The competitive advantage of Dutch Datacenters. Using Porter's Diamond to analyse the datacenter industry

(1)

The competitive advantage of

Dutch Datacenters;

Using Porter’s

Diamond to analyse the datacenter industry

Author: Koen Molenaar (s4063589)

Masterthesis Economic Geography Nijmegen School of Management Radboud Universiteit Nijmegen

(2)

RADBOUD UNIVERSITEIT NIJMEGEN

The competitive advantage of Dutch

Datacenters

Using Porter’s Diamond to analyse the datacenter

industry

Molenaar, K. (Koen)

Supervisor: Arnoud Lagendijk

(3)

Abstract

The digital infrastructure is one of the best performing industries in the Netherlands. With high grow percentages and a promising future this industry has been designated as the ‘third mainport’, meaning that there is a relatively large quantity of goods passing considering that it is in a small country. This research investigates how the datacenter industry gained this position and whether how they are spatially manifested in the Netherlands. Using Porter’s diamond for the competitive advantage insights have been given on how the competitive advantage came to be and what factors play a role. This research aims to give an insight in a very technological industry from an action based, geographical perspective.

The first and main conclusion is that there is a lack of data, the interviews that have been conducted have shown strong qualitative arguments but proving them with factual data has turned out to be almost impossible. In the interviews and existing literature strong correlations were found, however, data shows the exact opposite correlation. Three explanations can be given for this gap. The lack of accurate data (1) is caused by the recentness of this industry’ existence. Furthermore, the diversity in preferences for different types of datacenter (2) make it hard to generalize everything into one industry. Big financial investments that have to be made in order to construct and maintain a datacenter make that the industry is not always transparent (3). Deals are made behind closed doors.

(8)

1. Introduction:

We work in the cloud, we browse web pages, visit sites and discuss on fora. By using geographically defined terms we try to make the internet tangible. Society is depending on the internet, not only do we do our shopping or watch videos, our sewer system is connected to the internet, our electricity network relies on functioning digital infrastructure. We heavily rely on a constant flow of information, not just limited to the internet. You could even argue that the internet should be part of the Maslow’ hierarchy of human needs (van Mensvoort, personal communication, 19-05,2016). Even though the internet is widely accepted in society we hardly understand where it comes from. It is just there.

The infrastructure that makes the internet possible consists of many parts. Physical land cables or sea cables are just one part of the infrastructure that makes communication possible. Internet Service Providers (ISP’s), Internet Exchanges (IX’s) and datacenters play a crucial role in the process of data transportation. Together they are called the digital infrastructure, the backbone of the internet. This backbone is spread across the world, making some areas better connected than other areas. Allegedly, states have cut off parts of the internet and made it hard to access all data for some.

There is a hype which followed a longer trend of increasing data flows; the internet of things. Connecting devices, gathering the data and analyze it to improve processes. From trees in Amsterdam to wearable devices that monitor your blood pressure. The quantity of information that is being generated is immense and must be stored somewhere. For this, we have datacenters, big buildings full of computer equipment. Information gets more compact, what used to be stored in a giant computer room is now stored on a few servers. However, the square meters of floor space that is used is still increasing. A consistent, sharp increase, in datacenter space is expected in the upcoming years.

There is a very strong presence of datacenters in the Netherlands. Together with other European cities Paris, London and Frankfurt, Amsterdam can be seen as one of the datacenter capitals of the world. The Dutch digital infrastructure plays a key role in the internet eco-system of the world. Furthermore, internet exchange AMS-IX (Amsterdam) and NL-IX (Netherlands) are both amongst the top 10 worldwide internet exchanges (Financieel

(9)

Dagblad, 2015). One reason that has been mentioned is the presence of sea cables, coming from the U.S. It is very unlikely that this is the only reason, this research will find other reasons and will explain how the Netherlands has become a key player in the internet eco-system.

It was only recently (22-9-2016), that insight has been given in the locations of datacenters in the Netherlands. The map provided by the also relatively young (2 years) trade organization shows a clear pattern, 65% of the datacenters are in or around Amsterdam. The other 35% are based in other provinces, near big cities or at universities where they are being used as supercomputers. Hyper-scale datacenters such as Google’s newly built datacenter in Delfzijl and Microsoft’s datacenter in Middenmeer are exemplary of the global presence of the Dutch digital infrastructure. It is no coincidence that the headquarters of Netflix is in Amsterdam. As of late the Google datacenter is in the news. Allegedly bribes have been paid for the construction of the datacenter (Financieel Dagblad, 2016).

The recentness of this event demonstrates the speed at which the industry is developing. Growth numbers higher than 10% a year are not uncommon in the datacenter industry. To illustrate this growth, last week a new datacenter in Amsterdam was announced. Not a regular datacenter, a datacenter of 100.000 m2 _{(Parool, 2016). For now, it is not} important to discuss the actual size (or implications this will have on the industry) of this, what matters is how these immense changes affect this research.

It Is no coincidence that most datacenter activity is located near Amsterdam. The presence of a large internet exchange (AMS-IX) only partly explains this. To fully understand why Google is located on the outskirts of the Netherlands and thus not being part of the Amsterdam hub will be researched further.

This research will focus on the datacenter part of the digital infrastructure. To be more specific; on why the Netherlands has become a leading country regarding internet traffic and datacenter storage. The primary question of this research will therefore be:

Which processes have made the Netherlands a leading datacenter country and how are datacenters spatially manifested within the Netherlands?

(10)

1.2 Relevance and research goals

The processes described in the introduction largely cover why this research is relevant for, on the one hand, society and on the other hand the academic world. The more obvious reasons are not the only reasons why this research has been conducted. This paragraph will briefly zoom in on other ways this research is relevant.

As part of the digital infrastructure, datacenters will become increasingly more important. The storage of data and the connectivity they provide determine accessibility in the future. The debate about data privacy is lively, being discussed on the international, national and local level. Who can access this data and under what circumstances can governments or businesses use the data. This research does not aim to give an answer to this question. Nevertheless, understanding the industry partly comes forth from this debate.

