Cloud Wars, Microsoft and AWS - An application of game theory
Author: David Nagrosst, Associate at Portman Partners
In this paper, we review how strategic game theory and economics is used within the cloud computing services market between the two largest competitive players in that market, Amazon (AWS) and Microsoft (MS). This paper will be broken into four principle concepts and will be covering and referencing elements of the previous group assignment throughout this paper. We will first introduce cloud computing services and highlight some essential characteristics of the cloud followed by a review of the critical timeline as presented in our group assignment, and thirdly we will provide a narrative around our timeline using theory, research, and evidence to support the case of how game theory and economics are applied. Finally, a conclusion is provided for how Amazon and Microsoft are in a game and the importance of being aware that you are in one to proactively develop a strategy to achieve an optimal outcome.
(Wikipedia, 2018), describes cloud computing as “shared pools of configurable computer system resources and higher-level services that can be rapidly provisioned with minimal management effort, often over the Internet. Cloud computing relies on sharing of resources to achieve coherence and economies of scale, similar to a public utility.” While cloud computing has been available and previously sold in basic forms such as selling a fraction of a server’s resources through virtualization, “AWS was first to market with a modern cloud infrastructure service when it launched Amazon Elastic Compute Cloud in August 2006. Surprisingly, it took several years before a competitor responded.” (Miller, 2016) Amazon had taken a first-mover advantage (FMA).
How is the cloud different from what we had before and why is it so critical to innovation? Let's first examine Ronald Coase's work on why do firms exist and “...why transactions take place in firms. Are markets so efficient at allocating resources? His answer, yes, if simply, using the market is costly. The most important costs are (i) discovering what the market prices are and (ii) negotiating a contract for each exchange transaction in the market. According to Coase, these costs can be avoided inside the firm through a degree of vertical integration.” (McNutt, 2018 #1) In addition, transaction costs also include research, negotiations, monitoring, and the legal work required to contract, among other things. The cloud reduces and eliminates most transaction cost problems by creating pricing transparency making discovery easy, hourly charges with no locked-in contracts, standard online contracts and payments, and the vertical integration of all the resources required to set up any software over the Internet. The modern cloud makes the infrastructure needed for IT products like computing, storage, and Internet-accessible and consumable immediately to anyone with a credit card. The resulting innovation has been a vital enabler of the sharing and ‘gig' economy by reducing barriers to entry for innovative startups such as Uber, Netflix, and Airbnb, that have been born out of the cloud to launch and/or scale their software products.
Not surprising, by making IT accessible to so many, it grew rapidly, resulting in the score of competitors who wanted to capture the growth potential and profits of the cloud market as shown in Figure 1. Profits, after all, attract entry (Besanko et al., 2018), and many entered the burgeoning market; but with the investment required proving higher than many of them anticipated, they eventually exited. Economies of scale and a significant amount of upfront investment is needed to achieve a cost advantage over the competition. (Besanko et al., 2018) Cloud, as pointed out earlier, is very much akin to a power utility company regarding the similar type of high upfront investment required to compete and this ultimately creates a barrier to entry and forces exit to those who don’t have the stomach or pockets for the size of investment needed. As shown, competitors who did not commit to investing the large sums had to ultimately exit and are regulated to the fringes, offering value-added services around AWS and MS offerings.
AWS was able to achieve low pricing by developing self-service capabilities and vertically integrating and managing their supply costs making them a cost leader. (McNutt, 2014) Identifies 5 steps to being a cost leader: step 1 is having a global economy of scale, step 2 is optimizing labour productivity, step 3 is normalizing the wages to increase output efficiency, step 4 is controlling and decreasing fixed costs, and lastly, step 5 is about understanding capacity and managing it well. By many accounts, AWS is a cost leader because they build their data centres to massive economies of scale spreading their fixed costs, buy bulk amounts of Internet capacity, designs and manufacture their equipment to reduce overheads, and minimizes their labour costs through self-service and automation. (Richman, 2017), writes about Amazon's secret weapon, "Market-leading public-cloud company Amazon Web Services … design not only its own routers, chips, storage servers and compute servers but also its own high-speed network.” Amazon is currently the only cloud company to attain this level of cost-reducing vertical integration in the supply chain.
