Inflation, energy, CO2 taxes and the impact on Artificial Intelligence nasshliski

It continues to surprise me that central banks, calculating in the dynamic general equilibrium (DGE) models they use and complementing them with subsequent more micro studies of supply and demand analysis, did not foresee the speed of the reduction in inflation that is taking place . .

I am of the opinion that the central banks cannot be blamed for not even anticipating the rise, given the severity and duration of the war in Ukraine, and above all, the brutal increase in gas and electricity prices, as consequence of the cutoff of the supply of Russian gas to Germany due to the blowing up of the Nord Stream gas pipeline, it was not something foreseeable. What does seem worrying to me is that they did not foresee the speed of the fall, so linked to the effects of energy prices, which are definitely also key to underlying inflation.

This failure is, in my opinion, the consequence of a larger problem: the lack of attention of academic economists in their models to the energy variable, which is more important than we think. It is not well understood why we continue not paying attention to energy in our economic models, we do not even take it into account, we only talk about it when an inflation problem arises. And this is already a mistake, because it leads to a lack of understanding of how the interaction of energy works on all fronts, including productivity. If we did, we would understand how wrong some policies are, because they omit the relevance of this variable in the analysis of economic policy decisions. Also very dangerous are academic studies that, calculating on unviable assumptions, send messages such as that with a global tax on a ton of CO2 of 227 euros we will reach the famous net zero emissions in 2050.

Let’s go by parts. A clear case of error in assessing the relevance of this variable is observed in the European energy and CO2 pricing policies, energy-climate policies that are leading to losses of competitiveness and income in Europe, which will not cease to exist. increase in the future as you will see in the following paragraph in reference to Artificial Intelligence. Basically, this is due to political decisions that imply much higher energy prices than those of our partners.

These policies are very negative for our future because there are very relevant issues that are not discussed or analyzed by our authorities and that are key. For example, has anyone analyzed and contrasted whether productivity increases occur more in periods with low energy prices? Or that, probably, long-term productivity in Europe depends on the level of these energy costs and there may be a productivity gap compared to other countries for this reason.

Why are climate policies expelling sectors with high productivity and high energy consumption? And not only to current sectors such as aluminum or motor vehicles, but also to the economic sectors of the future. Let’s think about data centers, which are large consumers of energy; or better yet in the development of artificial intelligence. Does it make sense to invest in this sector in Europe, at these energy prices?

To understand the magnitude of Europe’s problem in this area, let’s look at the calculations of extra costs in Europe compared to the US for new chips specially designed for Artificial Intelligence (AI). The Nvidia H100 AI GPU card or chip includes technological innovations that combined can increase the acceleration of large language models by 30, maximizing what can be done in conversational Artificial Intelligence (AI).

It measures only 27 cm long, 11 cm wide and 2 cm deep. Nvidia is expected to sell around 2 million in 2024 in the United States alone. This is in addition to the 1.5 million that were sold in 2023. Well, each of these chips consumes up to 700w of power (maximum power). At 61% annual utilization, a single H100 GPU card would consume approximately 3,740 kWh of electricity per year. That is, a single chip approaches the average energy consumption of an electric vehicle. Be careful, more cards will be sold than electric vehicles in 2024 in the United States

The energy intensity of AI is enormous. What investments are going to be made in Europe in AI if electricity, just due to the effect of the price of CO2, (90 euros per ton in 2023) is already more expensive in many European countries than in the United States? But this is not the worst, the worst is that the marginal price of electricity, which is generated with natural gas, is today three times higher here than in the US.

In summary, we are talking about the fact that the variable cost of Artificial Intelligence (that is, the cost of electricity per chip or card) is four times more expensive here than in the United States! And we are not aware that this overpricing is largely due to current European energy and climate policies. Has anyone explained this to Mario Draghi, who is carrying out a study on the future of Europe? Don’t worry, if he is not aware, I will tell him ‘whatever is necessary’.

In addition to the wrong policies, we have wrong analyzes from social prescribers or experts who do not talk to us about this problem of CO2 costs and energy policies in Europe. Rather, they send the opposite message, which is good. Thus, a recent macroeconomic study that concludes with messages such as “with a CO2 tax of 227 euros per ton we reach the goal of zero emissions in 2050 throughout the world”, and everything is fixed, since it also does not have a high cost in terms of social welfare.

To begin with, the assumptions of the models must take into account that the evolution of CO2 reductions in Spain becomes possible in part because you transfer them to other places, increasing them more than those that are already generated, because they are less efficient and they use fossil fuels with more carbon content; Therefore, the net of well-measured CO2 reduction is much smaller than is thought and so is the technological improvement. The problem with expert opinion is that society receives a message of “yes we can”, which is absolutely wrong. To say that by putting the tax on everyone you achieve something is not understanding what we are talking about. To show it, I leave another example. Graphite is key in the energy transition, as it is irreplaceable in batteries. One ton of spherical graphite requires the consumption of 60 tons of high-quality carbon that “only generates” 180 tons of CO2. If the optimal tax is 227 euros per ton, then the price of graphite before starting to produce it is 50,000 euros per ton. At that price of CO2, what would happen is that the transition would be delayed, since it means multiplying the price of the smallest electric vehicle by three. If you do the same calculation with polysilicon, the Nvidia chip would multiply its price by four, which would also delay the introduction of AI. I don’t know if that’s what we’re looking for, that’s the same thing.

Leave a Comment