This research aims to provide insights into the different ways to look at datacenters and their specific requirements. There is a knowledge gap at the government. Therefore, it will end with recommendations. How come the Netherlands, as being a small country, plays such an important role in the internet infrastructure? Furthermore, the policy recommendations that will be provided in the last part of this research should give handles for policy makers. To remain one of the leading countries and to understanding the rapidly changing environment and the effects that come with it.

Furthermore, this research seeks to contribute to the academic debate whereas this is the first research to combine datacenters and geography. To dive deeper into the how’s and why’s of the datacenter industry and to add to the debate around the concepts that will be explained in the second and third chapter.

(11)

1.3 Subquestions

Multiple sub-questions have been setup and they will be presented in this paragraph. The final goal of this research is to provide policy recommendations for the Dutch government and related government organizations. These policy recommendations should provide a guideline on how to maintain (or even improve) the position as one of the world’s leading datacenter hubs.

The position a datacenter takes in a global, regional or local production network influences the way policy makers approach the industry. Datacenters may attract other businesses or can be looked at as a necessary requirement for other economic activity. The construction of a datacenter, the taxes paid and the jobs they create influence the effort a government puts in to attract and retain datacenters, hence benefiting the business climate for datacenters. Moreover, focused activities in a country may explain the relatively high number of datacenters in that particular case. This is, for instance, the case in Sweden where datacenters have been built since there was a surplus of power after the steel industry declined.

If the presence of a datacenter has a rather small effect on the economy, why then would a government make an effort to attract datacenter? Until recently the government has made no effort to understand the datacenter industry, A possible explanation could be; the datacenter industry is relatively new and is characterized by rapid change and innovation, leaving the government in the dark about their specific needs. The rapid change of the environment in which datacenters operate could be a reason for governments not to create policy regarding this.

(12)

The position of a datacenter within the digital infrastructure may, to some extent, explain the lack of interest from the Dutch government. Furthermore, the effect of the digital infrastructure on the ‘real economy’ should be taken into account here. This will be part of this research for two reasons, [1] to create a better understanding of the position of datacenters and the way in which they affect their surrounding and [2] to discuss the effect of datacenters on the economy and possibly explain the attention that has been given from a political perspective. This will be researched with the following question:

What (global) processes have led to the development of the datacenter industry and what role do datacenters play in the economy?

To fully comprehend the rise of the datacenter industry the question will zoom out to depict global trends that influence the datacenter industry. Several of these processes have been discussed in the introduction. These processes will be discussed briefly whereas they sketch global trends that do not necessarily affect the competitive position of the Dutch datacenter industry.

To further investigate the competitive advantage of the Netherlands it is necessary to understand what factors and processes shape the datacenter landscape in the Netherlands. First, to understand the variables and second, to see if there are any exceptions. Outliers or different cases that bias the datacenter landscape. This will be done with the two following sub-questions:

What variables have made the Netherlands an attractive location for foreign businesses and how are different types of datacenters spatially dispersed?

It is not obvious that there are no differences within a single country. Possible patterns of datacenters that can be seen might give new insights into why the Netherlands has become such a strong hub. Different types of datacenters have different reasons to invest in different places. The variables depicting those decisions all together form the landscape and can explain the competitive position. Regional patterns give more explanation power to the model whereas it is corrected for potential outlier cases and unforeseen effects.

(13)

To discuss those, several cases on the European level have been selected and will be researched with the following question:

What insights can other markets give in the determinants of an attractive business climate for datacenters?

1.4 Reader’s guide

The next chapter will give an overview of the relevant theoretical concepts that will be used in the analysis. The views of the main authors on competitive advantages will be discussed as well as the tools they have developed to research this topic in a structured way.

In chapter 3 the methods of this research will be discussed. The way in which the data is gathered affects the validity and trustworthiness of this research. Where the theoretical framework provides guidelines for the analyses of the data this chapter will further explain the consequences and the methods that will be used. Furthermore, it will zoom in on the different aspects of the data and the role of the researcher within in this research.

Chapter 4 will discuss the position of datacenters within the digital infrastructure. Answering the first two sub-questions this chapter provides insights into the business that goes beyond the datacenters themselves. The global processes that influence the datacenter industry and effects datacenters have on their surrounding might, to some extent explain why they are present at certain locations.

The analytical chapter that follows focuses on the Dutch case and an analysis of the existing policy of the municipality of Amsterdam, the NFIA and other involved actors that play a role in the datacenter industry. Furthermore, it will investigate how the Dutch hub has come to be and what makes it different from other countries/hubs. Chapter six will discuss multiple other European cases: Iceland, Sweden and Luxembourg. This comparative analysis will be carried out to further investigate what factors are important in the datacenter business climate. The final chapter will provide concluding remarks on the datacenter business climate followed by policy recommendations.

(14)

2. Theoretical framework:

In the theoretical framework, several terms will be explained. Terms that will be used will close the gap between datacenter terminology on the one hand and economic geography terms on the other hand. The main purpose of this chapter is to provide a framework from where an analysis can be made of why the Netherlands has become a key player in the worldwide digital infrastructure. Location factors play a key role in this process and will hopefully provide a better understanding of the competitive position of the datacenter industry. A good starting point is Porter’s Diamond of national advantage. This research will mainly focus on particular parts of the diamond, parts where the datacenter market distinguishes itself from other markets. By doing so this research seeks to offer new insights in this theory for a particular field. Furthermore, it hopes to provide policy recommendations for local, regional and national governments.

Figure 1, Porter’s diamond

Porter’s diamond is used mostly to compare the competitive advantage of nations. However; it can also be used for narrowly defined industries or even particular industry segments (Porter, 1998). In this case, we talk about the competitive advantage of a set of firms or industries rather than a comparative advantage (Wagner, 2003, p. 16). The comparative advantage of a country can be seen as the result of singular advantages from multiple industry segments. There are two types of advantages; industry specific advantages

(15)

for a country and generic advantages which a country has. These generic factors play a role, will be discussed but will not be thoroughly examined even though they play an important role in the business climate of the Netherlands. This will be done under the assumption that the countries which will be used for the comparative chapter do not differ too much from one another. Mehrizi and Pakneiat (2008, p. 86) found these generic factors one of the most important factors while trying to explain the Telecom sector in Iran, since they are extremely relevant in developing countries.

