Also, don't forget that many companies started with it as part of the Business Server suite and so all their software was built around it for little cost of the database itself. Then they grew into it.
I have spent a long time working in and around it, and respect it overall. Personally, I have moved a lot of clients out of SQL Server into more cost effective solutions when they have grown to a point that the licensing cost will outweigh the need to re-architect their system. That usually happens when they realize they need robust replication or a distributed solution which will take some architecture work anyway, and why pay MSFT huge licensing fees and architect for it when you can move to say Postgres and save the licensing fees, just as an example.
Enterprises use it because they sign contracts with Microsoft to use all Microsoft products, and then they get support from Microsoft when things go wrong.
Furthermore, "no one ever got fired for hiring IBM" (not sure who said that originally). Basically, when you use a big-name, enterprise vendor and the software fails, it's their fault. When you use a (F)OSS solution and the software fails, it's your fault.
You can go on and on with technology that sucks that's still used by large enterprise for the same reason.
As mentioned, you get core SQL Server, along with a number of integrated tools that range from adequate to best in class:
* SSIS for ETL
* SSRS for reporting - includes traditional paginated reports, Power View for visual analysis and exploration (in 2016 gets native mobile apps and major overhaul)
* SSAS - Multidimensional for traditional OLAP, data mining; Tabular for in-memory columnstore
* CLR - everything has access to CLR languages for customizability and scripting
* Azure - easy transition from on-prem to cloud either with IaaS and pre-configured DB VMs or Azure SQL and Azure DW for SaaS with syntax and feature compatibility; very strong hybrid infrastructure story
Features All of these are native
* In-memory OLTP - memory optimized tables, optionally transactionally persisted to disk or not, with sprocs compiled to native code
* Columnstore index - optionally memory-optimized tables with clustered columnstore for hybrid OLTP / OLAP workload
* HA groups for native replication and failover story
* Native R integration (coming SQL Server 2016)
* Integrated row-level security
* Encryption for data at rest and in transit (more of an Azure SQL piece)
* Native master data and data quality services
I don't know if you care much (but CIOs and IT directors of large organizations do), but SQL Server 2014 is the absolute leader in Gartner's magic quadrant for mission-critical application development, and Microsoft is in a leadership position for BI as well.
Believe it or not, TCO is one of the largest talking points for Microsoft data platform account executives and solution sellers. Especially when placed against Oracle that story is strong. We have customers (we being the company I work for - BI and analytics consultancy) coming to SQL Server from every platform you can imagine. There are also many customers leaving SQL Server for every other platform. At this point in time Microsoft is doing a pretty good job of making sure that first number is larger than the second by a healthy margin.
A lot of it is inertia, as well. People who have been writing software for MS SQL server for the past couple of decades probably aren't going to change any time soon.
Plus with MS Azure, MS SQL server is easy for a non DBA to setup, add failover and scale.
Also, if you're a Microsoft shop, it's easier to deal with than learning Postgres's unixy ways. And you can program sprocs using .Net
Why does Oracle, DB2, Sybase, or any other commercial RDBMS still exist?
* Different capabilities
* Different levels of support
* Previous lock-in
* Non-technical reasons