The never-ending story: Netflix and the ever-escalating cost of content

Regular readers will recall I have been following Netflix (NFLX) for many years (I wrote up some initial thoughts on this blog a few years ago), but as the service has grown from ‘challenger’ to ‘near-ubiquitous’, I find myself returning more often to the company, its growth philosophy, and the underlying existential question for any business: will it every be sustainably cash-generative?

This post was sparked by a recent discussion I had with a friend on the topic. More bullishly-inclined, he made the (reasonable) argument that NFLX was smart to swap profits/cashflows for more subscribers today, because once ‘at scale’ the company could trim its content (and marketing) spend and still provide a compelling service – so compelling, he thought, that NFLX could also raise prices substantially. My bearish counter was based upon two points: 1) the benefits of scale from this point have diminishing returns, because a huge amount of incremental content spend is not universal in demand (the incremental subscriber is not invariably a high-paying English-speaker and so needs unique local-language content that in most cases cannot be amortized globally); and 2) even if we granted them point 1) was theoretically possible, the recent history of the company – as it has grown from 50mm subscribers to 135mm in the last 5yrs – has actually demonstrated the opposite property: more incremental spend on content for less incremental benefit, not the other way around.

This second point in particular got me thinking – can we demonstrate, numerically, exactly how much more expensive it has been for NFLX to grow in recent years as the business has expanded beyond its home market (the US and English-speaking markets)? This is something I’ll explore in this post.

How much is a new subscriber worth? The ‘traditional’ method

Let’s start with some basics. For most subscription-based businesses, Silicon Valley types try to measure the value of the proposition by comparing the cost to acquire a new subscriber (‘customer acquisition cost’, or CAC) with the ‘lifetime value’ (LTV) of that acquired customer. CAC is quite easily defined: it is simply marketing expenses dividend by net new customers acquired in the period. This is how it looks for NFLX since 2014 (using TTM numbers to smooth quarterly fluctuations in marketing spend):

Screenshot 2019-07-08 15.55.44

So CAC has increased from <$50 to ~$75/paid member in the last 5yrs – not great, perhaps, but understandable in the context of a largely-penetrated home market that is subject to increasing competition (the US); and more aggressive international marketing in countries where NFLX was less well-known or where OTT TV was less understood (the rest of the world).

Of course, to provide meaning to the CAC, we need to compare it to LTV, since if LTV is similarly increasing, paying a higher CAC wouldn’t really be a problem. LTV, however, is slightly more complex to calculate than CAC, because we need to make an assumption regarding how many subscribers NFLX loses on an annual basis (‘churn’) – a number NFLX conveniently has declined to disclose since 2010. We also need to make an assumption regarding the future earnings power of each subscriber (measured in terms of gross profit per subscriber in a given period).

Still, we can take a shot at it. In the below, I have assumed annual churn is 25% (so just over 6% per quarter); this is actually lower than cable TV churn (around 7% a quarter), despite the fact that NFLX is cancellable any time without penalty, so I would argue this is not aggressive. I have also simply aggregated gross profits (again on a TTM basis to smooth quarterly fluctuations), and as a result LTV (over a similar period) looks something like this:

Screenshot 2019-07-08 16.04.01

You can see that despite the fall-off in LTV in recent quarters, the absolute number (around $180) is still well in excess of CAC ($75-80) – meaning when we plot the all-important LTV/CAC ratio chart, we see that despite the decline of late, growing subscribers still appears to be value additive to the enterprise (even if the pace of recent decline should be of some concern):

Screenshot 2019-07-08 16.07.54

So when measured in the ‘traditional’ way, NFLX’s astounding subscriber growth still appears to be generating net positive lifetime value per customer to the company (uncertainty around churn not withstanding), validating the bull thesis, right? Not so fast…

The increasing cost of content bogey and the need to adjust LTV accordingly

The main problem with the above approach is a conceptual one: once the customer has been acquired (through marketing spend), the assumption is made that – for those customers that don’t churn out – you don’t need to ‘re-acquire’ them – whatever you spent in marketing dollars to get them in the door, that is the total cost to you to keep them on board. But what about the cost of maintaining them as a subscriber? In many software/SaaS businesses (where this methodology originated), that may truly be a very small incremental cost. But NFLX makes no secret of the need to spend exponential amounts of money on new content that – purportedly – drives more eyeballs – but (and this is the key) in reality exists to maintain existing subscribers. I think it’s fair to consider incremental content spend by NFLX as an added cost to maintaining its subscriber base for three simple reasons: 1) not many people watch the same content twice (meaning once consumed, content needs to be replaced); 2) increasing engagement – something NFLX is ironically proud of – increases the velocity of needed content replacement, effectively driving up the maintenance flywheel against the company; and 3) alluding to my earlier point, a lot of the new content being created in recent years has a much smaller intended audience than earlier content (given increasing localization, etc).

In other words, the traditional metric associated with measuring the value of a customer needs to be modified, to account for the excess $$ being dedicated to growing the content pool which – even if amortized over an increasing number of paid users – should nevertheless be thought of as the cost to ‘re-acquire’ the customer, continually, to stop them quitting and going to another service (of which there are now many, and growing).

Here’s how I’m thinking about it. In the traditional LTV calculation, we look at gross profit per subscriber to estimate what the potential value that customer could bring in over its lifetime could be (before he/she churns out). ‘Gross profit’ is essentially average revenue per user (that is, average monthly subscription cost) less the total cost to deliver the content to the customer (the overwhelming majority of which is content amortization, in other words, content production costs spread over a 10yr assume content ‘life’).

But the strict income statement entry – ‘cost of revenue’ – is, as we have discussed, not the only run-rate cost of content to the company: NFLX has been spending more and more on content, in escalating fashion in recent years, and this is reflected in the difference between the content amortization cost in the income state (‘Cost of Revenues’ in the PnL) and the net incremental cash out spent on content costs (in the Cash flow statement).

Pulling it all together from NFLX’s financials, here is the incremental content spend (again TTM for smoothing) over the last five years:

Screenshot 2019-07-08 16.27.51

It’s worth dwelling on the significance of the above. Five years ago, NFLX was spending ~$500mm a year on content than it was currently amortizing; today that number is almost 10x greater ($4.5bn) and they are still guiding to spend more!

Of course, this escalating content spend has come along with massive subscriber growth so we need to relate the increased spend to the enlarged subscriber base. Below, I show the incremental content spend beyond the PnL, on a monthly basis (using the average TTM paying subscriber base as the denominator) – you can see that even on an equalized basis, NFLX is spending >3x the amount on incremental content beyond the PnL, per subscriber, than it was in 2014…

Screenshot 2019-07-08 16.34.02

Coming up with a modified LTV/CAC calculation

Considering the above, I think it’s fairly obvious that we need to modify the ‘traditional’ LTV/CAC metric to account for this increasing cash cost of content (even if it hasn’t made its way into the PnL yet). We can do this simply, by subtracting the incremental content spend, from the gross profits, per paying subscriber (we don’t need to change any of the CAC calcuation, since marketing costs don’t need adjustment). We do this, and voila, all of a sudden (what I would argue) the true LTV/CAC doesn’t look so rosy (including the ‘traditional’ LTV/CAC for comparative purposes):

Screenshot 2019-07-08 16.42.08

I consider the above chart fairly damning evidence against the NFLX bull narrative, as it demonstrates a) the business has essentially been adding negative value subscribers since late 2016 (when content spend really kicked into high octane); and b) the increasingly negative value nature of the proposition has been accelerating in recent quarters.

However if you were a bull, and even if you were willing to acknowledge my methodology had merit, you might push back in some or all of the following ways:

  1. this methodology doesn’t take into account future price increases;
  2. NFLX won’t have to spend current levels on content in the future;
  3. NFLX can increase revenues in other ways (advertising, etc).

Of these, I think 2) is fairly easy to dismiss, since NFLX has avowed publically they will continue to outspend all the competition on content, for the medium-term – and since competition is only increasing (Disney+, Hulu, Amazon Prime video, HBO go, CBS, Apple, etc etc), and since some of these competitors view OTT almost as a loss-leader (Amazon, etc), I don’t think it’s really credible to envisage a world in which NFLX voluntarily cuts content spend (as opposed to one in which the market forces them to slow down spending ). Similarly, 3) is fairly anathema to what NFLX is purportedly trying to create – ad-free high quality on-demand content – and I would expect churn would sky-rocket if they introduced ads onto the platform. And once again, given the competitive landscape, this seems unlikely unless most/all competitors move in a similar direction.

This leaves price hikes. I think the point is a reasonable one, and has been pursued by NFLX historically – the question is really one of scale. Certainly there is scope to hike prices somewhat, especially in developed markets; but NFLX is increasingly an international story (essentially all sub growth comes from outside the US now) where the legacy of high cable bills (and level of GDP wealth) is simply smaller. Today, over half the subscriber pie is international, and with the US mature, this ratio will only grow – limiting, I think, the upper bound of price for the increasing majority of subscribers

Moreover, we still need to question the scale of any price hikes in relation to the escalating content bill. The chart below shows the percentage of ARPU that is currently consumed by incremental content spend beyond the PnL: this number was 10% in 2014 but is nearly 30% today:

Screenshot 2019-07-08 16.57.14.png

That is to say – NFLX could manage to raise prices 30% across the board, unilaterally – and it would still only bring ARPU into line with the escalating cost of content (assuming it stopped where it is, which it won’t). That kind of price increase may be stomachable in the West (even at the cost of a big jump in churn, perhaps temporarily), but it would necessarily price out large portions of the emerging world (that NFLX needs for its growth story).