2.1 Other frameworks for industrial analysis

Porter’s diamond is not the only framework to analyse spatial patterns and competitive advantages. For instance, Mehrizi and Pakneiat (2008) have compared Porter’s diamond to a sectoral innovation system (SIS). SIS focuses on the innovation process and competence building. Something which is not taken into account in Porter’s diamond. They conclude that the diamond model is inappropriate to analyze sectors in developing countries for multiple reasons (Mehrizi & Pakneiat, 2008, p.84). They conclude their research by stating the following (p.84):

‘The diamond model is more suitable for analyzing production-related and market-intensive aspects of sectoral development which are more the case in short run’

The context in which developments take place influence the outcome of this research. There have been moderations to Porter’s diamond which will be discussed in the last part of this theoretical framework. The notion that Porter’s diamond is more viable for advantages on the short run is discussed there. Therefore, this will be an important part of the methodology and analysis later in this research. Before bringing up other critiques it is important to understand the aspects of the diamond to their fullest extent.

2.2 Factor conditions

Porter (1990. P. 77) divides factor conditions into two types, basic factors, and advanced factors. Resources, both human and physical, infrastructure and capital resources are all considered factor conditions. Infrastructure includes both physical infrastructure and regulatory infrastructure, this regulatory infrastructure includes government policies and programs (Ozgen, 2011). For this research, a broad definition of government policies will be

(16)

used. Another factor condition is the availability of financing [ CITATION Kul98 \l 1043 ]. Therefore we will also take a look at companies affiliated with the government.

As stated before this research will make a distinction between generic factors and industry specific factors. These industry specific advantages are for instance; highly skilled (specialized) personal, digital infrastructure, climate and the availability of (green) energy. Of course this is preliminary and these industry specific (mostly advanced) factors will be discussed in the following chapters.

2.3 Demand conditions

The quality of the demand is more important than the quantity. If the quality of the demand is high the products have to be more innovative hence the competitiveness of the country becomes higher. Demanding home buyers lead to product innovation and create technological advantages over other countries or industries. Insufficient demand from the domestic buyers forces businesses to improve technologically thus creating a competitive advantage over other nations. A change in demand can heavily influence an industry. If demand drops an industry has to adapt and find new ways to make money.

2.4 Related and supporting industries

Moon et al (1998) describe related and supporting industries as following:

‘Related and supporting industries are those whereby firms coordinate or share activities in the value chain or those which involve products that are complementary to the firms of a given nation.’

These supporting industries have to be in reach for the sector. They provide complementary services, technological advantages, buyers information or anything you can think of which may help the industry to gain a competitive advantage. One type of good which is of particular interest for this research are public goods; to be more specific, club goods. They provide network opportunities for existing and new businesses. The concept of club goods, as well as their effect, will be explained below.

(17)

2.4.1 Club goods

The underlying theory for club goods is club theory. Club theory examines club goods in economics and was first mentioned in the work of Pigou (1920) on tolls for congested roads. Club theory is predicated on groups creating utility through exclusive collective action[ CITATION Bru08 \l 1033 ]. Club goods are a subtype of public goods, excludable but not rivalrous (Cornes & Todd, 1980). They are only accessible by members of a club, these members must have chosen to become part of the club voluntarily and have the right to leave the club if they wish to do so. Economists generally define clubs as voluntary groups in which two or more individuals cooperate in order to create a commonly shared good [ CITATION Bru08 \l 1033 ]. Furthermore, access to club goods could be denied if a party chooses not to participate in the club. This is a very strict definition of club goods. [ CITATION Ols65 \l 1033 ] states that clubs are set up in order to: ‘exploit economies of scale and to

share public goods’.

A less strict, and more often used definition for club goods targets a specific type of good only beneficial for a specific branch. Those include; specific education, news websites for a particular field and trade organizations. A good can only be non-rivalrous when there is no congestion. As soon as congestion occurs the goods become challenged. This can, for instance, occur at high-end job seekers that provide special services.

One of the main issues of clubs is, when the extra value is created in the form of club goods, that not all members contribute for the greater cause. Some just take advantage of what the others have created, this is the free rider problem. The free rider is an agent gaining from the opportunities created by the other members without contributing. Weak exclusion criteria might lead to free rider behaviour. As a special private benefit deriving value, in part, from the fact that nonparticipants are excluded [ CITATION Bru08 \l 1033 ].

Club goods are generally seen as one of the reasons for a business to start in a specific area. It influences the decision-making process. By allowing members to draw from one another’s characteristics, clubs create private benefits for members. This research will examine whether club goods also play a role in the competitive position of datacenters.

(18)

2.5 Firm strategy, structure, and rivalry

2.5.1 Technology and innovation in Porter’s diamond

Innovation processes do not play a central role in Porter’s diamond. For that other models provide a better framework. However, Porter does argue that when there are more firms, geographically close together, the more rivalry there would be hence resulting in more innovation and a better competitive position. The higher the rivalry, the better (Mehrizi & Pakneiat, 2008).

Technological change may, according to Krugman et al (1993), lead to an effect called leapfrogging. The technological advantage gained in country A will be diminished if country B adopts the new technology to further improve its industry. The costs and even more important, the gains in country B will be higher than in country A, leading to a relative advantage over time. An underlying assumption in this model is that the new technology does not look appealing for country A in the beginning. Besides, the nation that is lagging behind will have relatively low wages for this industry, the combination of lower wages and better technology improves the relative position of country B (Krugman et al, 1993). This is in contrast with external economies of scale where the advantage of country A will become bigger over a given period of time. There is a divergence between countries.

