Thread on data centers

Cloud computing, over the top video consumption (Disney+, HBO max, etc), online gaming and connectivity apps are driving massive global demand for data center capacity.
Data centers are locations that house networking, server and storage equipment.

The necessary size and complexity of a data center depends on the type and nature of the computing and storage requirements of the customers that install their equipment inside the data center. Data centers are central locations which consolidate computing resources (servers),

data storage and efficiently manage network transport requirements. Various companies need data centers (media providers like Disney+, Internet service providers, SaaS which delivers software over the Internet, storage content and data companies).

Data center companies provide space, power and cooling to individual customers in a secure individual environment or cage within a larger shared (colocation) data center facility, interconnection (network peering with other customers within a facility), and security and

monitoring services. Data center operators provide the infrastructure (building, power and cooling capacity) for customer equipment to operate in a professionally managed environment but do not typically own any of the server, storage or networking gear that customers install

within the facility.

The infrastructure ecosystem
1)Wireless communication towers transfer signals to smartphones, offering mobile access to online content, while also allowing people to communicate between themselves.
2)5G communication will allow houses located in low-

density zones to benefit from broadband connection.
3)The increase in mobile data supports the establishment of a fibre connection between towers and data centers.
4)The optic fiber is the fastest technology which allows houses to benefit from a constantly growing volume of

online content and for companies to have access to cloud applications.

Data centers can be either managed internally or run by companies specializing in data center management. Internal data centers still make up 60% of total data centers but as data center complexity and power

requirements have increased and existing infrastructure has aged, companies have begun outsourcing their data center needs to 3rd party operators like Equinix/Digital Realty. They operate more efficiently since they can spread overhead costs over many tenants. Colocation in a

data center (DC) is attractive for customers whose DC needs are not great enough to have their own DC (structured as REITs)

Within customers, there’s wholesale/hyperscalers and retail/enterprise. Hyperscalers are large and prefer to manage the equipment within the data center

themselves, only outsourcing management of the data center infrastructure (power and cooling). Wholesale data center operators offer customers the ability to expand their infrastructure in new facilities without having to manage the building or equipment procurement process.

These contracts range 5-15 years, lowering salesforce requirements. Cloud-based AWS, MSFT, GOOGL, FB, CRM account for most of hyperscale demand.

Retail colocation targets customers that need a smaller scale data center footprint and the ability to connect their businesses with

other customers. Contract lengths are shorter here, 2-5 years. Retail colocation providers house many network operators within a facility that offers customers direct access to major global networks.

More on customers connecting with each other: customers don’t need to use the

public Internet by doing so, so they are more safe and have reduced latency & data center operators connect enterprises to each others within data centers. Companies are also attracted to data centers where many networks are connected and networks can also connect and exchange

Internet traffic with each other.

Power cooling and infrastructure and power costs combine to make up almost 40% of the total costs of a data center, so being in an area with low power costs provides a structural cost advantage.

The business model is strong: stable & recurring revenue streams, high switching costs, high incremental margins, low recurring capital intensity, high up-front costs.
On the switching costs: the cost to move one cabinet’s worth of equipment is 10 000$. And monthly recurring

revenue per cabinet is around 2000$, with retail contracts 2-5 years. Even if customers can save 10% of costs, it would take a lot of time of payback just to reimburse the initial moving cost. Plus moving servers critical to operations is risky and can compromise data.

end