Where does that leave us?

This has been a long exploration, but I’m fairly confident reiterating my early prognostication: NFLX is an amazing service, but I really can’t see how it will ever generate sustainable free cash flow, let alone an adequate return on capital – let alone a return adequate to justify the current $180bn valuation (!). That doesn’t make it a short (and I’m not), but unless you expect an uber-benign future where NFLX is free to raise prices with impunity whilst reining in the excessive content spend it and its competitors have been enacting exponentially, it certainly doesn’t make it a long either.

Disclosure: no position in NFLX












Tesla’s unit growth, and the art of spurious comparisons

There’s a fairly fundamental concept in most retail industries: the concept of same-store sales (‘like-for-like sales’, ‘equivalent-store sales’, ‘existing-store sales’, etc). This is a pretty intuitive concept, but to recapitulate by example, let’s say I’m a McDonald’s franchisor, and last year I had 100 stores that did $100mm in revenues ($1mm/store), and sold 250k Big Macs at $4 a pop (this is a very minimalist hypothetical McDonald’s :)). That’s great but I want to grow, so let’s say I open another 100 stores. Unfortunately though I over-estimated demand for Big Macs, such that the new stores underperform and draw traffic from my existing footprint (there are lots of other burger joints out there, of course), and I am forced to cut prices on my burgers to $3.5 each to maintain growth in my burger volume (necessary to keep all those grills in the new and old stores busy). At the end of the year I report 350k burgers sold and $122.5mm in revenues – both new ‘records’ for my franchise operation – but clearly a large decline in same-store-sales, and, as a result, profits (I doubled the store count and a huge component of fixed costs but only increased revenues 22%). It’s quite easy in this example to see how revenue growth for its own sake actually destroys value.

Which brings me to Tesla (TSLA), which just reported ‘record’ quarterly deliveries of 95k vehicles, versus the Q4’18 performance of 91k deliveries. Bulls are busy trumpeting this as proof that demand is strong and sustainable; but I think if you break down this theoretical record by analyzing the geographic footprint of where the sales are coming from actually suggests the opposite. Once you realize that unit volume is being driven by a combination of new market entries and price-cutting, it’s pretty clear that TSLA’s unit growth is only so many empty calories.

The analogy to a hypothetical burger chain is not perfect (because TSLA isn’t opening new factories in all of its new locales and therefore doesn’t face the same fixed cost burden on incremental sales), but conceptually it is somewhat instructive. Lost in the bullish headlines this morning is the simple reality that in Q4’18, TSLA was only selling the Model 3 in the US (its core market); by now the Model 3 is available worldwide (with the exception of some very small right-hand drive markets) – tenuously equivalent to opening a vast number of new ‘stores’ to achieve a marginal increase in sales. Instead, if we look purely at the US market, where the Model 3 has been out for the entire year (both 2H 2018 and 1H 2019), the unit performance looks unimpressive to say the least: Model 3 deliveries fell 42% half-on-half in 1H’19 versus 2H’18 (67k vs 116k, per InsideEVs). Incidentally, this half-on-half decline is mirrored in the performance of the Model S (-52% HoH) and Model X (-46%) as well.

In other words, in Tesla’s most mature market – and despite significant price cuts and the introduction of the much lower-priced Model 3 SR+ – unit sales are falling rapidly half-on-half. It is hard to argue there is any seasonality involved, either: since 2014, the average HoH change in US SAAR is negligible (see below), so it’s not as if there is an external reason for the large decline in ‘same-market sales’ in the US:

Screenshot 2019-07-03 10.33.36

Rather, it seems fairly self-explanatory what happened: there was a large, one-time surge in demand last year in 2H, as all the pent-up demand for the Model 3 was fulfilled in 2 voracious quarters; thereafter, run-rate demand (in the US) seems to have moderated to much lower levels, even at lower prices (actually there are still one-time pull-forward effects in the US in 2Q, like the halving of the Federal Tax Credit once again, but let’s leave this aside for now). This is not readily apparent in the gross numbers, because of the aforementioned opening up of new markets. But as we lap 2H’19 comps, and as Europe/China mature in the coming quarters, it will most certainly become much clearer (and signs are already pretty bad re run-rate demand in Europe, especially, as daily registrations have rolled over much faster than was seen in the US).

Thus in recent quarters, TSLA has found itself somewhat in the situation of our hypothetical burger franchisor: wanting, needing to show growth (to defend the stock price) but with underlying unit growth receding in its existing core, having to continuously open new markets, as well as cut prices, to maintain gross unit momentum.

The effect of price cuts should be readily understood, but is something the bulls conveniently ignore. Taking the simplest measure of average selling price (total automotive revenue divided by unit deliveries), TSLA’s ASP’s have been declining fairly consistently over the last 5 quarters (ignoring 2Q for now as we haven’t seen the print yet, though I expend the trend to continue):

Screenshot 2019-07-03 11.13.53.png

While of course a good chunk of this is due to the introduction of the Model 3 progressively from 2Q last year, we also know TSLA has cut prices at least 3 times in the last six months on all its models; and has been known to push large end-of-quarter discounts to clear inventory (as happened in 1Q).  Finally, it demonstrates the futility of TSLA achieving unit growth for its own sake, at the price of ASPs. Consider that in this 2Q, ASPs only declined modestly from 1Q (and thus are around $60k blended), TSLA will have shipped 5k more units than in 4Q’18 – a quarter that was only very marginally profitable ($200mm), excluding subsidies – but at the cost of 5k x ~$10000 gross margin per vehicle, or $500mm in total and thus putting the company back squarely in loss-making territory. That doesn’t sound much like a record quarter, to me…but it is a trick Elon and co will have to repeat, ad nauseam, to satisfy a stockmarket pricing in ongoing aggressive unit growth in the coming years.

Since we all learnt in Economics 101 that demand is a function of price, this all goes to the heart of the bull/bear debate and why the bear position – one I occupy – is so intractable. Any fool can tell you that if you sell Rolexes for $500 a pop your unit growth will be incredible; you’d also be hard pressed to find someone arguing such an idea would be a smart business move. Selling something, anything, of value demonstrably below cost will undoubtedly move units – but to confuse this with creating economic value is borderline insanity. If you are marketing a consumer product that is currently unprofitable (check), that has demonstrably sold its highest ASP products ever (check check) and cannot grow units on a like-for-like basis without cutting prices (check check check), how will it be possible to ever generate sustainable profits (let alone an adequate return on capital)? The market may close its eyes and bathe in the glow of ‘record unit volume’ for a day, a month, or longer, but in the end if the business proposition is unsustainable, the stock will follow.

Disclosure: short TSLA






NIO follow-up: fun and games in the 20-F

Due to the enthusiastic response to my first missive on NIO, I thought I would delight all my fans 🙂 with a quick follow-up post the 20-F (annual report) filing today. Here are my main takeaways; the overarching conclusion remains ‘alarm bells are ringing’ and if anything I think the cash needs at the company are more intense than I first anticipated.

In no particular order, here’s what jumped out at me from the filing (with pagination so you can find it within the doc if you like): –

Capex: they are guiding to $600mm in calendar year 2019, a good deal higher than my baseline $500mm cash burn/qtr run-rate estimated (I had been assuming capex would actually fall a little from last year’s level, $400mm, since they decided to kill the manufacturing plant). In other words, if they hit their capex spend and all other assumptions stay the same, quarterly cash burn is more like $570mm (not the $500mm I anticipated). The nature of the capex, too, seems of questionable value: only $26mm was spent on power solutions/infrastructure for their energy network, with ‘R&D equipment’ being a key expenditure (amongst tooling, office equipment, leasehold improvements, etc). They do not disclose how much R&D they are effectively capitalizing in this way. They explicitly mention that equity/debt financing will be conducted if the proceeds of their convert offering are insufficient to cover capex needs (p113).

Users on their app: I thought the disclosure around their user base was a little…strange. The bulls like to talk about the ‘800k+ community of Nio app users’ as evidence the company has built meaningful scale. Here is the precise disclosure from the 20-F (p. 69): We aim to engage with users and create an environment conducive for user interaction both online and offline. Our mobile application had over 760,000 registered users as of December 31, 2018 and over 190,000 daily active users on peak days in 2018 (my emphasis). Clearly this is not the standard definition of ‘daily active users’ as they don’t even specify the time period (typically this number would be calculated monthly or quarterly with the peak usage only incorporated on this basis). Putting aside how they define ‘active’, though, if they only had 190k users on peak days, who knows what the real DAU number is, 100k? 50k? In any case it is suggestive of a far smaller fanclub than the 800k number bandied about by the sell-side. Anytime the company can’t give you a simple number on a standard metric like DAUs, it’s a concern.