2.5.2 Connectivity & economies of scale

Datacenters are, as discussed in the chapter before part of a larger network. Datacenters are connecting to other datacenters and modes of the internet. This happens in the so-called ‘meet me’ room, the place where the cables enter the datacenter. By connecting to other datacenters and nodes of the internet higher connectivity can be reached. The speed of a connection is (besides many other factors) also depends on distance. A connection will get slower when further apart. Even though we are talking about milliseconds this might be a reason to locate IT in a particular datacenter. By clustering, datacenters gain benefits that are otherwise not there. AMS-IX (Amsterdam Internet Exchange) is one of the most important international hubs. Together with London and Frankfurt these places are of outmost importance for the digital infrastructure in Europe. A datacenter is a hub with multiple connections to the internet [ CITATION Dut16 \l 1043 ]. Usually a large datacenter has more of those connections making it more connected. Even though we are talking about

(19)

milliseconds this latency is important for datacenters and might be an important factor when it comes to site selection. There is an important distinction that can be made while further researching economies of scale; internal and external economies of scale. Both decrease marginal costs of production but in a different way. This will be explained below.

2.5.3 Internal economies of scale

Datacenters are more efficient when they are big. They can cut in power costs, building costs and cooling can be more efficient. Expensive equipment to ensure uptime can be purchased if a datacenters is larger. Internal economies of scale decrease the marginal costs of production. This can be achieved by better production processes. In case of datacenters think about cooling, technological advantages or higher skilled employees.

Other internal economies of scale that one could think of are large purchase discounts and cuts in administration costs. Larger companies with bigger scale effects might have more efficient funds for R&D and thus be more innovative[ CITATION Kle97 \l 1043 ]. Another effect which should be taken into account is the learning-by-doing effect. Failures and successes help actors to understand the things they are doing now and should do in the future. Learning by doing can make processes more efficient [ CITATION Bos12 \l 1043 ].

2.5.4 External economies of scale and first mover advantages

The Dutch datacenter industry is acknowledged to be one of the leading countries industries in the world. External economies of scale can be gained as the result of growth in the industry where the business is active. This can, for example, be the existence of a trade organization, a wide range of experts that can provide specific information. Another important factor that will be discussed and taken into account in this research is the role of reputation. The reputation of an industry has an effect on the way a sector is viewed. If there is a positive reputation this might lead to more investments. Another example is the negotiation of discounts considering high volume trades. This would be if more than business is involved, otherwise, it will be internal economies of scale.

(20)

2.6 Government

The government has two important components in the original form of Porter’s diamond. First, the government has to be a stable factor. A stable government lowers the risk for a business to invest in a country, in other words, there is less risk when it is foreseeable how the government will behave when certain changes occur.

The second component is the enhancement of factors. By investing the government can create a better business climate for companies to investment and therefore gaining a competitive advantage.

It must be noted however that the role of the government part of the diamond is heavily discussed in the academic debate. This debate is very relevant for this research since the problems that have been addressed in the introduction might very well have to do with the role of the government. Therefore, in paragraph 2.8.2 this discussion will be analyzed.

2.7 Chance

Chance is an external determinant, chance events just happen (Jin and Moon, 2006). An economic crisis, for instance, can influence the competitive position of a country or industry. Incremental inventions, as well as industry-changing innovations, also belong to this particular part of the diamond.

2.8 Critiques and discussion

One of the arguments against Porter’s Diamond is his focus on the ‘home base’(Moon et al, 1998). Porter uses the country of origin of the company is the reason to explain the investment climate. Therefore some industry can not be explained. For instance the competitiveness of small countries such as Korea (Moon et al, 1998) or export-oriented industries such as the dairy industry from New Zealand [ CITATION Car93 \l 1043 ]. In order to solve this issue, Rugman and D’Cruz created a double diamond model, taking into account diamonds of other countries, becoming globally competitive (Rugman & D’Cruz, 1993). By addressing foreign location factors the general level of competitiveness of, for instance, Canada can be better explained.

(21)

This does however not explain the competitive position of small countries such as Korea and Singapore. By further examining the domestic demand it is not logical that these countries are considered as a hub. Moon et al (1998) have thus created a generalized double diamond, incorporating multinational activity into the model. This includes demand from foreign actors hence it can explain the situation of Singapore and Korea. This could also be the case for many European countries where demand is not solely based on domestic income but where export and import play a significant role.

2.8.1 Datacenters and cluster forming

The determinants of Porter’s diamond create industry clusters and impact new business ventures (Lehtinen, Poikela, Pongracz, 2006). Their research was carried out in Finland, examining environmental businesses. Of course this is a difference, however, as will be discussed below there are good arguments to believe datacenter clusters have the same effect. Porter argues that the elements of the Diamond are strongest within clusters.

Porter (2000) defines clusters as following:

“Clusters are geographic concentrations of interconnected companies, specialized suppliers, service providers, firms in related industries, and associated institutions (e.g., universities, standards agencies, trade associations) in a particular field that compete but also cooperate.”

Networks of experts and specialized suppliers are extremely important, they provide advanced equipment and generate high-end knowledge. Clusters have multiple advantages. One of them is described by (Pike et al. 2006, p.110):

“Successful cluster can forge ‘first mover’ advantages and benefit from externalities and increase returns to establish their competitive advantage at the expense of other localities and regions”

The first mover advantages are closely tied to piloting and knowledge transfer. Best practices can be shared within the cluster. This is in line with the argument by the Dutch government; to boost innovation and create an extra advantage for the digital infrastructure. The geographic concentration of industry cluster helps ease technology transfer and innovation [ CITATION Jus06 \l 1033 ]. The accessibility to social capital and networks create favourable

(22)

conditions for the exchange of knowledge and creation of new knowledge which help in recognition of opportunities (Ozgen, 2011).

2.8.2 The role of the government

Another, widely criticized aspect of the diamond is the role of the government. Rugman and Verbeke (1990) have argued that the government factor is valuable whereas it influences a nation’s competitive advantage directly. Especially when they pursue interventionist trade and industrial strategies as has happened in, for instance, South Korea. Hence they argue in favour of adding government policy to Porter’s diamond. Porter has developed a hypothesis concerning government policy.

