Based on the analysis here, CRU estimates that true excess capacity was just under 300 Mt globally in 2016 and this should fall to ~115 Mt by 2021, assuming China is able to close down 160 Mt of capacity. Most this excess-capacity will be in China (i.e. ~35 Mt), although this will only account for ~3% of overall capacity in the country, a much lower share than for most other regions.
The steel industry and the issue of over-capacity
Over-capacity has been a major point of discussion in the steel sector for the last 30 to 40 years and certainly since large parts of the industry, outside of China, moved out of government hands through the 1980’s. During this period, the desire of national governments to maintain a ‘strategic’ industry was overridden by the economic imperatives associated with the ongoing financial costs to support inefficient and loss-making operations. Whereas before, steel sectors were configured to service national requirements and to provide employment, they were subsequently exposed to the commercial realities of the global market. The reactions of individual companies were broadly the same: reduce costs, improve efficiency, invest to maximise volume to achieve scale economies and improve quality and value-add. For individual companies these responses are perfectly rational but, they can be catastrophic for the industry as a whole, particularly when everyone is chasing the same vision and making investment decisions that have ramifications on capacity many years into the future – and certainly through the steel cycle. But is this something that can be avoided or is the steel industry, particularly susceptible to overinvestment? Furthermore, what is China’s role in all this? To answer these questions, we first examine the scale of the problem.
How much over-capacity is there and where is it?
The graph below draws on CRU’s Steel Capacity Database to understand capacity utilisation at the crude steel production (i.e. cast steel) level in key steelmaking regions over the last 15 years or so. This indicates that, for the 5 years prior to the global financial crisis (GFC), globally, the steel sector was operating at just under 80% capacity utilisation, however, since the GFC, it has been operating closer to 72%. So, even during the former period of strong demand, the industry overall was unable to achieve very high levels of capacity utilisation, partly due to the rapid build of capacity in China in response to significant increases in domestic demand. A simple interpretation of this graph implies that, at best, there is 20% of surplus capacity in the world today and at worst, as much as 30%, equivalent to 465 Mt and 700 Mt respectively. But why is this important?
Demonstrating why over-capacity is such a concern, this graph illustrates the notional profitability of the European steel sector at different capacity utilisation levels, based on output from CRU’s Steel Cost Review, and indicates that sustainable levels of profit (i.e. defined as 7.5% EBITDA margin or higher) are only achieved above ~85% capacity utilisation; this is a function of the cost structure and revenue curve of individual mills.
To illustrate why this is the case, the graph below sets out a generalised structure of revenue and costs for an individual steel mill. Thus, every steel mill has its own revenue curve representing each tonne of steel that is sold, from the most valuable to the least, and its own cost structure that can be broken down into fixed and variable costs. As illustrated in the graph below, even under trend conditions (i.e. represented by the blue revenue line), each steel mill will be selling some material at below full cost, with the last tonne sold at close to marginal cost (n.b. probably general purpose steels sold to more distant markets). This is rational, as it maximises production and lowers average costs, such that overall profit is maximised. However, as demand falls (i.e. represented by the lower grey revenue line), the final tonne of steel will still be sold at close to marginal cost, but the much shorter revenue curve and higher average costs (n.b. not shown in the graph) mean that a greater proportion of sales are loss-making. In this way, as capacity utilisation falls below 85% (n.b. because of lower demand or capacity growth or both), steel mills will fight for the fewer sales that are available and each will lower their prices to maximise market share, which pulls down the revenue curve and profitability. Specifically, what the above implies is that over-capacity is an issue involving all steel mills. That is, even under steady-state, or trend, conditions, most steel mills will be selling some proportion of their steel at below full cost to more distant, lower value markets and so each makes its own contribution to the overall capacity surplus. Therefore, it is not possible to single out specific mills that are wholly responsible for the over-capacity. But is this ‘over capacity’ at individual mills, perhaps defined as that volume of production sold below full cost under steady-state conditions, really excess-capacity? The following section attempts to answer this.
Is over-capacity always excess capacity?
This may seem like an odd question but, hopefully, the below will demonstrate that over-capacity is not always excess capacity and, in fact, that over-capacity is hard-wired into the cost structure of steelmaking. Let us begin by looking at the typical, underlying cost structure of an integrated steel mill. At a high level, a steel mill has both fixed and variable costs. In the short-term, the fixed costs cannot be changed, therefore, as the mill operates at lower volumes (i.e. lower capacity utilisation), unit costs will rise, as illustrated in the graph here. Additionally, other factors will contribute to the change in costs, such as yield and energy efficiency, which deteriorate at lower levels of capacity utilisation and vice versa. However, what is important to note is that there tends to be a minimum for costs that typically occurs around the 85% capacity utilisation level. This behaviour can be explained by the configuration of integrated mills and operational imperatives.
