Arbitraging AWS reserved instances

data points graphic

At Strategic Blue, we are often asked whether we offer RI Arbitrage services for other AWS Resellers and Channel Partners.  The answer is both yes and no, due to confusion around the term “arbitrage”.

As former commodity traders, Strategic Blue are very focussed on safely managing cloud-related financial risk, both our own, and that of our customers and partners.  So the issue with use of the word “arbitrage” is that it tends to be applied to situations where the profits resulting from a matching purchase and sale are not in fact risk-free.

Take the definition provided by

“Buy low, sell high” is the mantra of the stock market. Perhaps the most extreme example of this is arbitrage, the act of buying and selling goods simultaneously in different markets to gain an immediate profit. Impressive, but tricky. has glossed over the important point that the goods that have been bought and sold must be fungible, i.e. absolutely identical, in order to be an arbitrage.

The Oxford English Dictionary gets closer, by referring to the “same asset”, but still doesn’t emphasise the lack of risk (as a result of the asset being the same):

The simultaneous buying and selling of securities, currency, or commodities in different markets or in derivative forms in order to take advantage of differing prices for the same asset.

The relevance of this to “RI Arbitrage” is that the buying of AWS “Reserved Instances” at a low price, in combination with the selling of “On Demand Instances” is certainly not risk-free, as the Reserved Instance pricing instrument is not identical to the on-demand pricing instrument.

Let me back up a minute, and provide a history lesson.  I’ll start with the punchline – Barings Bank, once a major UK financial institution, went bankrupt as a result of a series of events that started with a misunderstanding of the nearly risk-free, so near-arbitrage, activities of a trader called Nick Leeson.  Nick was authorised by the bank to take advantage of the fact that Japanese Government Bonds could be bought and sold both in Osaka, and in Singapore.  The bonds were absolutely identical in both markets, thus it was absolutely possible to buy bonds in one market, and simultaneously sell them in the other, and if the buy price is lower than the sell price, then a profit was made.  The actions of Nick Leeson’s team at Barings, and others, who had traders operating in both markets, helped to keep the two markets trading with equivalent prices, or at least the prices in the two markets were close enough that the gap between them was kept inside the “arbitrage channel”.  So when the prices were inside the arbitrage channel, the transaction costs were greater than the arbitrage opportunity.

The reason why it was “nearly” an arbitrage, is that the trading pits were chaotic, and occasionally mistakes were made.  As a result, sometimes the traders held long or short positions, that they were not authorised to have, as a result of accidentally failing to buy or sell quite the right number of bonds in each market, at quite the right time.  For the rest of the story, about how a handful of small mistakes ultimately brought down one of Britain’s biggest banks, I strongly recommend reading the book or watching the movie Rogue Trader starring Ewan McGregor.

So there are a range of risks that an AWS Reseller runs when engaging in so-called “RI Arbitrage”.  

The default scenario is the the customer has been running a lot of on-demand instances persistently, and hence not using the flexibility that the on-demand pricing model affords the customer to turn the instances off, and not pay for them, when they are not needed.  If the customer has not bought Reserved Instances themselves, it is then very tempting for the AWS Reseller to make that 12 or 36 month commitment to AWS, in order to access the considerably lower pricing, while still charging the maximum on-demand rate to the customer.  The obvious risk here is then “what happens when the customer turns off the instances?”.  The answer should then be that the AWS Reseller loses money, unless they can replace that matching usage, under the same invoice, in a manner that does not breach the AWS terms and conditions.  This is because the AWS Reseller is “long” fixed price, fixed term cloud contracts for a very specific cloud “commodity”, and they had failed to “hedge” their position, by sharing the “arbitrage” with their customer, in return for securing a matching term commitment from the customer (what Strategic Blue call a Cloud Options subscription).

An alternative problematic scenario is that in the above example, instead of turning off the instances, the customer decides that they would like to buy the Reserved Instances themselves.  If this happens sometime after the AWS Reseller bought the matching Reserved Instances to take advantage of the “RI arbitrage”, then the AWS Reseller, who has been recognising profit on the arbitrage up to that point, suddenly loses the rest of the arbitrage opportunity, has to find a way to synthesise the effect in their billing systems of having matching Reserved Instances with a term that does not match that of the actual purchased Reserved Instances.  In this scenario, the AWS Reseller also picks up a short position in this scenario, which cannot (easily) be hedged, and will lead to the AWS Reseller buying at the prevailing On-Demand rate for the period between the end of the “arbitrage RI” and the end of the “synthetic RI”.  This will then create a loss that will wipe out some or all of the “arbitrage profit” that will have been booked in the previous financial year.

Hopefully the above explanation explains why Strategic Blue avoids referring to our cloud trading activities as “RI Arbitrage”.  Our approach to pursuing this opportunity, be it ourselves, or for our partners, is to put the customer first, giving them right of first refusal to convert their forecasts of future usage into commitments to us, or to our partners, in return for a low fixed price.  We make this as easy for the customer as possible, through our Blue Review service, and maximise what can be committed to, by synthesising commitments, which we call “subscriptions”, that can be for any term, and can even start in the future, should that match the customer’s forecast.  By entering into these cost-saving, budget-fixing subscriptions with customers, Strategic Blue picks up a position (usually a short position), which we then hedge as best we can by purchasing new Reserved Instances from AWS.  We then monitor our “Residual Position” very carefully, across all the many types of cloud resources we are trading, in order to ensure that we safely manage that position.  Running a Cloud Options business really is very much like what professional commodity traders do, applied to cloud computing contracts, in this case AWS Reserved Instances for EC2, RDS and even Redshift and Elasticache.

Related Posts:

About Us

Strategic Blue helps organisations buy cloud on terms to suit their needs, helping them accelerate innovation in the cloud and optimise long-term cloud costs.

Let’s Socialise:

Popular Posts: