This can be done with some limitations using pgfplotstable's \pgfplotstablecreatecol macro (also with create on use to lazily create the column data, but its contents would then be lost after returning from \pgfplotstabletypeset). According to the pgfplotstable documentation:
Currently, you can only access three values of one column at a time: the current row, the previous row and the next row. Access to arbitrary indices is not (yet) supported.
In the example below, I implemented the formula described in the question, initializing it with (arbitrary) value 100 in “row -1.” Since the initial table data is:
x y
0 1
5 6
10 11
the computed values are:
1 + 100 = 101
6 + 101 = 107
11 + 107 = 118
In order to access previously-computed values in the column being dynamically created, I store them globally (here: only the most recently computed value, using \xdef\myPreviousValue{...}) because pgfplotstable's \prevrow macro doesn't give access to values from the column being created, as it seems. If access to any previously-computed value in the column being created is desired, one could use a pgfmath array1 or an expl3 tl or seq variable, for instance.
\begin{filecontents*}{data.csv}
x y
0 1
5 6
10 11
\end{filecontents*}
\documentclass{article}
\usepackage{booktabs}
\usepackage{pgfplotstable}
\pgfplotsset{compat=1.16}
\pgfplotstableread[row sep=newline, col sep=space]{data.csv}\myTable
\newcommand*{\myPreviousValue}{100} % initialization (row -1, sort of)
% Dynamically create column z
\pgfplotstablecreatecol[
create col/assign/.code={%
\pgfmathsetmacro{\myValue}{int(\thisrow{y} + \myPreviousValue)}%
\pgfplotstableset{create col/next content/.expand once={\myValue}}%
\xdef\myPreviousValue{\myValue}%
}]
{z}\myTable
\begin{document}
\pgfplotstabletypeset[
columns/x/.style={column name={$x$}},
columns/y/.style={column name={$y$}},
columns/z/.style={column name={$z$}},
every head row/.style={before row=\toprule, after row=\midrule},
every last row/.style={after row=\bottomrule}
]{\myTable}
\end{document}

Note: the int() in the pgfmath expression I used might appear as unnecessary because by default, \pgfplotstabletypeset formats values using \pgfmathprintnumber and, again by default, \pgfmathprintnumber detects if the input has a fractional part equal to zero in order to special-case the printing of integers. For instance, \pgfmathprintnumber{118.0} prints 118 by default. However, without the int(), the values stored in \myValue—and thus also in the created in-memory column and in \myPreviousValue—would have a trailing .0; using the int() function prevents this.
Footnote
- Search the TikZ & PGF manual for “array access operators” in the Mathematical and Object-Oriented Engines part.
datatoolpackage, see the manual from page 76 onwards. Note that this is different frompgfplotstable. – Marijn Mar 20 '20 at 12:46\DTLforeach*{values}{\x=x,\y=y}{ \addplot coordinates {(\x,\y)}; }? – atticus Mar 20 '20 at 13:52