That is, no integrated mill has what might be called the ‘perfect’ configuration. All steel mill configurations will be the result of decisions taken over many years, compromises made due to available capital, land area constraints, local regulations etc. As such, actual capacity of the whole operation is unlikely to directly reflect the nominal capacity of individual facilities. To illustrate, the table below sets out the ‘ideal’ capacity of a specific integrated mill based on the nominal capacities of individual assets. The ‘ideal’ capacity assumes that the full output of the facility in question (e.g. coke plant) is consumed to make steel and no 3rd party purchases, or sales, are made.
The table indicates that the mill in question is not perfectly balanced such that, if operating at 4.7 Mt crude steel, the maximum output based on the size of BOF shop, then some 3rd party coke, agglomerated blast furnace burden and grid electricity would need to be purchased, which would incur a cost, and the casters would be operating below capacity. This would not be irrational necessarily, particularly if the steel market is performing well and steel prices high, as the added costs would be justified.
However, what we can also see from the table is that, more routinely, this mill has operated at ~87% of maximum capacity, probably because this is an operational and cost ‘sweet spot’ for the configuration in question. This supports the notion that over-capacity is built into the cost structure of steel production and, therefore, cannot be eliminated (n.b. even greenfield mills will suffer to a degree due to non-achievement of design specifications, but also as input price changes undermine design operating practices).
Further, for a mill, particularly of the complexity and sophistication of a modern integrated steel mill, to operate close to 100% capacity utilisation all the time would be difficult operationally and likely to be counter-productive, as well as potentially unsafe. Under these conditions, any disruption to production, minor or major, will likely have significant impacts on the ability of the mill to deliver and the supply steel would be affected. This would likely lead to higher prices and certainly price volatility (n.b. a similar situation existed in the metallurgical coal sector towards end-2016, with Australian exporting mines operating close to 100% capacity utilisation experiencing operational issues, which contributed to high prices). The costs (n.b. capex. and opex.) associated with the added maintenance to support production close to 100% of capacity would also be very high and any failure to sustain higher maintenance standards could lead to accidents. As such, it is my view that it would be unrealistic to expect the whole of the steel industry to be operating with such stringent operating requirements under steady-state conditions.
Finally, the characteristics of the steel market itself dictate that over-capacity is a pre-requisite in order to ensure security of supply and some degree of price stability. The graph here sets out quarterly steel demand relative to mean demand by select regions, chosen based on their stable demand situation during the last 12 years (n.b. China could not be used in this analysis because of the strong growth of steel demand during this period). What this shows is that quarterly, or short-term, steel demand can be >10% above mean demand, but can also fall to almost 85% of the mean. This is a wide variation, but the implication is that, in order to satisfy short-term steel demand variations, steel mills need to able lift production to as much as 12% above steady-state levels. The alternative would be to produce at a more stable production level, perhaps at higher capacity utilisations, but to make to stock – the variations in demand could then be met by stock build/drawdown, as appropriate. However, in order to work properly, such an approach would incur costs of its own and require a greater understanding or visibility of demand changes than is available today. The implication of the above graph is that, for the great majority of the steel cycle, steel mills will, in all likelihood, be operating well below their full capacity, but this is more out of necessity than choice.
Over-capacity and China
According to CRU’s Steel Capacity Database, globally, there were ~2.33 bn t of installed carbon crude steel making capacity in 2016, with 1.14 bn t in China, and the industry was operating at ~72% capacity utilisation, compared with 76% for China. This implies over-capacity of ~650 Mt globally, of which 267 Mt is in China. However, based on the above, we believe that up to ~15% of over-capacity is hard-wired into the steel sector and cannot be eliminated, therefore, the true level of excess capacity is closer to 300 Mt globally, of which ~95 Mt is in China (i.e. less than one third for a country that accounted for 52% of global carbon steel production). Having said this, 2016 represented a period of relatively depressed demand globally, so the true level of excess capacity is somewhat below this value, perhaps closer 280 Mt.
Based on CRU’s forecasts for steel demand in the medium-term and expected closures of capacity in China particularly, which we believe will amount to about 160 Mt (n.b. includes 2016 closures; is higher than the current government target, but is in line with provincial targets), we expect capacity utilisation of the industry globally to lift to ~80% and for China to be at ~82%. Assuming the steel market is closer to steady state demand at this point, this implies a true excess-capacity of ~115 Mt, of which 36 Mt will be in China (i.e. ~3% of local capacity), 18 Mt in SE Asia (i.e. ~32%), 16 Mt in North America (i.e. ~9%), 14 Mt in Africa (i.e. ~34%), 12 Mt in South America (i.e. ~15%), 8 Mt in the Middle East (i.e. ~14%) and 7 Mt in Europe (i.e. ~3%).Explore this topic with CRU
When we last looked at the IMO 2020 MARPOL Annex VI policy back in 2018 Q4, there were many sources of ambiguity in terms of the outlook.