Porter’s hypothesis predicts that environmental regulations can stimulate innovation in a sector. Porter & van der Linde (1995) describe the way of thinking before Porter’s hypothesis as following:

‘In this static world, where firms have already made their cost-minimizing choices,

environmental regulation inevitably raises costs and will tend to reduce the market share of domestic companies on global markets’.

Porter’s hypothesis predicts that when the strict policy is created it may affect businesses in the short term, leading to innovation, generating a technological advantage for the sector [ CITATION Moh01 \l 1043 ]. This innovation may lead to an early mover advantage as discussed in the paragraph above. Most work on Porter’s hypothesis has been conducted in the field of environmental studies. However, it is also applicable if we look at the datacenter industry, more on this topic will be explained in the fourth chapter where general knowledge about the industry will be given.

The main critique on Porter’s hypothesis is the assumption that businesses do not see opportunities to innovate themselves (Wagner, 2003, p. 6). If it would increase their competitive advantage in the long run, why not act on it?

(23)

2.8.3 Chance and entrepreneurship

Businesses are profit driven entities, therefore entrepreneurs seeking profit maximization are present. They seek or find markets which are not yet in equilibrium but where demand is higher than supply. ‘An entrepreneurial opportunity is a market imperfection that can be exploited bringing the market to equilibrium’ [ CITATION Kir97 \l 1033 ]. Those who want to open a datacenter have to spend a relatively large amount of money. Besides, it takes around 2 years for the construction to be finished and the get the datacenter operational. This is the moment where a datacenter can start finding customers, the market demands a quick response. However, if things are not going as expected and the demand is lower than the supply the datacenter needs to remain operational. The full datacenter, this includes the power costs. This financing might play an extremely important part in the decision-making process of datacenters and will, therefore, be further analyzed in the following chapters.

(24)

2.9 Conceptual model

The information and the concepts that have been given above form a conceptual framework.

Figure 2. conceptual model

The arrows in the conceptual model indicate the following relationships:

(1) The economic impact datacenters have will influence the attitude of the government. If the economic impact is high, the attitude of the government will be positive.

(2) A positive attitude is beneficial for the competitive position of the Dutch datacenter industry.

(3) The spatial manifestation of datacenters influences the competitive position of the Dutch datacenter industry.

(4) Regional processes influence the way datacenters are spread across the Netherlands. These regional processes are, for instance, ground prices, energy prices and availability of green energy.

(5) The presence of a functioning digital infrastructure has a positive influence on the overall economy. Not only a direct impact (jobs, taxes etcetera) but also the indirect impact by attracting related businesses. These effects will be discussed in the fourth chapter.

(25)

(6) Datacenters are spatially manifested in both the Netherlands and in Europe. The spatial manifestation in Europe influences the competitive position of the Netherlands as a

worldwide hub. The exact relation will be explained further on in this research, a comparative analysis will be made in chapter six.

(7) Regional processes influence both the national and international manifestation of datacenters. Processes involving law and data privacy heavily affect the datacenter industry (8) Not only regional processes play a role, international processes influence the spatial manifestation within Europe. Increasing dataflows are not regional but should be looked at from an international perspective.

(9) They not only influence the European manifestation. By being one of the leading datacenter hubs global processes directly impact the national ecosystem. One could even argue that discourse changes on the national level influence international processes. This, however, needs more argumentation which will be given in chapter five.

(26)

2

.9 Conclusion

The different aspects of Porter’s diamond all influence the competitive position of a nation. It can be seen as an accumulation of different industries. The abovementioned critiques do not outweigh the advantages Porter’s diamond has over other theoretical frameworks to measure the competitive advantage of an industry or nation. The theoretical concepts that have been presented offer a framework to answer the sub-questions and the primary questions that have been posed.

As shown in the conceptual framework there are, hypothetically speaking, causal relations between factors. It is, therefore, crucial to understand that outcomes in one area might correlate (one could argue that there is a causal relation between certain parts) with one another. Therefore, it has been made observable what factors play a role in this research. The indicators and the way they will be measured are shown in Appendix I.

The policy recommendations in the final part of this thesis have two goals. First, to optimize the behaviour of the government concerning regulations and second; to see whether the government should have a role in determining and possibly influencing the chance (and of course the more obvious government) part of the diamond. The next paragraphs will explain the relevant variables and how they will be measured to achieve abovementioned goals. They are based on the theoretical concepts that were described earlier. Furthermore, explorative interviews have been held to gain a better understanding of the different factors that are relevant for the datacenter industry.

It is not possible to ‘measure’ all relevant indicators. Some are qualitative, some are based on stories that cannot be proven by hard data. This critique and other side notes will be discussed in the methodological chapter following the indicators. All indicators are discussed briefly. In the following chapters, it should become clear why these factors will be taken into account and in what ways they affect the main question. Especially the fourth chapter should create a better understanding of the datacenter industry. The next chapter will discuss methodological considerations that have been made throughout the course of this research.

(27)

3. Methodology

:

This chapter explains why certain choices have been made and how they have affected the research. Information will be gathered through multiple channels namely; interviews with experts, policy study, literature study and observations. All these types of data will be briefly discussed in the second part of this chapter as the research material.

Throughout this chapter validity and trustworthiness of the research will play a central role. First and foremost, reflecting on the data gathering process. The first thing that will be discussed in this chapter is the role of the researcher, how does the researcher potentially influence the outcomes of the research? One of the research questions seeks to gain knowledge from a comparison between countries, therefore the case selection procedure will be explained. Finally, there will be a critical reflection on the gathered material and the effects this has had on the trustworthiness and validity of this research.

For this research, there will be no survey carried out. Response rates in the datacenter industry are low and trade organizations and other actors have an exceptionally clear view of what is happening in the datacenter industry (Grove, personal communication, 12-02-2016).

