search between text

Those steps work for me, though you'll have to adapt with your real data, and it won't scale well to large numbers of rows -  you would need to repeat the "Split by ;" once for each data row:

• Transpose to line up the Product, Country and Price as rows
• Split each non-Price column by ";" (and PQ adds the Changed Type steps)
• Transpose back to Product, Country, Price order
• Fill down the Price to now-empty rows
• Rename columns back to Product, Country, Price

Alternately, you can try something with Table.From Records - e.g. http://stackoverflow.com/questions/31885049/power-query-transform-a-column-by-multiplying-by-another-column

I'm not sure why it's duplicating but, with millions of rows, this method won't work as you need to split for each row (after it's transposed).  Another method came to mind, BUT it's dependent on the maximum number of Products/Countries per line - anything more than 3-7 per line will be unworkable, and it may be slow for millions of lines - can you fix this at the source.  

For what it's worth, for a maximum of three product/country per line, add these lines instead to your Power Query:

    #"Split Product by Delimiter" = Table.SplitColumn(#"Renamed Columns","Product",Splitter.SplitTextByDelimiter(";", QuoteStyle.Csv),{"Product.1", "Product.2", "Product.3"}),
    #"Split Country by Delimiter" = Table.SplitColumn(#"Split Product by Delimiter","Country",Splitter.SplitTextByDelimiter(";", QuoteStyle.Csv),{"Country.1", "Country.2", "Country.3"}),
    #"Renamed Price Column 1" = Table.RenameColumns(#"Split Country by Delimiter",{{"Price", "Price.1"}}),
    #"Duplicated Price Column 2" = Table.DuplicateColumn(#"Renamed Price Column 1", "Price.1", "Price.2"),
    #"Duplicated Price Column 3" = Table.DuplicateColumn(#"Duplicated Price Column 2", "Price.2", "Price.3")
in
    #"Duplicated Price Column 3"

Then REFERENCE that table THREE times to make three derived copies, once for each of (up to) three product/country per line.

In the first copy, keep only columns Product.1, Country.1, Price.1, then rename all columns to drop the .x suffix

In the second copy, keep only columns Product.2, Country.2, Price.2, then rename all columns to drop the .x suffix

In the third copy, keep only columns Product.3, Country.3, Price.3, then rename all columns to drop the .x suffix

Then APPEND all three tables to a New table (they should all have columns Product, Country, Price), and filter out those with "null" Product - these will be rows created for lines with < 3 entries per line.

That only took a few minutes to create but, again, it may not scale to millions of records and if your data has too many variations then consider a function around Table.FromRecords etc. as a next option.