You can allow the ROW_NUMBER window function to craft a ranking value for each of your items, ordered by usage. Then make TOP(1) WITH TIES filter values with respect to the ORDER BY clause, getting only the first value tied (row number = 1, or if you will, the greatest usage value for each item).

SELECT TOP(1) WITH TIES *
FROM tab 
ORDER BY ROW_NUMBER() OVER(PARTITION BY [Item] ORDER BY [Usage] DESC)

Equivalently, you can compute the ROW_NUMBER window function first, then select all rows that have ranking = 1 in a subquery.

WITH cte AS (
    SELECT *, 
           ROW_NUMBER() OVER(PARTITION BY [Item] ORDER BY [Usage] DESC) AS rn
    FROM tab 
)
SELECT [Item], [Usage], [Customer]
FROM cte 
WHERE rn = 1

Output:

Check the demo here .

Consider learning more about these SQL-Server tools at the corresponding linked web pages.

I have used data provided by you to create a SQLFIDDLE.
I want to only show the rows with the max usage of each item, to show the customer that used the most of the item.

This can be achieved as

select t1.item, t1.usage, t1.customer
from table1 t1
inner join (
  select item, MAX(usage) as maximum_usage
  from table1
  group by item
) t2 on t1.usage = t2.maximum_usage;

This gives me the expected output as :

You can see the demo in SQLFIDDLE