This research has been carried out in 2016. To be more precise, the data gathering was from January until October. One conclusion is that there is a lack of data about datacenters. The director of the trade organization for datacenters (Dutch Datacenter Association) has stated the same (Grove, personal communication, 10-02-2016). Besides several qualitative reports, there is no database containing information about where datacenter customers come from, how much capacity datacenters have or even more general data, how much income is generated from the datacenter industry. The recentness of the industry is an explanation. The Dutch bureau of statistics (CBS) does not have a code for the datacenter industry. The code ‘registrar’ is too generic and includes all those that register domains hence it is not a good measure. Another reason for this lack of data is the unfamiliarity with the industry. To have at least some understanding of technical details behind the processes that take place can be a solution. At this very moment, the CBS is working on a special code for datacenters, this was initiated by the DDA. The consequences of this ‘missing’ data will be discussed in the critical reflection.

(28)

3.1 Research strategy

In optimal situations gathered data will be checked by comparing it to three other sources; data triangulation. Unfortunately, this is not always possible, therefore the implications this has on this research will be discussed below.

3.1.1The researchers position within the research

There are positions from where a researcher can do his or her research. The background of the researcher as well as their position during interviews and observations affect the outcome of the gathered data. Creswell (2013), describes five different ways in which a researcher can be involved. The researcher can be a: ‘complete participant’, ‘participant as observer’, ‘observer as participant’ or ‘complete participant’. For this research, I am doing my internship at ISPConnect, the Dutch trade organization for internet service providers (ISP’s). All Dutch trade organizations involved in the digital infrastructure are part of a national organization called DINL (Digitale INfrastructuur Nederland). They cooperate to organize events, are responsible for representation in The Hague and bring together people with shared interests. Therefore, the researcher can be seen as completely involved in the industry in which this research will be conducted. However, the trade organization for internet service providers does not have mutual interests with the datacenter industry. ISP’s may be the customers of datacenters and therefore work together as part of the digital infrastructure but they have different interests in many other regards. Before every interview, it will be clearly stated that the research is intended for research purposes only. Furthermore, interviewees will be asked whether data can be published and if preferred, the name of the interviewee can be removed. An important note that should be made here is that the industry is small, actors are well-connected and aware of the industry in which they work. It might not be possible to make gathered data ‘untraceable’.

The position of the researcher can be linked to action research, actively being involved in the population that is being researched. ‘To actively engage and benefit groups outside academia’ (Wynne-Jones et al, 2015). Morely (1983) recognized the potential of this approach. ‘To exchange between geographical theory and method and multi-organizational problem domains that ultimately constitute the open environmental systems framing all human geographical research’ (Morely, 1983). One of the aims of this research is to

(29)

contribute to the debate evolving around action research as a geographical theory and practice. Indirectly, by carrying out this research with this approach.

3.1.2 Case selection

Three cases have been selected to further explore why datacenters have chosen the Netherlands. The cases will be selected based on their availability and their position within the network of datacenters. Ideally, they provide new insights in the way the datacenter industry has become so present in certain places. Therefore, these cases are preferentially outlier cases, generating new information about specific location factors, argued from a qualitative perspective.

3.2 Research material

The research material consists of; (1) existing literature, (2) policy created by Dutch and foreign actors, (3) interviews with experts and (4) observations in the field. This research will also use secondary datasets whenever possible, a short note on this and the lack of survey data will be presented in the last part of this chapter. A list of indicators, based on Porter’s diamond, is shown in appendix I.

3.2.1 Literature study

A literature study will be carried out to find out what research has already been done and how different theories relate. The goal of the theoretical framework is to make clear why and how geographical concepts play a role.

4.2.2 Policy analysis

This research will analyze Dutch and foreign policy. Furthermore, it will analyse business plans and websites to develop a cohesive understanding of datacenter investments. Besides these national policies, the policy of relevant organizations will be analysed. These include companies such as the Netherlands Foreign Investment Agency, InvestInHolland, Schiphol Group and AMS-IX. These are different organizations with different interests. This will be done wherever possible, not all organizations have a written policy document. Some only have an invisible set of rules and others have not written things down since the industry is changing rapidly. The lack of knowledge from the government side might also be a cause of this. More about this in the analyses of the Dutch digital main hub policy.

(30)

By investing and comparing different policies from related industries it should become clear what certain organizations offer for the datacenter market. The related and supporting industries are pointed out by Porter (1995) as one of the factor conditions. Furthermore, they might contain interesting arguments as for why datacenter are present at certain locations.

4.2.3 Interviews

Interviews were held with fully personalized interview guides. The interviews are semi-structured to provide enough space for discussion for the interview partner. A first, explorative, interview has been conducted with Stijn Grove, director of the Dutch Datacenter Association. This interview has given insights into the most dominant factors that have contributed to the Dutch datacenter climate.

Most interviews have been held on a non-official basis. Some phrases have been anonymised. Throughout the course of this research the researcher has visited many events where datacenter topics were discussed. These will be discussed in the next paragraph. However, some crucial information has been gained during presentations and general knowledge from within the industry has been put on paper in this research. One anonymous interview is listed below, this person has agreed to have the interview recorded but has later withdrawn his/her approval.

Table 1, interviews

Name Date Organization Appendix

Stijn Grove 10-02-2016

Dutch Datacenter Association (DDA) II

Maikel Bouricius 26-04-2016

Green IT Amsterdam II

Eric van Pelt 27-04-2016

Netherlands Foreign Investment Agency (NFIA)

III

Job WItteman 21-06-2016

Amsterdam Internet Exchange (AMS-IX)

IV

Anonymous - A leading telecom company V

3.2.4 Observations

Observations have been made throughout the course of writing this thesis. Most of the time in informal settings. Conversations at the events have not been recorded. In order to fully

(31)

understand the industry, the researcher has to become part of the industry. Multiple events have been visited to gain a better understanding of the industry. As stated before the researcher is working in the field and has daily contacts with the people working in the industry. The observations are listed below. Short summaries of the content are in the appendices.