Expensing of points: they disclose that one of the ways they get people to sign in to their app is for ‘Nio Credits’ which can be used to then get merchandise, other discounts, etc. There is one tiny disclosure on the cost of this program (F-19, in the notes to the financials) – they expensed 144mm CNY in 2018 on this program. This constituted the bulk of their Sales & Marketing expense (218mm CNY in 2018). But in any case, these Nio Credits alone meant that they spent $22mm to acquire their app users, oh which they had only 190k active on ‘peak days’ and thus, assuming 50-100k real DAUs, customer acquisition cost (CAC) from this cost line item alone is $220-440. This seems incredibly high…especially as it takes into account only a tiny part of the brand-building expenses actually being incurred (Nio Houses, etc).

JAC manufacturing fees/cost absorption: they disclosed a total of 223mm CNY payments to JAC, their manufacturing partner, in 2018 – broken down as 97mm CNY of manufacturing fees and 126mm of loss reimbursement. This equates to 8k CNY of manufacturing fee per unit prouction (fixed per vehicle), and 17k of loss reimbursement per unit. The most interesting thing about this is that they had accrued 65mm of loss reimbursements as of end June 18 – meaning 60mm CNY or so of losses related to 2H last year. This is a problem because it means their outsourced manufacturing solution is still unprofitable even with higher volumes (25k annualized run-rate production in 2H’18) and suggests they will be gross margin negative for 2019 as well (lower volumes on ES8, ramp/higher costs on the new model from 3Q).

Reservations: they stopped disclosing ES8 reservations in the filings (this was mentioned on the conference call at 4Q earnings) so not a surprise but there is nothing written down about ES6 reservations in hand either. These were specifically discussed on a conference call, and ES8 reservation numbers formed a meaningful part of the (written) IPO narrative. Furthermore the rationale given on the call for stopping disclosing ES8 reservations was that the vehicle was already in production. But that clearly doesn’t apply to the ES6, so why the unwillingness to put reservation numbers down on a formal filing? Are they suffering many cancellations? Are they worried about getting in trouble with the SEC?

Employees: this one really needs an explanation. In the detailed discussion of why expenses were up so much in 2018 vs 2017 (p108), they say an increase in the number of our research and development employees (including employees of our product and software development teams) by approximately 75% from 2017 to 2018 was one of the main factors. The only problem is according to their own disclosures later in the document, R&D employees only went up by 14% yoy (from 3052 to 3492) – that doesn’t sound like ‘approximately 75%’ to me. Overall employee numbers did skyrocket (+40% yoy to around 9800) but the vast majority of these were in ‘Sales, marketing & service’ – ie customer facing roles, which, primarily in China, you would think garner only a fraction of the salary of engineers. Notably, the overseas R&D employee numbers – where a bulk of the very high comp would theoretically sit – didn’t change that much (844 vs 724, so +16% yoy).

I am not sure if this is a big deal or not – but I was already concerned they were massively over-spending on R&D and now it appears they are at the very least unintentionally saying they’ve hired a ton more engineers than they really have. Who knows where the spending is really going – but if the reported numbers are to believed then average R&D employee salaries went from $49k to $79k in a single year, while average company-wide salaries went from $41k to $62k (+50% yoy, not bad)…and remember this is almost all in cash, stock comp was but a very small part of total comp (just 7% of total opex).

So all in all, many more yellow flags and not much to really assuage my worries that they will hit the market for another $1bn+ in cash as soon as practicable. Clearly stock has rallied (my timing is always impeccable ;)) but still a high conviction short.

Disclosure: short NIO

NIO: this will end in tears

Forget where the stock has come from and the fact that it is already down 50% from the highs. NIO (NIO, listed on the NYSE), the recently-listed Chinese BEV name, is a hype-driven, cash-incinerating JOBS-act IPO that may qualify as the best poster child for ZIRP’s distortionary capital allocation effect when the history books get written years from now (great blog here on this topic by the way). In the meantime, it’s a great short because it has no moat, de minimis brand, competes in the world’s most crowded EV market, has a structurally high-cost model that is a big competitive disadvantage longer-term, has disclosed weaknesses in internal controls along with black-box cost accounting and a number of other oddities, and will suffer badly from just announced cuts to Chinese EV subsidies. However, you can throw that all out if you like and just focus on the main thing: NIO is a cash-incinerating machine par excellence: since founding barely 4yrs ago they have run up $5bn in accumulated deficits, are currently burning around $500m PER QUARTER, and have no clear path to profitability in the near-term. The scale of the cash burn is such that they had to rush out a convert offering to raise $700mm, just 3mos after the IPO raised $1bn…relying on the artificially-inflated stock price during the pre-lockup period. This prompted numerous lawsuits (the bonds have since cratered with the stock) but the upshot is now, no matter what, they need to come back to market for another $1bn in 6mos (or sooner), such is the scale of their cash burn (I judge they’ve already burnt most of the convert cash they raised).

So to my mind you have multiple ways to win with this short. If China blows up, this gets crushed and/or disappears (luxury auto highly sensitive to the macro economy). As or when they run out of cash, there is a real chance there will be a ‘run’ on the company, perhaps by its suppliers (working capital is already massively negative), the Chinese banks (who are already significant lenders) or as the market realizes the US-listed paper is worthless. Or they simply need to come back to market in 6mos and raise another $1bn at whatever price the market is willing to bear (hint – its not a $5bn EV for a company burning $2bn cash a year). The convert window is now closed (given what happened recently, the extant bonds now trade at 9% yield) so that option is gone. It is hard to say what downside is in this market; I think the shares are worthless but a realistic target this year is still closer to 1x revenue I feel, so >60% downside from here.

The stock is incredibly liquid and borrow is readily available at 2%, so this is actionable for most investors. This writeup will focus on three main points (yes my wife is a management consultant):

– NIO is a bad ‘business’

– Cash burn rate is so intense it trumps everything else

– Accounting oddities/other yellow flags


Nio is structurally a bad ‘business’:

Perhaps lost in the IPO hype (until recently) were some of the more glaring structural impediments to NIO’s long-term success. What I mean by that is, NIO looks and feels like an automotive OEM but actually it is a different kind of business model: they appear to take on half the characteristics of a capital-intensive, operationally-geared auto company, but at the same time they defer to outsource key parts of the business. For example, they purchase all their own raw materials, they build their own tooling; and they design the vehicles – but they don’t actually manufacture their own cars. Rather, JAC, a government-owned ‘third-tier’ (sell-side assessment) OEM manufactures their vehicles for them, on a contract basis, at their own factories, using their own workers (supervised by a much smaller number of NIO managers).

This is a fairly strange arrangement, to put it mildly. I am not sure anyone noticed this in the IPO docs, but NIO agreed to reimburse JAC for any operating losses suffered in the first three years of operation – opening NIO up to potentially being on the hook for an escalating burden if demand doesn’t scale with JAC’s production capacity – and perhaps more importantly, they have agreed to pay JAC a fixed, per-vehicle fee to produce each car. They don’t disclose what this fee is, but my rough estimate is somewhere around 10-15k CNY per car – which, when you consider the ASP of their one model launched so far (the ES8) is only 430k, constitutes perhaps 3% of ASP and so a very significant headwind against scaling gross margins.

Thus, for all intensive purposes they take on many of the downside characteristics of a capital-heavy business – they still carry raw materials, tooling and inventory; they still have operational exposure to low utilization of their partner’s production lines – but without as much of the upside (due to the fixed per-vehicle fee). Perhaps this is why the business, small scale though it is, exhibited less than expected operating leverage at the gross margin line in the most recent quarter.

There are other aspects of NIO’s offering that appear troubling. Clearly they operate in an (already) very crowded market, and they have no infrastructure/charging advantange. According to NIO’s own IPO docs, there are already 15 different BEV cars on the market in China today, driven (until recently) by generous subsidies and forcing NIO into a small niche of an already small market – this number is likely to double in the next year or so as new models continually enter the market. I am bearish on TSLA, but if you were to cite perhaps the two strongest elements of the bull case for TSLA you would suggest the lead they have in the US on other BEVs (essentially no competition in the mass market, for now); and also its market-leading supercharger infrastructure network. NIO, however, very clearly has no proprietary charging network (they make it clear they rely on public infrastructure), and instead have invested in battery swapping points in strategic locations to enable drivers to conduct long drives from, say, Shenzhen to Shanghai. Upon closer examination, however, the prospectus discloses the total cumulative investment in battery swap locations and facilities was only $10.3mm as of June 2018 – hardly suggesting a sustainable competitive advantage over time. Perhaps proving this point, services revenue – of which battery swapping is a part – has remained de minimis even as deliveries have ‘ramped’ over the last couple of quarters.

OK, so sub-optimal production model, no moat and plenty of competition. What, then, has NIO actually spent its money on? Well, other than two arguably cool-looking SUVs, they have progressively hired swathes of engineers, all over the world – despite a purely China focus both today and in the future. The company has discrete ‘R&D centres’ in Silicon Valley, the UK, Munich, and China – four different R&D locations despite no overseas businesses or ambitions, and having sold a total of only ~10k cars or so (as of Dec’18). These overseas locations, by the way, are no one or two-man outposts: Silicon Valley has 560 employees (8% of company total at time of IPO) across two offices; Munich has 170 (90% of whom are engineers); and the UK has over 33 (at least spread over two offices, London and Oxford). I am not really sure why a domestic Chinese company with no brand nor ambitions beyond China needs to spend oodles of money to hire expensive engineers all over the world – it seems superfluous and unnecessary at best and there is a nefarious explanation at worst. . Otherwise, they are spending a lot of $$ on brand building, for example by developing an app to market the ‘NIO lifestyle’ and have opened a number of ‘NIO houses’, basically a combination of showroom/health club/daycare centre in Chinese cities to credentialize the aspirational nature of the brand. So far all this has generated a grand total of 11k deliveries and maybe another 15k in reservations…so colour me skeptical.

