AGL profits may be weak for years, and it might be time to split it in two

In ITK’s view, either investment bank utility analysts, amongst the most knowledgeable and smart people out there on their sector,  can see more positives in the outlook for AGL than us, or alternatively they are too optimistic and will end up revising estimates down. Our view is that the risks are materially biased to the downside.

If I was AGL CEO – as if

I’d be taking a strategy to the AGL board, after having been told by my friendly investment banker it’s a great idea to split AGL into two, “thermal co” and “new co”. I’d give the shares in thermal co away to existing shareholders and they can do what they want with them. This is a tried and generally successful strategy. The main risk is that “new co” is too small to take advantage of AGL’s market scale. Still if you load “thermal co” up with AGL’s modest debt load, new co would be in a strong position to grow.

Leverage to changes in generation prices

AGL has a variety of revenue sources, but as usual we are interested in what happens at the margin. Conceptually, every $1/MWh reduction in price flows through to a reduction pre tax profits. The percentage change is leveraged by fixed operating costs, fixed depreciation and fixed net interest. Will this actually happen?

AGL is vertically integrated. If AGL sees a change in the average price it receives for its coal fuelled generation, will it be able to pass it on to customers (upwards) or have to pass it on (downwards price)?

In the case where Hazelwood closed and prices rose sharply, the gentailers were able to pass the price increase on to customers and profits rose commensurately

In the case where prices fall, it’s not quite so obvious. The reason is that the large gentailers have a large retail share. In the extreme case, imagine there is only one retailer with 100% of the retail customer base. Then, even assuming generation prices fell, the retailer could leave prices to customers unchanged and profits would increase. Over time, some customers would shift to behind the meter and new entrant retailers would see an opportunity and our monopoly retailer would gradually lose market share. But retailer price stickiness could still easily be a NPV [net present value] maximising strategy.

Because of federal and Victorian government law there is now a default market offering [DMO] set either by the Victorian government or the AER. For NSW and Southern Queensland, where the DMO is determined by the AER, the wholesale power price is calculated by ACIL Allen. Reading ACIL’s latest report I was reminded that as well as the flat load price we also need to allow for hedging costs and line losses. So in the Ausgrid area the allowed wholesale price for the DMO for FY20 was $103/MWh and this compared to the actual time-weighted average pool price of $72/MWh.

In relation to hedging costs, two of the components are the caps that have to be purchased by a retailer. These protect the retailer against most extreme price events and more importantly when forecasting coal generator profits, the forward contracting strategy.

We think, and this is generally accepted in the industry, that generators won’t risk contracting all of their output. The reason is if the generator breaks down and prices rise, the generator will be exposed to the loss of buying in the market to satisfy the obligation. Equally, the future is uncertain and mostly when prices are low customers want to lock in long-term contracts and when prices are high generators want to lock in long-term contracts.

So, in practice, what we used to model is that a generator will sell 25% of expected production three years forward, 25% two years and 25% one year forward with 25% sold at spot during the year. Now, though, for simplicity we switch to a three-year hedging model.

But, of course, as capacity factors for NSW and Queensland coal generators are relatively low and falling, this will also have an impact on the absolute level of contracting.

We can see the impact of contracting by looking at Vales Point B. The accounts for Sunset Power International, yet to lodge FY20, are basically for just that one generator, plus head office.

Figure 1. Source: ASIC, Nem Review, ITK calcs

When the dust settles this will end up as a brilliant investment for the owners. Putting my investment cap on, the only thing to do is doff it to the owners and management who have run the business well. However, the point to be made is that the actual price received caught up over the 2018 and 2019 financial years with the jump in the spot price that occurred FT17, albeit still remaining below the pool price. It’s one of the many things that make forecasting difficult for outsiders.

Over the next couple of years we expect that situation to reverse. Sunset’s Ebitda (earnings before interest, tax, depreciation and amortisation) went from $55m in FY17 to about $220m in  FY19 on unchanged volumes. It would have been even higher but for some expenses characterised but unexplained as “one time”.

Sunset Power’s main retail contract has historically been with ERM, though whether that will change now ERM is part of Shell is yet to be seen. Overall though the point is it shows that, in practice, generator profits do correlate strongly with pool prices but with a two-year lag. The theory is borne out in practice.

Price data used for numbers

The prices we are using in this note are shown in the graph below. It’s important to understand that these prices are an imperfect series. There is some history, some estimates, and some model results.

Figure 3. Source: Nem Review, ITK

We have used futures for FY21 and FY 22 because they are market consensus, and it’s likely investment banks use them, and most importantly because ITK numbers for those years are similar to consensus. For FY23 on we used ITK’s estimates.

We also note in our forecasts that Matt Kean, NSW energy Minister, is hoping his program will lead to NSW prices in the $45 -$51/MWh. That is broadly in the range ITK estimates. So there are lots of reasons for investors to take it seriously.

Getting back to AGL

There is much more to AGL than just wholesale prices for coal-fired electricity. Still if we look at that in isolation:

There are two obvious points to make, firstly the announced closure of Liddell and secondly the complete public uncertainty about what is going to happen to Portland smelter. The current $200m support package for Portland expires in June 2021. ITK’s very limited understanding is that if a smelter was contemplating closure it would tend to turn off the capital expenditure program years ahead of time. Specifically it would stop replacing pots (pots have a about a six-year life and some are normally replaced every year).

According to an AFR story from October 20 there are indications that the federal government as well as the state government are in talks. ITK’s view is that rather than just a power price subsidy both the federal and state governments might choose to couple support with a commitment from Portland to green up the smelter. I would argue that it would be worth doing a study to see whether using the smelter as a power sink could help with what we see as otherwise intractable ramp-rate issues at Yallourn.