Name Date Organization Appendix

Hostingcon Amsterdam Europe

1-10-2015 HostingCon VI

De rol van het lokale datacenter

10-02-2016 EvoSwitch VII

Opening nieuw datacenter (omgeving Haarlem

26-04-2016 Green IT Amsterdam VIII

World Hosting Days 14-03-2016 -18-03-2016

WHD.global IX

Amsterdam als digitale hotspot van Europa

27-04-2016 Netherlands Foreign Investment Agency (NFIA)

X

More-IP Peering into the future 19-05-2016 Amsterdam Internet Exchange (AMS-IX) XI Presentatie mainport voorbij RLT 01-07-2016 RLT XII

Opkomst van het lokale datacenter

22-09-2016 Twin Datacenter Arnhem & ISPConnect XIII Hostingcon Europe Amsterdam 12&13–11-2016 HostingCon XIV Table 2, events

Finally, this research has been presented in a preliminary phase at an event where the rise of the proximity datacenter was discussed. The feedback from the present datacenter professionals has been considered in this research.

3.3 Critical reflection

In optimal situations gathered data is checked by triangulation, getting the same result from three different sources. However, constraints by time and means limit the ability to do that. As discussed in the previous paragraph there is a lack of ‘hard’ facts about the datacenter industry. That research that has been previously conducted is mainly carried out by consultancy agencies that do not share their sources. This lack of data can partially be

(32)

addressed to a lack of understanding at the relevant actors and the recentness of this industry becoming mainstream and relevant.

With hindsight, this research should have been carried out in a later stage. The data that was lacking at first has become available in the final weeks of this research. This rapid change will continue and offers opportunities for researchers. Therefore, the recommendations that are focused on further research are extra valuable. They will be given at the end of this research.

(33)

4. The backbone of the internet, the position of

datacenters within the digital infrastructure

So what exactly is the role of datacenters within the digital infrastructure and even more important; what is the effect of the digital infrastructure on the ‘real’ economy? As stated before the digital infrastructure can be considered as the third mainport of the Netherlands. The effect datacenters have on the economy can, as hypothesized, explain the business climate in the Netherlands. By answering this question, it should be clear what the added value of datacenters is for society and what their functionality is. The outline of this chapter will be as following; first, the role of datacenters within the digital infrastructure will be provided. Second, an exact definition of the principle of ‘mainport’ will be given. Why did the Dutch government decide to make internet infrastructure the third mainport. What implications does this have for policy and the future perspective. The last part of this chapter will give the possible effects datacenters have on the real economy.

4.1The role of the datacenter within the digital infrastructure

In order to understand the effect of a datacenter on the real economy, it is crucial to get clear what role a datacenters plays in the digital infrastructure. The decision by the government to make the digital infrastructure the third main hub does have an effect on datacenters but it might not be direct. “The Dutch Digital Infrastructure is part of a global backbone for

delivering digital services to enterprises and consumers on a variety of devices” (Deloitte,

2014). This digital infrastructure is the backbone of the real economy and plays a part in internal and external business processes. The role of the datacenter within the digital infrastructure can be described as following [ CITATION Sti15 \l 1043 ]:

“Together with internet-exchanges, connectivity providers, hosting and internetservice providers datacenters form the digital infrastructure”

However, datacenters are not all the same. Authors use different distinction criteria to set up a categorization for datacenters. Lesser (2012) makes a distinction between four types of datacenters;

(34)

- scientific computing centers (national laboratories)

- co-location centers (private ‘clouds’ where servers are housed together)

- ‘in-house’ datacenters (facilities owned and operated by company using the servers)

The type of datacenters influences the position within the digital infrastructure. In recent years, there has been a shift towards different types of data storage. If we believe the news articles that are written and the ad campaigns, we see on the internet the in-house datacenters are disappearing and the public cloud providers flourish. This shift towards more cloud-oriented structures will be discussed in paragraph 2.6. Another fascinating thing is happening, we see a rise in the use of local datacenters. The so-called proximity datacenter. Datacenters near the core business facility, preferably in the region. This is not necessarily the cheapest solution but we can certainly speak about ‘the rise of the proximity datacenter’.

Two types of datacenters are of particular interest for this research, the public cloud providers datacenters and the co-location (colo) datacenters. These are chosen for multiple reasons. Public cloud providers build large datacenters, the larger the better, on locations which are the cheapest and well connected, their investment behaviour strongly differs from the other types of datacenter. They have special needs and are responsible for a large part of the infrastructure. Furthermore, they will play a key role in the shift towards the cloud and are relatively footloose when it comes to site selection and are looking to maximize profits and cut costs where possible.

Colo datacenters are multi-tenant and offer infrastructure space for all those in need. Small size businesses, medium size, and enterprise businesses are all welcome to host their infrastructure and put their servers in the datacenter. Colo datacenters offer a different level of services, from just providing space to managed hosting. It is this type of datacenter where we can see the most rapid changes. From simple Infrastructure as a service (IaaS) to managed hosting where the customer has outsourced the entire node of the production network to the datacenter.

(35)

4.2 Datacenter sector; what makes it different from other sectors?

An important part of Porter´s diamond is the structure of the market. Demand conditions are seen as a separate factor and rivalry is part of the more strategic side of the diamond. Specific players (i.e. rivalry) however will not be part of this. By describing common processes for this particular industry an understanding will be created on how it differs from other sectors for which Porter´s diamond is relevant. However, this paragraph will describe the status of the datacenter industry and the differences with other sectors. The datacenter market in Europe evolves mostly around 4 clusters, Frankfurt, London, Amsterdam and Paris (FLAP).

4.3 Digital infrastructure as a mainport

The digital infrastructure of the Netherlands has been selected as the third mainport after Schiphol and Rotterdam. Deloitte (2013) describes a mainport as following:

“First, they have a direct impact on the economy, for instance through the employment of

people working at the container terminals in Rotterdam or at Schiphol airport. In addition, there is also an indirect effect as these mainports attract various economic activities for which the presence of the mainport is conditional. The ecosystem of economic activities related to a mainport is a source of entrepreneurship and creativity.”

