AMD and the death of the crypto-mining GPU

(AMD stock and its options are extremely liquid as AMD is one of the top 10 most traded stocks.)

Previously, I’ve written about how AMD’s profitability has been inflated by the cryptocurrency bubble.  This will likely come to an end in the coming quarters as the market for video cards will shrink thanks to the emergence of ASICs for top cryptocurrencies.  We are already seeing the effects of this as prices of video cards continue to normalize at the retail level.  From PC Part Picker:

Most investment bank analysts anticipate that AMD’s earnings will go to the moon.  While that type of behaviour may help their employer earn underwriting profits if AMD chooses their employer for an equity raise, these optimistic projectons are unlikely to materialize.

I/B/E/S estimates for AMD, courtesy of Sentieo.

In my opinion, AMD is a compelling short as its earnings will likely shrink rather than grow in the coming quarters.

Cryptocurrency mining explained (recap)

I’ll use Bitcoin as an example.  Bitcoin is designed so that a certain number of new coins will be created in 2018.  If those 657,000 new Bitcoins are worth $3,500 each, then that’s $2.3 billion in Bitcoin that will be given away to miners.  The miners will fight over the fixed number of Bitcoins that will be created.

I’ve explained why cryptocurrency mining exists previously in the post “Blockchain is a useless technology“:

The idea is that users would use their CPU to perform a lot of math calculations to support the integrity of the blockchain.  Record keepers in the network will trust the side that has poured more computing power into its calculations; cryptography allows each node in the network to easily verify the amount of math performed.  If the honest users do more math than bad actors trying to disrupt the system, then the Bitcoin network will function as intended.  If the good guys don’t dedicate a lot of computing power to Bitcoin mining, then a bad actor can simply round up more computing power than the good guys and gain the trust of other nodes on the network.  To try to prevent this, there’s an incentive system for the good guys to build up a large standing army of CPUs.  Inflation is designed into Bitcoin to encourage the good guys to maintain a large army.  This is done by giving out Bitcoins to Bitcoin miners depending on how much work they put in (the number of Bitcoins given out in a year is more or less pre-determined).  Because the price of Bitcoin has skyrocketed, billions of dollars in Bitcoins will likely be given away this year to miners fighting over their share of Bitcoins.  The ‘mining’ analogy is used because Bitcoin miners put in work and are rewarded with digital ‘gold’.  Unfortunately, the current system results in an overhead of billions of dollars for a very low volume of transactions.

There are some subtle economic differences between the various cryptocurrencies.  For example, Bitcoin is designed so that the coin creation rate will be cut in half in set intervals.  Eventually, no new coins will be created.  These economic differences likely will not matter much as cryptocurrencies will fail; they are technically inferior to payment networks that have been tried in the past such as E-Gold.

Why the GPU market for crypto is dead

Originally, Bitcoin was mined with CPUs.  Once software for mining Bitcoin with GPUs became available, mining shifted towards mining Bitcoin with GPUs from AMD and Nvidia.  Then market shifted yet again when companies such as Bitmain designed ASICs for mining Bitcoin.  Because ASICs specialize in doing only one thing really well, their economics are far superior to more generalized computer chips such as GPUs and CPUs (CPUs are the most flexible and least specialized chips in a computer).  Because ASICs are far more cost-efficient than GPUs, they have pushed GPUs out of Bitcoin mining.  This occurs because the competition over the fixed number of Bitcoins can be intense.  As competition increases, the amount of computing power needed to mine a single Bitcoin goes up.  This increases the cost of creating a Bitcoin, the biggest of which is electricity.  Only the most cost-efficient methods of cryptocurrency mining survive, pushing out less efficient methods such as GPU mining.  Where ASICs exist, they will push GPUs out of the market.

Bitmain’s IPO prospectus disclosed that Bitmain began shipping ASICs for Ethereum (the second biggest crypto mining market) in March 2018.  This effectively killed the market for GPUs when it comes to mining Ethereum.  Bitmain makes ASICs for Bitcoin, Bitcoin Cash, Ether, Litecoin, Dash, and Zcash (see page 161 of the PDF of its IPO prospectus).  As Nvidia noted on its Q2 2018 conference call, the crypto market for GPUs is dead:

It is the case that we benefited in the last several quarters from an unusual lift from crypto. In the beginning of the year, we thought and we projected that crypto would be a larger contribution through the rest of the year, but at this time, we consider it to be immaterial for the second half.

Options

As mentioned previously, AMD is one of the most actively-traded stocks at the moment so its options are fairly liquid.  It may make sense to target options that expire 2 earnings releases from today.  (I happen to mainly own options expiring in Jan 2020.)  I am speculating on continued earnings misses to act as catalysts for the stock to fall.

Unfortunately the options are quite expensive as AMD has been an extremely volatile stock; the stock regularly moves several percent a day on no news.  Nonetheless, the options seem compelling to me because this is a short position with a well-defined catalyst.  Note that trades involving “well-defined” catalysts often don’t work out.  When China Medical defaulted on its bond payments, it seemed like a compelling short as bankruptcy should drive the stock towards 0.  While that trade was slightly profitable for me, weird things happened.  A wine mogul who had a large long position in the stock moved his shares into a cash account, triggering a short squeeze that caused many short sellers to lose money.  Weird things regularly happen when you bet against stocks.  If you bet against AMD, be prepared to lose money as your trade may not necessarily work out.  Also note that I’ve blogged extensively about Pretium Resources and that trade has not worked out the way that I thought it would.

The bottom line

With a TTM P/E ratio of around 55.8 (mkt cap $20,088M / $360M TTM net income), AMD should trade at a below-market P/E ratio given its declining earnings going forward.  Over the past three decades, AMD has significantly underperformed the S&P 500.  In the past decade, it has lost billions of dollars (on a GAAP basis).  In a normal environment, AMD would not be a good stock to own and clearly does not deserve the inflated valuation that it currently has.

The market is wrong about AMD being a growth stock.

Disclosure: AMD is currently my largest short position.

7 thoughts on “AMD and the death of the crypto-mining GPU

    • Nvidia is better managed than AMD. Nvidia doesn’t have a unique ‘moat’ like AMD’s x86 license (only AMD and Intel can make x86 processors that are backwards compatible with a lot of software that has been written). Nonetheless, Nvidia has managed to make a lot of money over time and continues to innovate. Their CUDA offering has created a new market for their video cards / GPUs. They are also making accelerators for machine learning, chips for mobile devices like the Nintendo Switch and tablets, etc. etc.

      Over time, I expect Nvidia to outperform AMD.

  1. I’ve heard something like that intel has no server chips due within a year that are competitive with AMD’s. Are CPU sales projections not significant enough to affect your short thesis?

      • To expand on that a little more… the best content would be benchmarks on these server chips doing real-world tasks. Nobody is going to use a high-end server to do ray-tracing. People will use high-end servers for things like databases. Journalists don’t create that type of content as the market for that content is extremely small.

      • I haven’t read that, all “I’ve heard” comes from semiaccurate.com, whose articles over the recent years have painted a bad picture for intel and don’t support an expectation of superior products in the near future. I trust semiaccurate for tech predictions more than for any market predictions (intel has remained dominant even in times when its tech was inferior), but it’s probably worth reading over some of their AMD and intel stuff before shorting AMD.

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 )

Google+ photo

You are commenting using your Google+ 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 )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.