It is not scaremongering, far from it, and although there will be sectors that will lower their prices because they are involved in the purest speculation, the rest of the industry is having severe difficulties to maintain current prices, where some have been blown up like that of the motherboards. Everything is influenced by the foundry industries, where costs do not stop rising and the question already arises: is this crisis of the chips the regulation of prices of the components for the future?
It is assumed that as the nanometers are reduced a chip should be more efficient in terms of performance / watt as well as its cost, because more chips are manufactured per 300mm wafer. But there is a variant that we are not taking into account and that no longer even belongs to the R&D of companies …
The cost of FABs multiplies
Staying on top of the wave is going to be really tough for Intel, Samsung and TSMC. We are currently in an all-out war where the foundations of the future are already being laid and the semiconductor crisis has only done for these FABs is overwhelm them and have to accelerate the planning and development plans of the foundries. This implies settling in strategic places against their rivals in favor of their interests, since logically no one wants to be left behind and that is why countries are struggling to obtain the services of private companies such as the three mentioned.
Yes, it is the countries that are going to give tax incentives and subsidies to companies that are not even from their own country, all for having the territorial hegemony of the technology of the future, a cold war that is shooting costs to levels never seen before.
After this brief introduction of the world panorama we go with the data, because it is terrifying … From 90 nm, manufacturers are working with 300 mm wafers as we saw in a specific article for it, and although this in terms of investment did not represent a Very large growth percentage, if we look back and thanks to the data provided we see that they have not stopped increasing.
The jump from 28 nm to 20 nm broke the scale in expenditure, where again the passage from 14 nm to 7 nm represented a greater investment and now from this at 5 nm the largest of all ever recorded is produced. Speaking of figures, at 90 nm a FAB cost 2.4 billion, at 28 nm it rose to 6 billion dollars and the new TSMC FABs are already 16 billion.
Investments from profits, until when?
The problem is that Intel, TSMC and Samsung spend a lot on FAB, then they earn huge amounts of money, but this is not the problem, but the fact that of all that income currently the investment for new foundries exceeds the 50% in the three giants.
In other words, more than half of the earnings must be used to create new factories and since the cost is increasing and also the time that a node lasts is increasingly narrow, the cost / benefit is decreasing alarmingly. Let’s no longer talk about the cost of development and research, because that is another topic apart, or the efficiency problems of the equipment, the incremental price of ASML EUV scanners or simply the salary increase that the industry is suffering to retain the talent and not lose workers to the competition.
The problem is that one of the three companies has a certain advantage and this is nothing more than Samsung. While Intel and TSMC are dedicated to selling or manufacturing chips (either for themselves or for third parties) Samsung not only does this, but also has different businesses in multiple sectors, from NAND Flash to TV panels, through other minor tasks, not to mention that it is almost a Korean public company that employs thousands of government employees.
What does this mean? Well, its income not only depends on the foundries, but it is diversified and also its investment power is greater and is increasing, so it is difficult for the prices of its components to vary.
Prices and components
You do not have to be very smart to understand where the industry is going: constant and gradual price increase of any component for PC that integrates a semiconductor. The problem is that the industry subcontracts to other companies that in turn need subsidiaries that work with raw materials where today they are still in deficit in terms of supply and demand, but when they manage to equalize them, it is unlikely that prices will fall.
Everything that goes up must come down, but no one wants to lose profits and go into the red because the number of units of a product sold is lower. They will take the opportunity to sell at current prices, arguing that they are asked for fewer pieces or tons of material and the wheel will continue to turn as long as someone does not decide to lower prices considerably and thereby take customers away. But for this, years will pass, beyond 2023, surely by 2025 when everything returns to its normal course of the year 2019 and by then inflation will have taken a toll, so the prices of the components could remain on the same path.