INTC, AMD, NVDA: A Zero-Sum Gain?

Advanced Micro Device (NASDAQ: AMD) has virtually followed the same dramatic path of NVIDIA (NASDAQ: NVDA) in the recent period (see Figure above).  It is logical to look at AMD and NVDA the same way.  On the other hand, NVDA and AMD’s sharp movements are in direct contrast to the lackluster performances of Intel (NASDAQ: INTC) and Qualcomm (NASDAQ: QCOM) which are supposedly their competitors.  In order to do a proper valuation of AMD, the first task is to identify proper benchmarks for comparisons.

What Are AMD’s Benchmarks?

To make a sensible value comparison, we need to assume that all potential companies make relatively homogeneous products.  Further, product similarities are reflected in relative stock valuation in that the most relevant choice of the benchmark should be based on shareholders’ perspective.   A company is only relevant to AMD if its stock moves with or against AMD in reacting to the same industry events.   As a result, I first identify which stocks are relevant to be compared in valuations.  For AMD, it is widely assumed that INTC, NVDA, and QCOM are the natural competitors due to their product similarities.  We verify this assertion by correlating both the monthly returns between any two stocks over the last 5 years.   In Table below, the monthly return correlations suggest that all four stocks are somewhat related.  However, after the common market component (NASDAQ return) is taken out, only INTC, NVDA, and AMD are statistically correlated (in green).   Thus, these three stocks should be comparable benchmarks to each other.

corr

As investors in the tech industry love to use PE ratios to value tech stocks, comparing the absolute level of PE ratios seems less meaningful as tech companies, by design, are heterogeneous in growth potential.   Therefore, we elect to use a relative valuation approach, Growth Duration Model, with the assumption that the relative valuation of two entities is determined by their relative growth rates in earnings in the next n period.  A generic version of the model can be shown in equation below.

holtA typical example for the use of this model is that AMD is often compared with NVDA, as their relative market share in the discrete GPU space is considered a zero-sum gain.  Figure below indicates that the timing of the market share shift between AMD and NVDA is remarkably and closely mirrored with their relative stock price movements.

NVDA AMD Side by Side.v4

Triangular Valuations

The comparison becomes tricky at this time, since there are three stocks involved in the zero-sum game.  If there is some merit in the approach described above, it stands to reason that AMD, INTC and NVDA should be priced in some relative fair relations which reflect the different expected growth rates, assuming the underlying risk of each growth is comparable.  As there are three stocks which are fundamentally and historically related, there should be one relationship between any two stocks out of the three stocks.  So, using all the known numbers in Table below, the following three equations should be held among INTC, AMD, and NVDA.

PE

For AMD and NVDA: 

49/30 = [(1 + gamd + 0.00)/(1 + gnvda + 0.0054)]n

For INTC and NVDA: 

12/30 = [(1 + gintc + 0.029)/(1 + gnvda + 0.0054)]n

For INTC and AMD: 

12/49 = [(1 + gintc + 0.029)/(1 + gamd + 0.00)]n

To make the case more interesting, an extra wrinkle is added.  The usual way to use this model is to compute the fair value on the left side of the equation by plugging in the growth rate estimates on the right-hand side.  Assuming that the current market price is already at the fair level, there is a unique growth rate associating with each market price.  In other words, the stock has already priced in such a unique growth rate.  Therefore, there are three unique growth rates which simultaneously maintain the three relative valuations. The implied growth rates are considered more “unbiased” since the calculation does not require individual or subjective estimates.

How Much Earnings Growth Has Been Priced Into the Stock Prices? 

The remaining work of solving these three growth rates simultaneously is simply numerical.  The only variation for the solution is the assumption for the number of years of the workout period, n, which indicates how many years the difference in two growth rates will last.  In short, the numerical process generates a unique combination of implied growth rates for AMD, INTC, and NVDA, given a certain workout period.

ig

So it looks like, for the next one year, AMD’s stock has already priced in 210% earnings growth annually, given that INTC and NVDA have also priced in 3% and 110%, respectively.  Mind you that it is logical to derive a slightly lower growth rate for a longer compounded time period, so the AMD’s stock reflects an annual 210% growth rate for the next 1 year, 81% for the next 2 years, and 51% for the next 3 years.

Does This Make Sense?

Thus, in order to “justify” today’s three stock price levels, AMD’s annual earnings “needs” to grow 51%, INTC to grow 3%, and NVDA to grow 32% each year for the next 3 years.  In the meantime, let’s look at the following fact,

  • The historical annual growth rate for the last five years is -43%% for AMD and 30% for NVDA.
  • AMD’s last 1-year growth rate is 15% and NDVA is 55%.
  • The analyst 1-year (5-year) estimate for AMD is approximately 314% (-98%). For NVDA, the estimates are 18% and 12% respectively. INTC may grow between 5% and 8%.

While market price implied earnings growth rate in Figure above may make mathematical sense, does it make common sense?  Further, for all scenarios, AMD’s growth rates are consistently over 50% and NVDA’s growth rates need to be over 30%.  You can judge it yourself if it is too aggressive.

One obvious explanation for the seemingly large discrepancy is that the market has not incorporated all current information, resulting in one or both stocks being overvalued.  We concede that this possibility directly challenges the only assumption of this analysis (the three stocks are fairly priced).  This may explain why the more recent posts of both stocks have been bearish.

The other logical conclusion is that the market has incorporated more information than the analysts have used, that the current market prices are not overvalued.  For example, AMD’s new rollout Vega may deliver a better than expected. Incidentally, the primer author of this site. Alex Cho, has conducted an AMD pro forma estimate of over 300% earnings growth by 2018, which is in the bulk part of the analyst’s forecast,  though, the evidence here suggests that the market only priced in two third of his forecast (210%/300%) due to the uncertainty of the MPU space and GPU trends.   As a result, his AMD $20.42 (2017) and $29.89 (2018) price targets may be a bit ambitious.

That being said, NVDA appears to have a more promising outlook for the higher “unaccounted” growth rates. The future payoff from the investments in VR, AI, and self-driving technology is very likely to explain the difference between the low 30% versus the street estimate of 18% earnings growth, which is more achievable compared to AMD’s over 50% long-term growth rates.  In summary, from the imbedded growth rates already priced in the stock prices, if both AMD and NVDA are overvalued, AMD is more overvalued than NVDA.

Finally, it appears that the market has a similar concern.  Notably, both AMD and NVDA have had an over 20% pullback from their respective recent highs.  As there has been no new fundamental information, the drop in prices is consistent with a typical “correction” for the magnitude of 15-20% range.  Although, my conclusion is based on the stocks prices at March 3.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s