In any case, its worth keeping in mind just how much they are spending in absolute dollars: in the year leading up to the IPO, NIO spent 6.1bn CNY – around $900mm – in total opex, of which basically half was R&D. By way of contrast, when TSLA IPO’d – at a similar stage of development with zero automotive revenues – they spent $50mm in R&D that year and ~$65mm in total opex, or 1/13th the number. Hmm…note also that >90% of these expenses are cash expenses – perhaps not surprisingly, only a small % of compensation has come via stock options/RSUs, and so there are real cash consequences to all these hires – again, this is a bit unusual for a venture-stage company, in that normally equity/non-cash comp is (in my experience) a good deal higher than 5-10% of total opex. Maybe all those new hires don’t want company stock because they don’t think it’s worth anything?

This is probably the right time to mention that this is a Chinese JOBS Act IPO, listed in the US, not China (#thechinahustle) with an admitted uncured failure in internal controls – so alarm bells should be ringing. But let’s come back to that point later. The main point for now is that a huge amount of R&D and SG&A spend is on salaries, is cash-based, and, therefore, very close to fixed costs (per the prospectus, 40% of R&D and SG&A, respectively, were for salaries). They have continued to hire with abandon, as recent disclosures suggest the employee count is up to 9k (versus 7k at the time of the IPO, last September). Even if the topline eventually grows, these fixed costs necessarily limitt the ability to scale profitably such that I think they would need to sustainably sell 100k units/yr to generate operating profits…

It’s worth mentioning at this stage the most recent regulatory changes (announced yesterday). Basically, the Chinese government cut subsidies on EV purchases for cars with a range >250km (ie, including NIO’s two models) by 50%, after a transition period; including the elimination of some local subsidies, the total subsidy amount will decline say 65%, increasing the cost of the average car by around $5k lets say (against NIO’s ES8’s current ASP of $67k, so this is pretty meaningful). This is bad, of course, but its not as if NIO was selling many cars before this cut so while I don’t think it helps I also don’t think the thesis rests on the macro here at all. Instead, this thesis is all about cash burn…


NIO is incinerating cash at an amazing rate:

The main reason I like this trade – and even if you think they have a shot at building a viable brand, or like the car, or whatever – is because even if that happens eventually, they will need to raise so much $$ in the meantime that supply/demand in the stock alone is likely to send the price lower from here. Consider the following:

– NIO has burnt >$5bn in 4+ yrs of operation, ie >$1bn/yr on average;

– they burnt $1bn (my estimates) in its first two quarters as a public company, just after raising $1bn in the IPO – cash burn is increasing because of the aforementioned big increase in hiring and cash nature of those expenses;

– the accelerated burn rate forced them to raise ANOTHER $700mm in convert debt before the IPO lock-up expired (with no new disclosures as it was an overnight deal), that was promptly followed by a guidance cut;

– Q1/Q2 delivery guidance implies at least $500mm/qtr in further cash burn and gross cash today (allowing for the convert but before 1Q cash burn) is maybe $1.9bn.

In other words, assuming nothing changes, by June 30th – we are already half-way there – gross cash should be below $1bn, ASSUMING no deterioration in working capital (which has blown out to negative $500mm due to a big spike in payables and accruals in the most recent quarter) which seems at least partially likely given the elevated levels in Q4 and the guided big sequential decline in revenues. In reality we should see at least some normalization of working capital which would imply gross cash well under $1bn – ie less than 2 quarters of runway from there – and, with subsidies going lower in 2H, no guarantee that orders/deliveries/demand will ramp.

Since they still owe the Chinese banks 3bn CNY, or $450mm, I think the minimum cash balance they would allow would be at least $500m (given how Chinese banks like to be collateralized by cash, cf TSLA), meaning realistically NIO may be operating with barely 1 quarter’s worth of cash, run-rate, as soon as mid-summer. Clearly that would precipitate a hugely dilutive offering of – I think – at least $750mm-$1bn, basically at any price the market would accept (in my view, much lower than $5). This is likely to be in the context of souring EV-related demand more generally (as I think TSLA well and truly cracks by then) and since NIO will very much be a price taker, I cannot see how they can pull the same trick twice (essentially using the hype to raise capital at the wrong price) since the convert deal bombed so very badly. We have seen how the stock took the advent of IPO lock-up related selling recently; how do you think it will react to a price-insensitive further $1bn in stock being pushed into the market?


Other accounting yellow flags:

I don’t want to spend too much time on the conspiracy theory (though I have alluded to it slightly throughout the writeup). Suffice to say you also have the following going for you on the short side:

– this is a JOBS act IPO;

– the IPO disclosed material weakness in internal controls in both 2016 and 2017, that may or may not have been fixed (not mandatory because of JOBS act IPO provisions);

– this was a Chinese co IPO’ing in the US, raising new money from offshore, mostly US/European institutions and not domestic onshore Chinese investors;

– the only non-Chinese executive, Padsmaree Warrior, quit the board shortly after the IPO;

– the founder, Bin Li, has no experience in automotive manufacturing or indeed any manufacturing (his previous venture, Bitauto, is an ecommerce business);

– the $ value of PP&E, considering that the company does not have its own factory, seems very high ($550mm) and there is minimal disclosure around what is capitalized here versus expensed;

– ditto for the R&D expenses (other than salaries), there is minimal disclosure and no idea why the $$ amounts are so high relative to other start-up auto companies;

– the co shoves a lot of standard payables into ‘Accrued and other Liabilities’ (things like payables for R&D, salaries, accrued expenses);

– standard onshore/offshore VIE treatment for ADR owners (essentially you own nothing of the onshore assets), so if anything goes wrong, you’re more screwed than usual (par for the course with Chinese ADRs).

In other words, I am not saying this is another China fraud, but the conditions clearly exist for it to be a fraudulent enterprise, defrauding offshore investors to the benefit of Chinese entrepreneurs. I don’t think you need it to be a fraud to win, however.

Disclosure: short NIO

Stranded on the Westshore

Westshore Terminals (WTE, listed in Toronto), is by all appearances an amazing business. Located just south of Vancouver, WTE operates one of the largest coal export terminals in the world, loading around 30 million tons (mt) per year while operating almost continuously, 24 hours a day, 365 days a year, either from trains arriving directly at the terminal or from storage facilities that WTE maintains onsite. Until recently, it has been the only coal terminal of scale capable of serving Alberta’s export-oriented metallurgical (met) coal mines, giving it significant pricing power over key producers in the region; it also handles coal for export railed to the coast from the Powder River Basin (Wyoming/Montana in the US). WTE operates the port on a throughput basis – getting paid a fixed, and often escalating, fee per ton loaded – and never takes ownership of the coal – thereby taking no coal price risk and making the business ‘capital-light’. Meanwhile, underlying capex requirements are low, since invested equipment has a long usable life and WTE leases the right to use the port from the local government. These features have allowed the business to historically generate >50% EBITDA margins, convert >65% of EBITDA to free cash flow, and generate sustainable returns on invested capital (ROIC) of >20% across the cycle. Moreover, the company distributes most all its free cash to shareholders in the form of dividends and buybacks, because it has no debt.

Sounds great, right? Well, despite all of the above, I think WTE is an incredible short opportunity today. Oftentimes the best shorts occur not when a middling or poor business inflects lower (although I like those too ;)), but when a heretofore excellent business undergoes a fundamental change. Just such a change is coming to WTE, and will be inexorably reflected in the company’s earnings power in the next couple of years. I believe the sleepy, dividend-oriented Canadian ownership currently in the stock simply hasn’t cottoned on yet – they have been clipping their 3-4% yield consistently now for a decade – and that creates the opportunity we have today.

In a word, the problem is customer concentration (well, two words, I guess). WTE relies on one customer, Teck Resources (TECK, listed both in Toronto and New York), for >60% of its volumes; and another, Cloud Peak Energy (CPE), for ~15% of its volumes. This would be concerning for most businesses – as customer concentration increases the volatility of any company’s earnings stream – but in this particular case it is downright alarming. That’s because CPE is about to go bankrupt; and shortly thereafter, TECK’s loading contract – responsible for so much of the business’ stability over the last decade – expires in March 2021. Both issues are likely to prove huge negatives for WTE’s future prospects. Let’s look at the CPE impact first.

Impact of Cloud Peak’s impending Bankrutpcy

Cloud Peak is a distressed Powder River Basin (PRB) thermal coal producer. Thermal coal – used to produce electricity, as opposed to met coal, which is used to make steel – has been progressively losing share to nat gas and renewables as a generation fuel, both because of the environmental imperatives of many developed countries but also because of the increase cost competitiveness of gas-fired plants. Without going too much into the weeds of how CPE came to be in the situation it is in (see here for more background), suffice to say, the recent 10-K made it very clear CPE that is on its last legs as a public company and will need to be restructured shortly.