By selecting the Dutch infrastructure as the third mainport the Dutch government acknowledges the importance. There will be special conditions linked to this. Innovation will be boosted and there might be more space for pilots and tests financed (partly) by government bodies. This research will try to find out what these effects really are. However, this change in policy is very recent and effects might be part of a longer-term discourse. The policy document by the government states the following:

“samen met relevante belanghebbenden een economische visie te ontwikkelen om de positie van de digitale mainport te versterken; • en te investeren in de uitvoering van die visie”

What this exactly entangles and who those stakeholders are will be analysed thoroughly in the next chapter.

(36)

4.4 The effect oi datacenters on the real economy

Many consultancy bureaus have calculated the potential gains for companies when they move their hardware to a datacenter. However, little research has been done to estimate the effect of the datacenter itself on the economy. Due to multiplier effects this is hard to do. This paragraph will mostly draw from work conducted in Sweden where an extensive analysis has been made of a Facebook datacenter.

Lenio (2015) created an overview of potential gains for the real economy from the construction and operations of a datacenter (shown in table 1 on the next page). Since he did this for the U.S. market there is a slight difference with the Dutch, or European, market. The U.S. version makes a distinction at the state level and local level.

Revenue at the state level: Revenue at the community level:

Sales taxes on construction materials Sales taxes on construction materials Sales/use taxes on equipment purchases Sales/use taxes on equipment purchases Sales taxes or franchise fees on power

consumption

Sales taxes or franchise fees on power consumption

Personal income taxes from construction and permanent jobs

Local income taxes from construction and permanent jobs

Unemployment taxes from construction and permanent

Real estate taxes on a newly constructed or renovated building

Personal property taxes on computer servers and furniture

Table 3 Income generation from datacenters (Lenio, 2015)

These are all short-term, direct effects of the opening of a datacenter. There are also longer term effects which should be accounted for. Keep in mind that datacenters are part of a larger infrastructure and a necessary tool for many businesses in their daily operations. They provide the infrastructure which allows other businesses to develop and therefore play a key role. They can also uplift real estate prices and the surrounding area [ CITATION Sha07 \l 1043 ].

The effect of a datacenter on the real economy has been researched before. Two datacenters have been extensively researched in earlier work. One datacenter in Lenoir

(37)

(North Carolina, USA) and a datacenter in Luleå (Sweden). The results of those case studies will be discussed and analyzed in chapter 5. The analysis will also include a case study from the Netherlands. These two cases are characterized whereas they replaced a declined or declining former industry. In the case of Lenoir the furniture industry [ CITATION Ric07 \l 1043 ] and in the case of Luleå the former metal industry. The consequence of this is a power grid which is useful for datacenters. Nevertheless, the are also other factors which might contribute to the popularity of these two regions. Therefore an analysis of these regions will create a better understanding of this research.

By researching the effect that a datacenter has on the real economy, an analysis can be made on whether or not this is part of the decision-making process of governments to create a policy to attract or retain datacenters. The next chapter will focus on the linkages between datacenters and geographical concepts.

(38)

4.5 Datacenter sector; a rapidly changing industry, a short insight into global processes that shape the datacenter landscape:

30 years ago, there were no colocation datacenters, only in-house datacenters. Nowadays, there is a (some argue rapid) shift towards the cloud. There are multiple forms of data storage in the cloud; public cloud, private cloud, and hybrid cloud. All of them require the storage of data in datacenters or other locations. Important here is that the hybrid cloud requires a constant connection between data in different places. For private cloud, a business only needs a fraction of this connectivity.

Larger amounts of data are transferred every day. Streaming services such as YouTube (Google) and Netflix require a high speed, constant connection. Preferably located as near as possible to the end user. This gives local datacenters the opportunity to become involved, get more direct links to the digital economy and become hyper-relevant. The rise of the proximity datacenters is a key subject throughout this research and will be explained further in the analytical chapters.

Another upcoming issue is the storage of data and the privacy consequences. The storage of data in the cloud is still housed within a datacenter. Governments order businesses to store the data on known locations. The exit of the United Kingdom from the European Union might offer a new opportunity taking this process into account.

(39)

4.6 Conclusion

The main goal of this chapter was to give an insight into the datacenter industry and the role datacenters play as part of the digital economy. The subquestion is:

What (global) processes have led to the development of the datacenter industry and what role do datacenters play in the economy?

Multiple regional and global processes have been discussed. Afterward, several processes that are likely to occur in the future have been highlighted. The effects of datacenters are hard to measure. Their direct effect is not as large as one would expect to look at the investment. Indirect effects, however, should be considered when addressing the business climate and incentives to attract the datacenter industry.

(40)

5. Analyses of the Dutch datacenter business climate

This chapter will describe the Dutch datacenter industry and the actors that play a key role in creating it. How did the datacenter industry become so important for the Netherlands? What is the importance of the organizations that are becoming more mature for the datacenter industry? Furthermore, an analysis will be made of important factors that have or will play a role in the attracting of foreign capital and influence the digital infrastructure.

A distinction can be made between two types of factors that play a role in investment climate of datacenters; generic and sector specific factors. The generic factors are applicable for an entire country or region whereas the sector specific factors only play a role for the datacenter industry. This research will take into account the first type of factors. However, the focus will lie upon the specific factors that apply within a region or country.

5.1 Early days of the datacenter industry

The datacenter industry is relatively new, it only exists for about 30 years. Before it was common to have the IT equipment on premise, at the company site. These so-called in-house datacenters still play a role in the datacenter landscape. Not all businesses have moved their IT infrastructure to a datacenter. Long term processes as the way to the cloud and the internet of things contribute to a shift. However, in order to understand the consequences of these processes, it is necessary to understand the creation of the strong Dutch datacenter climate.

5.1.2 Competition from London

The most important competitor for AMS-IX was the London Internet Exchange LINX. Established in 1994 LINX was the biggest internet exchange in Europe by the number of members and the speed.

‘London was the go-to-market for American datacenters’ (Witteman, personal communication, 21-06-2016). It was perceived as the gateway to Europe. Despite this favorable competitive position the situation has changed and Amsterdam has grown to be one of the leading countries. There are multiple explanations for this. First, the ground price around London is higher. Second, the 2012 Olympic games have an impact on the power supply in and around London. London has been a no-go area for datacenters for this