Many researchers use their own hardware, whether personally owned or company/university owned but for private use. This provides the most flexibility and also the most responsibility and personal time investment in managing the system.

If your organization manages the computer for you, you will spend less time on I.T. and have more time for research. The down side is you cannot always get what you want installed on the machine and if you can, you may have to wait.

The advantages to managing your own hardware and software can be significant, but only if you know how to do it well. Many researchers don't know how to choose and secure an operating system, fail to keep up with security updates, and rely mainly on caveman installs for their scientific software. Most such people would be better off letting I.T. manage their machines.

If you are diligent about keeping your system secure and up-to-date and manage your software via package managers, you may find yourself with a significant edge in the race for research funds.

Most colleges and universities maintain computer labs with software to serve the needs of their students. Check with your instructors or department office to find out what's available to you.

Some campuses may also have clusters and grids available for parallel computing. If you need to run large simulations, parameter sweeps, or Monte Carlo simulations, it may be possible to run hundreds at a time instead of one at a time on your PC or a stand-alone lab PC.

The National Science Foundation funds several very large clusters on campuses around the country for general use by researchers on other campuses.

Use of these resources is free for academic researchers. Small allocations of computing time are easily obtained, while larger allocations require a more extensive proposal.

A number of commercial services are also available for those who have the ability to pay as they go.

Amazon EC2, Google Cloud Platform, Azure, and other cloud computing services allow researchers to create their own custom virtual machines and even virtual clusters. Users pay for CPU time used.

Users of these services can quickly configure virtual machines with a wide variety of configurations such as the number of CPUs, the type of CPU, the amount of RAM, and the amount and type of storage. The cost is typically a fraction of a penny per CPU-hour.

The problem for academic researchers is that funding is usually fixed by research grants, so the unpredictable pay-as-you-go model can be problematic. There may also be significant bureaucracy involved in paying the fees through department channels.