The real question however is not the ultimate filing, but the form the company takes during/after restructuring. Last year CPE produced 50mt of coal, from three different mines (Antelope, Cordero, and Spring Creek) – but they only shipped 4.6mt through WTE (all from Spring Creek). This shipment number is meant to escalate to as much as 10.5mt by 2022 – meaning as soon as in a couple of years, this mine alone was meant to account for ~30% of WTE’s then-notional capacity! (35mt per company guidance), As far as WTE is concerned, it is really only what happens to Spring Creek (the one Montana mine with coal of quality/type suitable for Asian utility customers) post B/K that matters, but I believe it may be quite difficult to ‘extract’ the Spring Creek operations from the husk of the rest of CPE, as we shall discuss.

In many bankruptcies, companies will use the process to shed themselves of uneconomic assets and restructure around a smaller, more profitable, core. Here, you could argue that CPE could perhaps shut down/curtail the Antelope and Cordero mines – which the 10-K makes clear are the main sources of the company’s current liquidity crisis and are the most structurally-unappealing assets in CPE’s stable – and simply operate Spring Creek, gradually ramping export production as that appears to be the only profitable market left for its coal.

But there are a few intractable problems in this case:

  1. Remedial obligations: coal companies are required by law to ‘fix’ the land after mining is complete (basically this means trying to restore the land to closely resemble what it was before mining began). This may indeed be futile from an environmental perspective, but it is mandatory, and more importantly for our considerations, it is expensive. As of the 10-K, the dollar value on the balance sheet of these liabilities was $93mm – but note that this is a discounted number: simply the present value of a number estimated to be paid in the far future, using a ‘credit-adjusted, risk-free rate.’ We do not know what the actual, absolute number is (it is undisclosed), but it should be much higher than $100mm. We know this because the company has $408mm of reclamation and lease bonds – a number that only required $25mm of collateral until recently (due to the perceived creditworthiness of the company) but now, as the company’s distress has become more urgent, is being requested to be topped up:

Screenshot 2019-03-22 11.07.33

Translating this into normal English: in order to keep mining, someone needs to provide more funds to third-party insurers of CPE’s remedial obligations because they are worried they will have to cover these costs if the company is not around to pay the tab (which could be much sooner than the nominal long mine lives left, if the company folds and no-one else wants to mine coal in the PRB anymore). This is all a long way of saying that it is not simply a matter of shutting down the ‘bad’ assets: someone still needs to pay for the remedial work, and this number is in the hundreds of millions. Moreover, since the company’s mining permits depend on reclamation liabilities being funded, there is a reasonable chance that Spring Creek’s continual operation rests on the willingness of someone to fund the clean-up costs at the other two mines. Inevitably this affects both the price paid and the economics of any ‘good’ assets that remain in operation (since they would have to presumably fund the cleanupfor anyone willing to buy them). That alone has dire consequences for WTE…

2. Spring Creek economics: another big problem is the underlying earnings power of CPE’s export business as it stands today. CPE reported the following in their 10-K:

Screenshot 2019-03-22 11.14.04

(‘logistics business’ refers mostly to tons for export). During 4Q, the Kalimantan (Indonesian) benchmark index for this type of coal declined aggressively because EM currencies got killed and also because Indonesia stopped a ban on coal exports (which had been supporting prices). Elsewhere in the 10-K, CPE discloses that its cost of production on a per ton basis was $54/t last year – hence the uneconomic nature of these exports even close to $50/t. While prices have since recovered to the high-50s, it’s quite clear that even CPE’s ‘good’ export business remains priced out of the global market with its current cost structure when Kalimantan prices are below the mid-$50s/t. This conclusion is also problematic for WTE, as transportation/loading costs constituted 70% of delivered cost to customers in 2018 ($183mm out of total export coal opex of $266mm). Because logistics costs are such a massive contributor to all-in costs, the solution to anyone in bankruptcy is simple: use the process to recut all the logistics contracts (indeed, this is primarily what B/K is designed for, to allow distressed businesses to exit onerous contracts).

3. CPE is burning huge amounts of cash: this point may seem somewhat facetious but I think it’s important to emphasize that CPE is not entering B/K with stable operations. They disclosed $93mm of cash on hand at Dec’18, but have no sources of liquidity beyond that (since they canceled their Credit Agreement as they were going to bust covenants). At the same time they disclosed they only have $65mm of cash on hand as of early March:

Screenshot 2019-03-22 11.26.12

They thus appear to be burning >$15mm PER MONTH and so if/when they enter bankruptcy will require significant and immediate a cash injection to maintain operations. They may well negotiate some kind of DIP financing – despite the issues mentioned – but I think it is worth highlighting that the cash needs alone could dictate much lower near-term (or mid-term) production as they simply don’t have the liquidity needed to support operations.

What does it mean for WTE?

To summarize all of the above: CPE has unavoidable, large, remedial obligations and continues to rapidly burn cash. Even in the scenario where the export-oriented mines continue producing (and this is not a given), the economics of that operation need to be significantly improved to maintain that mine’s ability to produce throughout the cycle. Since WTE’s terminal loading fee is one of the most significant cost items contributing to CPE’s export coal delivered cost, it appears inevitable that the estate will use the bankruptcy process to renegotiate WTE’s loading fees significantly lower. To me, this represents the best possible outcome, as there remains a very real chance CPE assets exit the market either partially or completely.

But how much does WTE actually make off CPE coal right now? That number is not disclosed, but in 2018 WTE generated CAD $11.7/t of coal handling revenue on average across all its customers. The CPE agreement has been renegotiated a few times, but I think CPE pays somewhat less than the average fee (mostly because TECK, WTE’s main customer, is locked into an above-market contract, more on that later). I estimate CPE is paying ~$11/t, or $51mm CAD (perhaps a little more with commitment fees, etc), which – using 77c USD/CAD avg in 2018 – implies ~$40mm USD, or around 22% of total transportation costs (the rest being the cost to rail the coal from minehead to port) and 15% of total opex for the export business. Note that total transportation costs for the exported coal last year were $40/t on a per ton basis (including both rail cost and WTE’s loading fees).

So, how much would cash costs have to come down to satisfy new owners, and how much of that burden would WTE have to share? While this is a difficult question to answer precisely (made more difficult without knowing what form the new cap structure would take and indeed if CPE could even garner new capital), consider the following points:

  • even well-capitalized, low-cost, through-the-cycle coal producers (Warrior Met Coal, Peabody, Coronado Resources, etc) trade with extremely low multiples (3-4x EV/EBITDA or lower) and implied high costs of capital (low to mid-teens %). Anyone potentially bankrolling a restructuring high-cost 100% thermal producer would of course demand an even higher cost of capital as well as a huge improvement in pro-forma operational returns;
  • there are limited direct comparables, but Consol Energy owns and operates a smaller coal handling terminal in Baltimore that charges $5 (USD)/t all-in loading fee – significantly less than CPE is paying today for similar services in a similar market;
  • discussions with Australian coal-handling infrastructure providers suggests they charge at least 30% cheaper than the rates WTE is earning today (this is, in essence, one of the reasons TECK is so unhappy).

None of this provides a straight answer, but pulling it all together, if we applied, say, a blanket 30% cut to all logistics costs (meaning the railway company shares the pain), then on a per ton basis this would fall from $40/t to ~$27/t; all-in costs (assuming no other efficiencies on the non-logistics costs) would fall from $57/t to ~$45/t; and since depreciation is ~10% of costs, this would imply a low-$40s cash cost number (versus mid-$50s cash cost number today). In other words even if Kalimantan coal went back to the high 40s (where it was in December), the mine should still be generating 15-20% EBITDA margins – not too far below what a somewhat-comparable PRB producer like Peabody (BTU) is earning today. Something like this may be an acceptable business to consider recapitalizing with new money (relative to simply investing in an existing asset like Peabody or Coronado).

In this scenario, WTE’s take would be haircut, clearly, by 30% – thus falling from say CAD $11/t to $7.7/t – which sounds fairly drastic but is still a near 20% premium to what the East coast terminal owned by Consol is charging (in USD terms); and indeed a similar rate to what WTE used to charge back in 2011 (before many years of price escalation):

Screenshot 2019-03-22 12.16.11

Such a move would immediately remove ~$15mm CAD from WTE’s EBITDA, if volumes remained flat at 4.6mt rates – this is only 7% of FY18 EBITDA, so not too bad, right? Maybe not…but since we know CPE is trying to get to 10.5mt over the next few years, the implications are actually disastrous if applied across the pro-forma larger loading volume – almost 1/3 of WTE’s loading tonnage in 2021-22 – because effectively they will be displacing TECK volumes rolling off at much higher rates, and fixed costs are very high, as we shall see.

We will come back to the all-in economics, but let’s turn to the second issue, TECK, now.

