In a strongly-typed language like C#, the compiler makes no assumptions about what you intend to do with your data. You must make your intent explicit through your code. Something like this should work:

string csvData = "A,B;C,D";
string[][] sep = csvData.Split(';') // Returns string[] {"A,B","C,D"}
  .Select(str => str.Split(',')) // Returns IEnumerable<string[]> {{"A","B"},{"C","D"}}
  .ToArray(); // Returns string[][] {{"A","B"},{"C","D"}}

Rows are separated by semicolon, columns by comma?

Splitting by ';' gives you an array of rows. Split a row by ',' gives you an array of values.

If your data has a consistent schema, as in each csv you process has the same columns, you could define a class to represent the entity to make the data easier to with with.

Let's say it's customer data:

John,Smith,8675309,johnsmith@gmail.com;

You could make a class with those properties:

public class Customer
{
    public string FirstName { get; set; }
    public string LastName { get; set; }
    public string Phone { get; set; }
    public string Email { get; set; }
}

Then:

var rows = csvdata.Split(';');
List<Customer> customers = new();
foreach(var row in rows)
{
    var customer = row.Split(',');
    customers.Add(new()
    {
        FirstName = row[0],
        LastName = row[1],
        Phone = row[2],
        Email = row[3]
    });
}

Now you have a list of customers to do whatever it is you do with customers.

Here is an answer to present a few alternative ideas and things you can do with C# - more for educational/academic purposes than anything else. These days to consume a CSV we'd use a CSV library

If your data is definitely regularly formed you can get away with just one Split. The following code splits on either char to make one long array. It then stands to reason that every 4 elements is a new customer, the data of the customer being given by n+0, n+1, n+2 and n+3. Because we know how many data items we will consume, dividing it by 4 gives us the number of customers so we can presize our 2D array

var bits = data.Split(';',',');
var twoD = new string[bits.Length/4,4];
for(int x = 0; x < bits.Length; x+=4){
    twoD[x/4,0] = bits[x+0];
    twoD[x/4,1] = bits[x+1];
    twoD[x/4,2] = bits[x+2];
    twoD[x/4,3] = bits[x+3];
}

I don't think I'd use 2D arrays though - and I commend the other answer advising to create a class to hold the related data; you can use this same technique

var custs = new List<Customer>();
for(int x = 0; x < bits.Length;){
    custs.Add(new()
    {
        FirstName = bits[x++],
        LastName = bits[x++],
        Phone = bits[x++],
        Email = bits[x++]
    });
}

Here we aren't incrementing x in the loop header; every time a bit of info is assigned x is bumped up by 1 in the loop body. We could have kept the same approach as before, jumping it by 4 - just demoing another approach that lends itself well here.

I mentioned that these days we probably wouldn't really read a csv manually and split ourselves - what if the data contains a comma, or a semicolon - it wrecks the file structure

There are a boatload of libraries that read CSV files, CsvHelper is a popular one, and you'd use it like:

using var reader = new StreamReader("path\\to\\file.csv");
using var csv = new CsvReader(reader, CultureInfo.InvariantCulture)

var custs = csv.GetRecords<Customer>().ToList();

...

Your file would have a header line with column names that match your property names in c#. If it doesn't then you can use attributes on the properties to tell CsvH what column should be mapped to what property - https://joshclose.github.io/CsvHelper/getting-started/

Here's the simplest way I know to produce a 2d array by splitting a string.

string csvData = "A,B,C;D,E,F,G";

var temporary =
    csvData
        .Split(';')
        .SelectMany((xs, i) => xs.Split(',').Select((x, j) => new { x, i, j }))
        .ToArray();
        
int max_i = temporary.Max(x => x.i);
int max_j = temporary.Max(x => x.j);

string[,] array = new string[max_i + 1, max_j + 1];

foreach (var t in temporary)
{
    array[t.i, t.j] = t.x;
}

I purposely chose csvData to be missing a value.

temporary is this:

And the final array is this: