investor dollars are turning into even more dollars, right?
But I think a lot of people will do it, and I actually think you could make money buying your own GPUs, buying your own CPUs , little mini PCs or something like that. And you could make money by launching your own Chinese models and providing services built on that. You don't have to pay Anthropic or Claude, or, well, you don't have to pay Claude or GPT for all of this.
This is a win, and it's deflationary. Now I will say that Chinese models are not ready for everything yet. For example , will I be developing advanced 3D animation or any other design I'm working on with Claude Fable, or will I be using open weight? Oh, you'll spend money on Fable, because it's still a lot cheaper than, you know, a designer who'll spend 6 months and figure it all out .
Without any discrediting of the designers. I'm just saying you could be busy with something else. But in any case, Lisa Cook is essentially arguing that aside from the geopolitical issue, there are many signs that both large and small companies are using AI to save money. Evidence shows that AI adoption is happening faster than the PC or Internet adoption cycle.
This is a great achievement compared to other historical moments. She also says it's encouraging for small businesses because we know that some of the most important breakthroughs in this country that we've ever seen have come from startups that are using cutting-edge technology in new and innovative ways.
This makes me happy. AI can help us generate more ideas faster, which suggests that we are in a ROR moment. So what is the ROR moment? The ROR moment is when people's ideas lead to innovation. Generating ideas is perhaps the most important thing for a small business. Coming up with really good, competitive , valuable ideas that can then be put into practice.
Like the people I see with many people, I can realize so much with AI. That's good. But it's a combination of things. You need to come up with an idea that 10 million people haven't already thought of. Then you need a team to implement that idea. So you need people with programming skills to create something that is really functional and user-friendly for a large number of people to use.
Uh, and you know, then you need to spread it around, and we'll get money to market things, which are a lot out there. So ideas are a really important part of this. Uh, and she's basically saying that if you could solve this and spread the ideas more widely, hopefully other things would come out. We could become more productive.
You could develop the economy. This is kind of a key point - the idea, this idea. Do you understand? But in any case, AI can allow small companies to scale faster and ultimately create more jobs. This is not a guarantee, but it is a reason for hope. That's exactly what she's doing here. It's all like the second half of the article here is, in fact , entirely devoted to hope and how resilient the job market is, and so far there 's reason for hope, blah blah blah blah.
But the main point of this is, let's put it all together. The main point of her opinion is that we cannot lower oil prices . But beyond the drop in oil prices, we must be very careful not to make mistakes. Because if we are growing due to AIL inflation and the supply problem solves it on its own, then you are making a mistake.
And you have three main risks: that the gains from AI will be concentrated among the rich, who then don't spend their money, which means you get deflation because it's not reflected in wages; and therefore wages are not rising, which is also deflationary or at least disinflationary, right? So, disinflation, right ?
Prices are falling, or, or, or prices are rising at a slower rate , and deflation is simply a decrease in prices, but all of this reduces inflationary pressure. And then of course, job reallocation , where people who are not very suitable for an AI startup , say, have to be retrained, and they lose their job, maybe, say, in customer service, and now you have to retrain to work at an AI startup; it will be completely different .
It's going to take some time to re -train on that, so I think this is a pretty clear signal from the Federal Reserve that, aside from the geopolitical challenges we're facing right now, most of what's ahead is deflationary, not inflationary. So I want to go back to this part about memory chips again, because I'm going to try to explain it properly now, because I came across it at the beginning of the segment, and it bothers me.
So, okay, first derivative. I know how it relates to derivatives . Uh, I'll just call it the first d. The first d is the price change, right? So, this is, uh, this is a change. Uh, so it would be, you know, whatever we change from $18 to $20.20, so $20.70, for the price of memory. How much is that divided by 18?
That's about a 15% change, right? That's it. This is a 15% change. Good. But if the second derivative is the change of the change... So if suddenly it changes, say, from 15 to 30%, then to 60%. The second derivative in this case would be doubling, right? Uh, it's going to be a big... big change, but it's going to be a constant and a doubling.
If suddenly it drops to a negative value , meaning this right here becomes negative, then suddenly it drops, this line falls off the cliff, because the growth rate has just completely changed. You have gone from positive to negative at a rapid pace. So a little clarification on this matter. Uh, but that's what you see here.
And this, uh, who did this? It was, let's see, Jeff. Jeff sees this, and that's what we 'll be looking for in Micron's earnings report tomorrow, and also what we're seeing is not just price growth year over year, looking back, which should look great. I mean, we should still see a price increase of somewhere around 15 percent for high-speed memory , but, uh, what kind of forecast are we seeing?
This will be a key indicator for Micron. But it all settles down as comments on the proposal online, which is exactly what Lisa Cook says. So, I think that's the exciting part, when you put all of this together, the Fed. This is what they are talking about. And for me, the main takeaway is that the damn deal with Iran can't happen fast enough. .
Comments 0