Teck Resources is going it alone

In 2011, Teck Resources (TECK), a Canadian mining conglomerate and the second-largest met coal player in the seaborne market, signed a binding, take-or-pay, 10yr contract with WTE the load at least 19mt/yr of met coal. This agreement expires in March 2021. TECK’s annual met coal production is unlikely to change much in the medium term (around 27-28mt/yr). TECK has already notified WTE that it intends to ship less than 19mt once the contract expires (this was disclosed in WTE’s recent annual report), and indeed TECK has been quite vocal about its alternate plans: it has prioritized increased the capacity of its own fully-owned loading facility, the Neptune Terminal, located in WTE’s backyard of Vancouver harbour, a mere 40km away:

Screenshot 2019-03-22 18.59.45.png

TECK has committed to spend the capital necessary to get capacity at Neptune from its current 12mt/yr up to at least 18.5mt/yr by 2021 (fully permitted for this amount already), and maybe to as much as 20mt/yr. At the moment, Neptune only handles 6mt/yr – largely because TECK is locked into the WTE contract for now – but once the expansion is complete, there is simply no reason to think TECK wouldn’t maximize handling at its own, fully-owned and operated facility, Neptune. Indeed, TECK management has been quite vocal on numerous investor calls (see the most recent call, for example) regarding the rationale to make the investment ($85mm+) to expand Neptune. It has been driven both by a desire to control their own infrastructure – and thus maximize flex deliveries around surge pricing, and also because, frankly, WTE has underperformed their expectations (eg by commingling high value met coal with other clients’ thermal coal; or prioritizing the shipment of other clients’ lower-value thermal coal instead of TECK’s high value met coal during peaks in the market). Given the volatility in global coal prices, these operational missteps, even if of a matter of days delay in delivering to customers, can be hugely impactful to the PnL of the business. Another important reason motivating the move is simply that TECK feels WTE is charging them too much (much higher than equivalent Australian or East Coast ports would, for example), and thus has gained far too much value out of TECK over the 10-year contract.

Thus, post March 2021, at a bare minimum TECK will rededicate as much WTE throughput towards Neptune as Neptune can handle. Assuming TECK can successively execute on the capex over the next 1.5yrs (and they have already started the project), this to me would mean at least 15mt/yr going towards Neptune (at 80% utilization) – meaning an incremental 9mt/yr on top of the 6mt/yr already going there. This would clearly come out of WTE’s volumes, reducing TECK volumes at WTE from 19mt down to 10mt/yr on an annualized basis. And once Neptune is built, these volumes will never come back (thus permanently raising the risk profile, and lowering the multiple, of WTE as an asset). Thus, even just this highly plausible scenario could see WTE lose 1/3 of its annual volumes, for good, in under 2 years, and permanent raise the cost of capital for WTE as a whole.

High fixed costs are a double-edged sword

Part of the reason WTE has been such a good earner until now is also the reason it could all fall apart in a hurry in the near future – namely, the cost structure of the business.

Screenshot 2019-03-22 19.21.10

Looking at the above, it’s quite clear most all the costs are fixed, not variable:

  • administrative costs: clearly mostly if not entirely fixed;
  • depreciation: entirely fixed;
  • terminal leases & fees: there is a fixed portion payable no matter what volumes are (about 60% of the total); then the balance varies depending on volumes about 17.6mt/yr. So still, mostly fixed;
  • salaries & wages: we don’t know precisely how much is fixed or variable, but evidence suggests this too is mostly fixed. For example, look at a chart of tonnage loaded versus total salaries & wages over the last eight years:

Screenshot 2019-03-22 19.28.36

Annual tonnage has moved around a fair bit, but salaries and wages have only kept going up (the jump after 2016 was due to a new union contract), and in general show zero sensitivity to annual declines in volume. In my subsequent calculations I am estimating 70% of salaries are fixed, though this should be conservative based on the above.

The obvious corollary to all this is simply that lost volumes will have a much bigger downside impact on WTE’s financials than even the percentage decline in volumes would suggest. Hence if or when TECK moves its volumes away, WTE could be losing ~33% of its volumes (10mt out of 30mt) but – as we shall see – a much, much larger % of its earnings power.

Putting it all together – EBITDA ~50% lower post TECK expiry, CPE restructuring

Let’s take a look now at what the post-2021 WTE business could look like, adjusting for what I think happens at CPE and TECK. A few notes on my base case assumptions:

  • TECK: as discussed above, I contemplate only 10mt of the 19mt currently at WTE, moving to Neptune from Mar21; the remaining 10mt I estimate will be restruck at current WTE ‘average’ rates (CAD 11.5/t);
  • CPE: I’m giving substantial credit for the planned ramp in volumes, estimating 9mt/yr from the current 4.6mt/yr rate (versus CPE’s goal to get to 10.5mt by 2022). In return for this/to make this economic through the cycle, as discussed, I envisage a cut in the loading rate to CAD 7/t (ie slightly more than the 30% cut I discussed conceptually earlier);
  • Others: no changes in tonnage nor estimated rates on the rest of WTE’s customers (this is also actually bullish as a number of these customers are also swing thermal coal producers)
  • Costs: 70% fixed costs share for salaries/wages (this could be generous); and depreciation rises to maintenance capex levels (20mm CAD/yr).


Screenshot 2019-03-22 20.21.52

Screenshot 2019-03-22 20.22.26

All in all, and because of the very high fixed cost nature of the assets, the decremental margins on lower volumes (throughput in this scenario falls to 25mt/yr from 30.5mt/yr in 2018) are very high. As a result, EBITDA falls from ~195mm to 99mm (-49%) and the business is doing barely 50mm of FCF sustainably in this base case. Even allowing for the unlevered balance sheet and some continued reduction in the share count between now and then, this implies a near 25x P/FCF multiple for a somewhat stranded asset post TECK’s pivot and CPE’s restructuring, far more exposed to lower-quality, swing producers and thus deserving of a lower rating from the market. I think a fairer price would be somewhere between 10-15x FCF in this scenario – implying at least 40% downside in my base case – and there are clearly many scenarios where earnings fall a lot farther than this (TECK moves more volumes away; renogotiated rates on the balance are lower than 11.5/t; or CPE goes into liquidation/production decline post B/K).

What I like best about this trade, though, is that the sell-side is so asleep at the wheel. Most analysts don’t seem to be considering what even happens beyond next year, and if they do, they assume the gentle escalators WTE has enjoyed in recent years simply continue – they seemingly cannot countenance what is right in front of them (probably because it is still 2yrs away). But given the importance of the TECK contract to WTE, this could really be renegotiated/announced in some form much sooner than March 2021 (thus making clear the fundamentally lower earnings power of the go-forward business), and is why this is such a compelling short in the near-term. That, and also you get full exposure to the consequences of the impending CPE bankruptcy filing.

Disclosure: short WTE CN

Tesla unsecured debt: recovery tops out at ~30c (if you’re lucky)

I’m dusting off the cleats (yes, it’s been a while), as I’ve had a large number of incoming questions regarding what happens to Tesla (TSLA) unsecured debt if or when they file for bankruptcy. This will not be a post about the long list of existential risks facing TSLA at the moment: there will be no commentary about how fundamentally flawed the business is; how they play fast and (very) loose with accounting rules, securities laws and their customers’ lives; how they’ve seen more executive turnover than the first class lounge at London Heathrow; nor how they’ve been running the business on the fumes of fumes for at least the last couple of quarters. For more on any/all of these topics, and a lot more, feel free to mosey on over to Twitter and follow $TSLAQ; or read some excellent blog posts here or here, for example.

Rather, since I consider a TSLA balance sheet restructuring basically a fait accompli at this point, to me the more interesting question than ‘where does the equity go?’ (answer: zero) is what happens to the unsecured creditors. As of the Dec’18 10-K, there was about $5.8bn of unsecured debt with full recourse to the TSLA balance sheet, and today the cheapest of these unsecured bonds trades around 88c on the dollar (the 5.3% ’23 long bond); note also that this category of debt includes the converts trading above par, due to convertible option value even though the strike price is still well out of the money.

I don’t want to spoil the lede, and I should disclose at the outset that I am (significantly) short TSLA stock in various forms (cash equity as well as via put options). But in terms of risk/reward, shorting the bonds may be even more attractive than shorting the common, because – as we shall discuss – I think ultimate TSLA recovery on unsecured debt will be very low (max 33c, likely much lower); you can generally get more leverage on credit shorts than equity shorts (for institutional clients, at least, via CDS); and the payoff on even outright bond shorts is much more asymmetric than shorting the stock (even if TSLA doesn’t file, it is most unlikely TSLA credit improves much in the next 1+ years so the likelihood of a large loss as TSLA credit reprices massively tighter is in my view remote).

Additionally, while not too complicated, TSLA’s cap structure is not exactly clean – so I hope by unpacking it here to demonstrate at least some of the thought process around credit analysis (in particular recovery analysis) and how it can be helpful in guiding not just bond but also equity investment decisions. With that out of the way, let’s dive in…

Step 1) Start with the Balance Sheet

Pretty simple really – any analysis of what creditors (of any stripe) will get depends on what the assets are, right? (Actually its more complicated than this, but let’s work with this standing assumption for now).