Meanwhile, Microsoft and its far deeper pockets were able to deploy the necessary capital to achieve economies of scale to compete against giant AWS and made their entry in 2010 to take on a second-mover advantage (SMA) and become the rival competitor of Amazon. Due to their economies of scale and capital investment deployed, both AWS and MS's long-run average cost structure is far superior to those competitors who just dipped a toe in the water. Low long-run average costs profit the bold. AWS is the real cost leader because they vertically integrated their supply chain, but MS is not far behind and nor do they need to be the cost leader because of their additional SaaS offering discussed later. Cost leader types may have so much demand that they must reserve capacity in advance making production drive demand. (McNutt, 2014) Thus we see MS and AWS deploying huge infrastructures and incurring massive costs before there is demand, and just as before, once available demand follows.
Next, we’ll examine two critical timelines (CTL) for Amazon and Microsoft, one which we provided in our group assignment as shown in Figure 2 and an expanded CTL with added detail and recent events as shown in Figure 3. Figure 3, which is based on announcements rather than availability, adds depth and lightly explores two other sub-games. In Figure 2, we can see the early game, with AWS taking on an FMA in 2006 and Microsoft coming late into the game in 2010. However, Microsoft only starts to accelerate its market expansion in 2014 with the installment of Satya Nadella as CEO of Microsoft.
When MS first entered the market, MS and AWS competed on price. AWS then quickly started to differentiate and "introduce new products" providing customers with more choice and "uses complex price actions" to make their products appear less expensive by de-bundling and giving customers the ability to customize. (Rao, 2000) MS’s entrance also triggered AWS to invest heavily in geographic expansion, expanding into eight locations and six countries in seven years from 2006-2013. Using their FMA and the strategy of "get big fast – by aggressively growing market share and keeping prices low…."(Halaburda, 2014) MS responded in 2014 with a barrage of geographic expansion and product development to catch up and surpass the number of AWS locations and markets. During this time between 2014 and 2017 – AWS continued to reduce its price. Amazon is known as the price leader for cloud and signaled to the market in their 2015 shareholder letter that they have “… dropped prices 51 times, in many cases before there was any competitive pressure to do so.”(Amazon, 2016). In the CTL, we can also see that MS eventually catches up to AWS concerning market locations, pricing, and features, bringing it to near-complete parity with AWS, except for market share.
On May 3, 2017, AWS announced another price reduction published by (Barr, 2017). Microsoft responded on May 19, 2017, with not just a price decrease, but also a signal with the release of the following public statement, "We believe in providing the most innovative cloud offering, yet have a longstanding commitment to match AWS on price drops." (Herskowitz, 2018) In Rao’s work on, "How to Fight a Price War," (Rao, 2000) one of the tactics to avoid a price war is to "reveal your strategic intentions" and this is precisely what Microsoft did. MS signaled that they intend to compete on quality rather than price, and since then, AWS has limited their general price reductions.
Another tactic, (Rao, 2000) mentions in his work is to "reveal your cost advantage." We will touch upon Microsoft’s not so secret weapon related to this tactic. In addition to traditional cloud services, Microsoft also offers Software as a Service (SaaS) products that have gained significant popularity like Microsoft Office 365. Microsoft SaaS offerings are deployed into the same locations and leverage their cloud investments. This affords Microsoft higher margins over the same fixed investments, and even though AWS has overall lower costs, MS has the ultimate cost advantage because their fixed investments spread over more revenue and profits. Amazon, on the other hand, has less leverage over their fixed investments which are limited to their e-commerce and prime offerings. Microsoft, therefore, can respond to AWS price drops without impacting their SaaS revenue which shares the same fixed investment.
In addition to the game of geographic expansion, price, and features of their core offering, there are also at least two sub-games that I provide some surface-level details and is included in Figure 3. Firstly, we have the Internet of Things (IoT) service offering which Amazon launched in Oct 2015 and Microsoft followed in Feb 2016. Lastly, we have technology adoption laggards of large multinational corporations with enterprise workloads which is a vast and relatively untapped market. AWS announced a partnership to handle enterprise workloads in Oct 2016 and Microsoft followed with its announcement in July 2017. There are even more such subgames for areas outside of pure cloud technology, but for the sake of brevity, we will not dive further into them.