This is the case in point for the uselessness of forecasts. There is a basic issue of whether 4.5TWh of demand is actually going to be there, or not, in a couple of years. It’s long odds on that if Portland closes it will take Yallourn with it.

For this exercise ITK assumes Liddell closes and Portland stays open.

In the end, for a conceptual exercise, we made a straightforward assumption that the actual price received by AGL is an equally weighted average of the current year and the two prior years. This smooths out the price effect in any one year but over a decade should give a reasonable indication.

AGL will also avoid some capital and operating costs when Liddell is closed. Looking at the Sunset Power numbers I think an operating cost of say $40/MWh is good enough for this exercise. Liddell’s number may well be lower due to costs shared with Bayswater, in which case the negative impact on profits of closing it will be larger.

In regard to capital costs for the past few years ITK has been using $10/MWh as the likely average maintenance capex cost of coal stations across the NEM. Vales Point has been running about $3/MWh but capex is lumpy. Using $10/MWh is $90m at Liddell which seems too high for a power station at the end of its life where safety is likely the only capex justification. So lets say $50m, but that is not to the profit account anyhow.

Using this approach we see coal generation revenue earned by AGL, net of operating cost savings from Liddell, falling by almost $700 m over the next 5-7 years and neither do we forecast any recovery. To reiterate, the volumes are flat, save for Liddell closure, prices used are a three-year rolling average, and cost savings from closing Liddell are added back.

Remediation

Another thing to think about with AGL is remediation costs.

AGL’s Environmental restoration provision at June 2020 is about $340 million and represents the NPV (net present value) , pre tax, of the expected cost of restoration of all power stations and gas fields.

If you grow $340m @ 5% for 20 years its over $900 m. In AGL’s public view the Loy Yang A brown coal generator is not expected to close until 2048 and thus its environmental restoration cost doesn’t start to fall due until then. Bayswater is scheduled for 2035. No one seriously believes LYA will be going until 2048 but you can’t prove otherwise. Still it makes me smile.

For a couple of years around FY14 and FY15, AGL published a separate sustainability PDF which had, as I recall, total environmental costs, undiscounted at about or over $900 million. However, I have been unable to locate that document or confirm that number in recent online searches.

Nevertheless I feel confident, particularly, as the known Hazelwood clean up cost was estimated at $743 million back in 2017, that AGL’s undiscounted costs of remediating Liddell, Bayswater and LYA will be at least $900 million. The discount unwind each year goes to the P&L interest expense. Utilisations of the provision don’t go to the P&L but to cash flows. The NPV at time of establishment goes into the asset cost on balance sheet and is depreciated at the same rate as the asset.

Figure 7 source: Company

Clean up costs are obviously a major issue for many power station owners and typically will act as a drag on closure decisions. AGL will be incentivised, from a P&L perspective, to allocate as much of the combined Liddell/Bayswater cleanup cost to Bayswater. From a tax management perspective it’s the reverse.

The point here is, to the extent that closure of LYA and Bayswater is accelerated by more renewable energy, then to that extent the NPV of the environmental costs increases.

There is more to AGL profit estimates than just coal fuelled generation

In this note we have discussed coal fuelled generation and AGL. We have not talked about the gas business, and I mean wholesaling and retailing of gas, as opposed to gas fired generation, nor have we talked about AGL’s wind farm revenue and costs. Here, it’s probably worth mentioning that AGL has a number of wind farm PPAs where it has contracted to pay higher than current market prices for wind output.

Readers may recall that the Macarthur wind farm, still the largest operating project in the NEM, has seen its “value” increase every time a piece has changed hands. The reason for this is that buyers are getting a fixed revenue stream and can finance that with ever lower interest rates. The flip side is that it’s AGL making the cash payments every year. We think the negative value of these wind PPAs (revenue less payments) represents the great bulk of AGL’s $138 million at 30 June 2020 onerous contracts provision.

Nor have we talked about the outlook for AGL retail volumes and margins and still less on recent “toe in the water” telcos “initiatives”.

As far as AGL’s historic and consensus forecast EBITDA goes the chart looks as follows:

Figure 8. Source: Factset

The shape of consensus earnings which are for the consolidated results of AGL matches our view of coal generation. In regard to FY21 outlook AGL stated at the time of its results presentation;

The underlying NPAT estimate was in the range of $560 – $660 m (mid point $610)

That range includes a $80 – $100 m after tax benefit from insurance claims. (ie the excluding one off underlying profit guidance was $610 – $90 = $520 m)

Wholesale gas margins were expected to be $150 m lower and

Wholesale electricity margins were expected be $150 m lower.

ITK’s “model”, which ignores changes in coal costs that could benefit AGL in FY21 and maybe FY22, would have electricity margins down by $271 million in FY2020 and by a whacking $436 million in FY2022, and a still material $200 million in FY23.

So in the extremely unlikely event that ITK’s very simple sensitivity was capturing what’s going on we would see consensus as still on the high side, or alternatively that there is upside in those out years from factors we haven’t considered in this note. Without having had the benefit of discussing these numbers with management, and that could well lead to a better and revised understanding, our view is that the preponderance of risk is to the downside compared to consensus.

Other things could change the outlook. The most likely, in ITK’s view, is early closure of a generator as well as Liddell. Although that wouldn’t happen before 2025 prices could react well before that. ITK’s broad estimate is that the closure of a typical generator would move prices up by say $10-$15 MWh for a couple of years. The market is much better prepared for another closure as compared to Hazelwood.