As of Dec-18, TSLA’s assets looked like this (the left side), and then on the right, adjusted ONLY for the repaid convert ($920mm that matured Mar-1). We will make other adjustments as we go, but here’s what it looks like for now:

Screenshot 2019-03-04 21.12.19

OK, so ~$30bn in gross assets and ~$29bn adjusted for the convert repayment. Sounds like a lot, doesn’t it? Hmm…hold that thought 🙂

2) Isolate assets associated with leases, as well as collateralized against non-recourse loans

Of course, a large portion of TSLA’s assets have been financed by leases – clearly these assets are ‘spoken for’ and unsecured creditors will have no recourse to them (similarly, lessors have no recourse beyond them). At the same time, TSLA has incurred a large amount of so-called ‘non-recourse debt’ (a lot of it related to the SolarCity takeover), which is simply a fancy way of saying certain assets on its consolidated balance sheet have been financed by loans secured against only those assets. Since we want to figure out what’s left for the guys at the bottom of the totem pole, we need to remove all this stuff.

This is where it gets a little tricky. TSLA’s disclosures are not perfect (!) and I have had to make some simplifying assumptions. You can dig through the devilish details in the below table, or you can just rely on the summary of how I have treated most of these obligations:

  • short-term and long-term restricted cash are in effect completely consumed by some combination of short-term and long-term build-to-suit (BTS) leases, as well as short-term and long-term residual value guarantees;
  • SolarCity related asset-backed securities are collateralized at a 70% average LTV (this is within the range of 65-80% disclosed, occasionally, for some of the transactions, but not all of them);
  • The Automotive Asset Backed Notes have only a small amount of overcollateralization (I assume LTV 90%) – this is based on some isolated ABN documentation I have seen suggesting around a 10% O/C cushion.

The below chart summarizes the treatment of all the various leases and non-recourse lending that consumes a good portion of TSLA’s assets, already. Apologies if it’s a little confusing/messy, but the overall takeaway is TSLA has already pledged $5.1bn of consolidated assets against its non-recourse obligations and lease liabilities:

Screenshot 2019-03-04 21.25.31

Double-checking this number, page 123 of the 10-K suggests the co has pledged $5.2bn of assets thus far, so I think we are on the right track:

Screenshot 2019-03-04 21.26.38

3) Isolate assets associated with secured (recourse) debt

While TSLA has a ton of secured, non-recourse debt, it doesn’t have much secured recourse debt. This makes this part relatively straightforward. There are really only two obligations, one very significant (the Credit Agreement backed by inventories, receivables, and some PP&E, known in the business as an asset-backed loan, or ABL), and one small other obligation.

Again, the precise over-collateralization on the ABL is not public (since the loan docs disclose 85% of ‘eligible’ asset classes ‘less reserves’ is the borrowing base, without making it clear what those reserves are, for example, but given the quality of those underlying assets – inventory, mostly, recovery of which even in clean bankruptcies is still really low, 50% or lower – I think estimating a 70% LTV on the outstanding balance is reasonable if not generous. Hence:

Screenshot 2019-03-04 21.34.02

4) Remove all collateralized assets from existing gross asset side of Balance Sheet

Now we need to put it all together. We simply subtract – from the relevant asset categories, based upon 10-K disclosures as well as reasonable judgement – the collateralized (‘spoken for’) assets, from the gross totals presented in the consolidated accounts. See below for my attempt at this, with some clarifying comments. I have included the original Dec’18 asset side of the balance sheet for ease of comparison. The key number to focus on – at the bottom in yellow – is that post this process of elimination, TSLA’s apparently generous $30bn in gross assets is really only $18bn once existing secured claimants of various types have been removed:

Screenshot 2019-03-04 21.40.52

5) Haircutting remaining assets for value in recovery

OK, we are about half way there. Next thing to do is to think through what values the remaining assets may have to creditors. This is the part of the analysis that becomes much more subjective and variable depending on the ultimate form of restructuring (is it a pre-pack filing? Will the business continue as a going concern or not? Will it be liquidated or not? etc). I have personal views as to how this restructuring is likely to play out (to be discussed later), but for the moment I want to emphasize that even if the ultimate form of restructuring is, say, more favorable to TSLA creditors, they will all necessarily apply some conservatism (read: haircuts) to expected asset values in and around any bankruptcy/restructuring process, and by definition unsecured bonds should trade down to/or below (to allow for a risky return to market participants) levels implied by this inherent conservatism. So, the analysis is important and valid, even if by nature imprecise.

That said, you can see below the haircuts I have made to various categories. Some of these haircuts are fairly obvious (intangible assets aren’t worth anything to creditors; recovery rates on inventories and captive non-land PP&E (for example machinery and leasehold improvements would in principle be very low). NOTE ALSO that I have taken a further $1bn out of existing cash – this is NOT because I think the Dec-18 reported cash number is bogus (although I do), but is simply my estimate for the gross cash burn likely to be reported in 1Q (mostly operational, and through restructuring cost, not W/C build), so I think this is a defensible adjustment and may end up being conservative.

As you can see, even the $18bn gross available to unsecured creditors could end up being more like $9.5bn, or less, post recovery impairments to stated asset values on the last balance sheet:

Screenshot 2019-03-04 21.52.55

6) Estimating unsecured claims…

OK so we’ve figured out the asset side (‘assets available to unsecured creditors’), but what about the creditor claims themselves?

Again, this is an area where the form of the restructuring could have large implications for ultimate recovery. For example, in any kind of ongoing-business situation, trade claimants (accounts payable, accrued expenses, a few others) would effectively be treated senior to all other claimants simply by the fact that unless they get paid the company will go into liquidation. In the below I do NOT assume that is the case – in other words my base case assumption is all unsecured claimants get treated equally, after administrative claims, as would be the case in a liquidation. As we shall see, this is likely an assumption that FAVORS non-trade unsecured creditors in this case (ie, bondholders), but in any case I will expand on my reasoning for this in a little more length in a moment.

For now, we simply need to tot up all the remaining balance sheet liabilities that are neither leases, nor non-recourse debt, nor secured recourse debt. Then – and this is very important – we need to estimate the off balance sheet liabilities that become unsecured claimants post filing. Again, another area of huge judgement required (estimating the liability related to all the shareholder lawsuits re ‘Funding Secured’, for example), but once again, I have tried to be as ‘pro-creditor’ as possible in my approach. Even so, the following large items make it onto the unsecured claimants list:

  • trade claims (about $6bn);
  • customer deposits ($0.8bn) – yes this is simply an unsecured claim (gulp);
  • unsecured debt ($5.8bn left post recent convert repayment);
  • Panasonic purchase obligations ($13.2bn under ‘binding and enforceable contracts’, clearly the massive elephant in the room);
  • Other purchase obligations ($2.4bn);
  • My estimates for the NPV of the Buffalo Gigafactory breakage cost ($200mm) and legal liabilities ($500mm) – both of these numbers probably end up higher.

NOTE I have not attempted to ‘true up’ the warranty provision or make any other grand assumptions re the state of the numbers, today; rather I am simply trying to assess where the numbers naturally fall as they currently stand, and see what that says about recovery value before anything (else) goes wrong.

As a result, we get something like this:

Screenshot 2019-03-04 22.03.29

Hence, my conclusion that unsecured recovery tops out at 33c and in reality is likely to be much, much lower. Perhaps it’s worth expanding on why that may be the case.

If you’ve been following closely thus far, you can see that the residual assets available to the unsecured class are essentially cash, then SolarCity assets and then PP&E. First, the easy part: even if we thought the $3.7bn cash balance (pre convert repayment) was ever real, the chances of that being there when or if they file (given how Musk is barrelling the company right over the cliff) is, let’s say, quite low. And as for the fixed assets – as a general rule, recovery values on fixed plant tend to undershoot on the downside. Simply put, hard assets like buildings and special purpose machines are not worth anywhere near as much to a new buyer as they are to an ongoing business. Finally, when it comes to the SolarCity assets, since SC was such a liberal user of asset backed financing when it was a going concern, I find it hard to believe there is much ‘real’ (read – to a third party) asset value in the un-collateralized assets today. If there were, I believe they would have already been siphoned off for additional liquidity by now…

Liquidation or DIP+Restructuring?

…Which leads us to a further point about why, to my mind, this is more likely to be a liquidation scenario than a going-concern restructuring. In order to restructure and keep the business is operation, TSLA, once it files, will need a MASSIVE debtor-in-possession (DIP) financing – maybe $3-5bn, by my estimate, simply to fund ongoing operations. This may not be impossible to conceive, but consider the following:

  • TSLA runs massively negative working capital ($1.5bn or so currently by my estimate) and has $6bn of trade claimants, today. Also, many of those creditors have given back large rebates to TSLA simply to allow them to survive until now; many of them have not been paid on time, ever. How much of that DIP goes straight out the door to keep the lights on – $2bn? $3bn? Why wouldn’t cash on delivery conditions exist for any go-forward business post-restructuring without a massive decrease in credit extended by suppliers?
  • DIPs are generally only provided when management remains in charge. I find it hard to believe Musk & co will be entrusted to run the company if/when it files (especially if there are fraud implications, which also seem highly likely). That necessarily makes it harder to get a DIP;
  • Keep in mind that Musk and co still run the company today and by the time it actually files, even if this is just in a couple of quarters, the liquidity/asset quality/brand quality of whatever is left is, judging by current events, likely to be exponentially worse and thus less salvageable (from the perspective of any potential new credit provider);
  • While not impossible it is generally harder to get a DIP in cases where there may have been malfeasance/fraud (in TSLA’s case, this could crop up in a multitude of different ways);
  • A DIP would have to prime the other secured lenders (currently the banks backing the ABL). While if these same lenders provided the DIP (as often happens) this may be doable, consider that – as we have discussed at length – there aren’t really many other assets left to the estate worth collateralizing that gives the ABL lenders reasonable security that their position is not being made worse than a liquidation (where they would most certainly be made whole);
  • A large DIP probably only accompanies a search for a buyer of part/all of TSLA. That means financing is contingent upon a third-party being willing to underwrite, say, the value of TSLA’s brand – such as it is post the price cut abomination that Musk unfurled last week – but at the same time likely means an end to the Model 3 (which essentially is what killed the company). However to my mind there are very few ways to keep the company going and viable as a brand, without continuing to service/repair/warranty the 250k+ Model 3 lemons on the road (imagine the brand damage of that approach…). But what third-party buyer is going to be willing to underwrite that open-ended, multi-year expense?