Next, we’ll have a look at anticipating the next moves. As shown in Figure 3 below, AWS has not yet responded to the MS announcement of expansion in Switzerland. What will the AWS response be? We can guess it will be within 6-15 months based on historical values of the site is unexpected, which I believe is the case here given Switzerland’s small relative population size. On the other hand, MS has yet to respond to the AWS announcement of expansion in Indonesia as reported by (Mah, 2018), but believe there may be a smaller gap in time between announcements given the population size (4th most populous in the world) and high GDP growth of Indonesia. See the predictions shown in Figure 3.
With ample evidence provided, we can demonstrate that Microsoft and Amazon play a game competing on price, location, options, and features and interdependent with each move carrying weight and impacting the other in a zero-sum game. An example of this is captioned in the news release," Amazon lost cloud market share to Microsoft in the fourth quarter: KeyBanc." (Novet, 2018) We see the game within a blend of competition and monopolistic competition when the full functions of the cloud are used and bounded to the platform and then are technically differentiated via API's and the human talent pools within their network to manage it. While public cloud providers can exploit the elasticity of demand by lowering prices to capture more workloads from traditional IT loads sitting in data centres on legacy infrastructure, its response will be delayed from the complexity of IT.
Further to this, cloud price transparency is high since its price lists are published, its “no contract terms” and “pay per use utility” model lowers most transaction costs, and 3rd party tools and technology developed to simplify moving IT workload between different cloud competitors make cloud a perfect medium for price wars to occur. Alas, I have demonstrated in this paper that they indeed have happened, only recently reaching a Nash Equilibrium on price. Lastly, they are sending each other numerous signals through marketing press releases and other public announcements on pricing, market expansion, product improvements, and entry into different business segments with the most significant signal coming from Microsoft indicating they will like to end the price war and compete on quality instead.
Next, we look at who is sending the signals. The management of Microsoft and Amazon and their respective CEO's all being players in the game is ultimately responsible for setting strategy for others to follow are thereby responsible for these ‘signals.' Jeff Bezos, CEO of Amazon, is notorious for embracing low margins and creating excellent customer experiences with quotes like "your margin is my opportunity" and backing this up with constant price cuts and constant product development and innovation. Amazon tends to focus on organic business growth and low margins. Microsoft is led by CEO, Satya Nadella, and brought a focus on cloud with quotes like “the world is about cloud-first, mobile-first” and from 2014 goes on to demonstrate the full weight of Microsoft by executing a swift global geographic expansion of their cloud to a point where AWS lost cloud market share for the first time in its storied history. Microsoft is all about aggressive growth and expansion. (Lee Wan Kuin et al., 2018) We can see from the CTL we constructed that after MS's aggressive expansion in 2014, AWS continued to keep pace with their previous style of one to two expansions per year keeping with their style until 2017 and matching investment for their low margin strategy. Meanwhile, Microsoft matched the aggressive approach chosen by Satya with a rapid global expansion that Amazon tried to prevent by expanding early to capture market share before Microsoft entered but ultimately could not stop Microsoft's rise.
In conclusion, the body of this paper demonstrates how "modelling behavior using game theory is proving especially useful when applied to economics." (Montenegro, 2018) Furthermore, to enjoy the optimal attainable outcome, management must first understand that they are a player in the game and their interdependence with other players. (McNutt, 2014) Lastly, the gathering of information related to the geography or segment of the game, corporate intelligence, identification of the close, competitive rival, and the understanding of the historical actions of the close rival and their key management to aid in the construction of the CTL. (McNutt, 2014) It helps us to understand what has happened to find the patterns and predict the next move to reduce uncertainty and obtain the best attainable outcome. After much research, it is unavoidable to pick a side, but ultimately Microsoft has played the game well and is in a good position to build off its SMA and rapidly expand its market share over AWS.