Michael Mauboussin's 2014 paper on Calculating ROIC is an excellent guide to just about everything you need to know about this important metric and how to make sense of it. Here are my summary notes from reading it:
ROIC is a key indicator of a company's ability to create value, its competitive advantages and capital efficiency. What's important is: 1) the absolute spread between the ROIC and the WACC, and 2) how much capital can be deployed at that spread (reinvestment rate)
ROIC = NOPAT / Invested Capital (IC)
where NOPAT = EBIT - Cash taxes (ie. cash earnings before financing)
IC can be thought of in two ways: 1) the net assets a company needs to operate 2) the amount of financing a company's capital holders need to supply to fund the net assets
There are a few practical issues however.
Firstly - excess cash.
He argues you should exclude excess cash and only consider the cash a company needs to operate. This ensures separation between the operating efficiency of the biz and capital allocation decisions (buybacks etc)
I think this makes sense in theory however in practice it becomes quite subjective - what you consider to be excess cash? Think about all the large tech companies and the big cash piles they are sitting on. He provides some guidelines below:
Next is goodwill. If there's been M&A then the goodwill on the BS may have a large impact on IC & ROIC as shown in the Cisco example below. If you don’t expect many large deals going fwd, then the operating ROIC (ex-goodwill) may provide a better picture of incremental returns
If you expect the company to be acquisitive, then the fully loaded ROIC incl goodwill may be more relevant, however there is more complexity here because as outlined in his note below, ROIC isn’t a good way to assess M&A (too low initially, too high in later years)
What matters actually is not just the absolute ROIC but the change in ROIC – Return on Incremental Capital (ROIIC), as this gives a sense of incremental return on new investments and the direction that the overall ROIC is trending
Due to the volatility in annual ROIIC, it may make sense to calculate ROIIC on a rolling 3-5 year basis, as shown in the Cisco example below
ROIC can help shed light on a company’s competitive advantages and strategy. There are broadly two sources of competitive advantage:
- Consumer advantage: habits, high switching costs, search costs etc
- Production advantage: scale and cost advantage, access to inputs, IP etc
By breaking down ROIC into its two components - NOPAT margin and capital turnover - we can glean insights into a company’s competitive advantages:
For example, a business that has a high ROIC due to a low margin but a high sales/capital turnover, like a low cost retailer, probably has some key advantages on the production side which should be analysed
A business that has a high ROIC due to a high margin but low sales turnover, like a luxury brand, probably has some key advantages on the consumer side which should be analysed
The below chart summarises these relationships:
Finally, ROIC is a key driver of valuation, something that is often forgotten when ascribing multiples to a business:
$SE The word "efficiency" was mentioned 27 times during the call. Management talked repeatedly about the cultural changes that are now permeating all aspects of the business and decision making. As a result, we now seem to have line of sight on profitability by end of next year
On Shopee. Right now the battle is just to generate any growth in GMV given repeoning+macro+reining in of S&M spend. However the monetisation levers are being pressed hard – according to my calcs this was just about the largest quarterly increase in take-rate recorded.
Takerate + reduction in S&M has seen the biggest quarterly reduction in both EBITDA loss and S&M per order to date. This also includes most of the $77m of one-off severances for all the job cuts, so if you exclude that the overall EBITDA loss per order reduces closer to 20c / ord
Michael Mauboussin's 2014 paper "What Does a P/E Multiple Mean" is a great guide to understanding the link between valuation, multiples and business economics (ROIC).
Here are some of my notes from reading it:
Multiples are a shorthand for valuation, not the valuation itself. Valuation should be driven by earnings and CFs
The value of a firm can be thought of in two parts:
- steady state value where NOPAT is held constant into perpetuity
- future value creation (growth opportunities)
The steady state component simply assumes that there's a core base of a business, the NOPAT of which can be maintained forever, and ROIC=cost of equity, so the steady-state PE multiple can be expressed as 1/CoE. eg. if the CoE is 8%, then the steady state PE mult should be 12.5x
Updated market share data of public cloud GPU and CPU deployments
TLDR: $NVDA and $AMD continue to take share.
Strong month for Nvidia with them supposedly claiming 100% of all incremental accelerator deployments in July, accounting for 84% total share (up from 82% I had in my deep dive).
This is interesting seeing Nvidia's share broken out by chips. Its flagship A100 has seen a fast ramp since its launch in 2020 but still only c.10% share. Quite surprised to see the older gen V100 (for training) and T4 (for inference) dominating and still being bought in droves
Very interesting data from Jefferies showing CPU and GPU instance share in the top cloud providers. Haven't come across this level of granularity before $NVDA $AMD $INTC
CPUs - clear trend of AMD taking share from Intel, but also AWS's Graviton being pushed strongly as well
Nvidia unsurprisingly dominates accelerator instances with over 80% share across all cloud providers, however interesting how much AWS is pushing its inferentia chip for inferencing workloads
Which means that Nvidia's share of AWS accelerated instances is falling and is the lowest amongst all the cloud providers (though still ~70%). Similar marginal share erosion happening at GCP with Google's TPUs, but to a lesser extent
$SE Overall a decent result given all the headwinds and tough comps. Some updated charts and KPIs that I track below along with some commentary:
Shopee GMV sequential decline was well expected by now but still 38% growth yoy and 64% rev yoy was a bit better than I expected. Take-rates flat qoq but these should see a nice increase next quarter as commissions have been raised across the board in the last few months
Good to see S&M per order reached its lowest level ever this quarter, although if i had to nickpick at anything it's the large increase in HQ costs which offset that somewhat, which meant total loss per order of $0.40 was largely flat.
The TLDR for those who want to cut straight to the summary at the end.
Feedback welcome as always.
The next article will cover Nvidia's expansion beyond the GPU, its software layer and full-stack computing approach, and competitive landscape.
If you enjoyed this, a retweet or a share would be greatly appreciated