None of this is to say a DIP, and a sale or recapitalization as a going concern is impossible (hell, Enron and Worldcom both got DIPs) – just that in this particular case I think it is clearly not the base case outcome.

But the overriding takeaway is simply this: even if you think it DOES remain an operating concern and they DO get DIP financing then the unsecured recovery will most certainly be a lot lower – since something like $3-5bn of super-senior secured financing will come in ahead of the unsecureds; and that new cash likely makes its way, mostly straight out the door to other unsecured creditors (trade claims) before the bondholders (and maybe Panasonic, for that matter – yet another reason why the DIP solution may not be forthcoming). Meaning, of course, if somehow a DIP does get done, unsecured recovery will be much, much lower than the 33c best case this exercise has suggested.

Disclosure: short TSLA everything (stock, straight bonds, converts, the lot).

Bracell: a pretty juicy merger arb play

I don’t often pursue merger arbitrage (the practice of buying stocks with announced takeovers pending, and/or shorting the shares of the acquiror, hoping to capture the spread to the deal price if/when the deal closes). I am not an expert in the space, and oftentimes the returns don’t justify (to me) the real risk that deals can fall through. But every now and then there is an exceptional case, and Bracell (1768.HK) looks like one such opportunity.

Bracell is an interesting small-cap stock. While listed in HK, the company has most all its production facilities and assets in Brazil; even more interestingly, it is majority owned (85%) by an Indonesian tycoon, Sukanto Tanoto – who also controls one of Bracell’s key customers (this is not normally a good thing, but bear with me).

Bracell manufactures dissolving wood pulp (DWP), a naturally occurring fiber used in a variety of applications. The higher grades (“specialty grade pulp”) are used for cigarette filters, eyeglass frames, pharmaceutical tablets, cords in tires, and the like. The lower grades (“rayon grade pulp” or “commodity viscose”) are a feedstock for the textile industry. As you can imagine, specialty grade commands a higher price and margin, while the commodity grade is of lower value, more competitive, and possesses much lower margins.

Both commodity and specialty pulp markets are quite competitive – there are at least five or six large-ish players globally – but Bracell possesses two key structural advantages. Firstly it is fully integrated, owning all its own forestry assets (trees are the key raw material input as you might expect) – this is necessarily an advantage when it comes to input costs. Secondly, these forests are entirely located in Brazil – a double advantage both because of the recent weakness of the local currency (BRL) versus most all export markets for DWP, and also because of the nature of the trees grown there. South American eucalyptus trees grow to maturity in ~8yrs, versus the 30yrs for North American hardwood used by a number of Bracell’s competitors, meaning Bracell has a long-term cost advantage as the effective yield of a eucalyptus plantation is so much better. Thirdly Bracell has an off-take agreement with an affiliate of its major shareholder, to sell as much commodity product as it needs to to maintain full utilization of its facilities (at the moment ~25-30% of its tonnage goes towards specialty, the balance towards commodity grades). This provides a further layer of operational security in times of low utilization.

It should be no surprise then that Bracell is effectively the low-cost competitor in the space, and can profitably produce specialty grades of DWP at prices ~$200/t cheaper than its North American rivals (Rayonier, Tembec, etc). With specialty grade prices around $1200/t, this is a massive cost advantage.

This is all by way of background. The key takeaway is that within this market, Bracell possesses a number of quasi-structural advantages that look likely to persist in the medium term.

Recent events and the opportunistic bid for the company

This brings us to recent events. For much of the past year, Bracell stock drifted along unnoticed in the $0.9-1 range, despite solid operational results and high cash flows. In early June, however, the company announced 1H profit (to be fully reported in August) would post 100-150% higher year-over-year, a function of the lower BRL, price stabilization in specialty pulp, continued cost discipline and (in my view) further customer wins due to their cost advantage. The stock immediately took off, rallying from ~$1 to the $1.5-1.6 range – a level that still looked very cheap on fundamentals (as we shall discuss).

At this point the company announced that the majority shareholder had bid to acquire the remaining public float (just ~15% of the outstanding), sending the stock spiking to $2 as arbitrageurs piled in, before settling back to $1.6-1.65 once the offer price (just $1.78) became known. This is effectively a ‘take-under’, given where the stock was trading pre-price announcement.

So this is where we stand today (mid-July). The stock is currently at $1.62; there is a $1.78 bid on the table from the current 85% shareholder. The company has hired Morgan Stanley to evaluate the offer, and according to HK Takeover rules, there is a requirement for 75% of MINORITY shareholders to approve the bid, and ALSO for <10% of shareholders to veto the deal outright. While not insurmountable, these are reasonable protections for minority shareholders, in my view.

It is worth considering now how cheap this $1.78/share bid is. Last fiscal year (Dec’15), Bracell generated ~$172mm of FCF, which equates to ~39c per HK share – or a LTM FCF multiple of just 4.5x. You would be hard pressed to buy a dying, heavily levered business at <5x cash flow. Bracell is quite the opposite – it is the low-cost producer in a commoditizing industry, possesses very little leverage today (~0.8x EBITDA), and will be effectively net debt free by year end 2016. You could easily argue a 10x FCF multiple for this business is too cheap, even absent an acquisition premium.

Of course, we also know 2016 earnings numbers will be better than 2015 (due to the 1H profit guidance). In the below abbreviated CF statement, I assume a decent step-up in cash taxes, capex spend, AND no working capital benefit but still think Bracell will throw off at least another ~30c per share (or ~$133mm) in FCF:


The majority owner is thus offering to pick up the remaining shares in what could become an industry leader at 5-6x FCF with no leverage…clearly very opportunistic, to say the least

Heads I win, tails…I win more?

This brings us to the source of the opportunity. As we sit today, I estimate the following three possible outcomes:

  • minority shareholders simply don’t care that they are being expropriated – they vote the deal through. As such you receive $1.78 for the $1.62 paid today, likely in <6 months (this is not a complex deal), thus ~22% annualized return or better. Not bad at all.
  • MS suggests the bid is too cheap, and, fearing losing the shareholder vote, the majority owner raises the bid. I estimate fair value much closer to $3 than $1.78 for this asset, but even a nominal bump in deal price would lead to significant returns. For example, 7x FCF would equate to ~5.5x EBITDA, which looks cheap versus comparables but is not as egregious as the current price. That alone would increase gross returns (vs $1.62 current) to 35.8%, and, if the deal closes in <9months (again, very conservative, it should be faster than this), to 48% annualized returns. Again – not too shabby.
  • MS suggests the bid is wayyyy too cheap, and the majority shareholder drops the bid. While this seems a fairly bad outcome, I really don’t think the stock has that far to fall. Recall this business is already trading at 4.2x LTM FCF and ~5.5x current year FCF, with no net debt from end-2016. It also pays a ~2.5% dividend which will surely be raised if the takeover doesn’t go through, AND it was trading in the $1.5-1.6 range BEFORE the takeover bid was announced. Finally, since the majority owner owns 85% of the issue, there simply isn’t much stock for arbs to dump if the deal breaks (frankly I think most of them bailed once the deal price was announced). Adding it all up, I don’t think the stock price falls that much if the deal breaks (5% or so?) and if it did I would buy more.

We must also consider the motives of the major shareholder. As of today, the entire market cap is $715mm (USD), and the EV is ~$850mm. At deal price, the implied EV is $923mm (looking at LTM numbers), or $785mm looking out to year 2016 end (due to the FCF generation). Of this EV the major shareholder already has $668mm invested (his 85% stake at $1.78) thus leaving only ~$118mm of ‘value’ being paid out to minorities. Increasing his bid by say 25% thus only amounts to an incremental ~$30mm being paid out, or just ~3.8% of EV at current deal price and <5% of the value of his stake at deal price. This does not seem like a large incremental amount to pay to achieve closure for this valuable asset, in my view.

Really the main limiting factor here is liquidity (the stock trades ~$200-300k USD per day, or less), and then timing (it is as yet unclear when the vote will be, though I anticipate it will be ❤ months). However for smaller funds/private investors, the risk/reward looks compelling.

Disclosure: long 1768.HK