diff --git a/pkg/sql/logictest/testdata/logic_test/udf_regressions b/pkg/sql/logictest/testdata/logic_test/udf_regressions
index 25bfcb0adc67..892c7154667e 100644
--- a/pkg/sql/logictest/testdata/logic_test/udf_regressions
+++ b/pkg/sql/logictest/testdata/logic_test/udf_regressions
@@ -580,3 +580,48 @@ statement error pgcode 22P02 invalid input value for enum e105259: "foo"
 SELECT f()
 
 subtest end
+
+
+# Regression test for #108297. UDFs with VOID return types should succeed when
+# the last statement returns columns of any type.
+subtest regression_108297
+
+statement ok
+CREATE OR REPLACE FUNCTION f108297() RETURNS VOID LANGUAGE SQL AS 'SELECT 1'
+
+query T
+SELECT f108297()
+----
+NULL
+
+statement ok
+CREATE OR REPLACE FUNCTION f108297() RETURNS VOID LANGUAGE SQL AS $$
+  SELECT 1, 'foo', NULL
+$$
+
+query T
+SELECT f108297()
+----
+NULL
+
+statement ok
+CREATE SEQUENCE s108297
+
+statement ok
+CREATE OR REPLACE FUNCTION f108297() RETURNS VOID LANGUAGE SQL AS $$
+  SELECT nextval('s108297')
+$$
+
+query T
+SELECT f108297()
+----
+NULL
+
+# Invoking the UDF above should have increment s108297 to 1, so calling nextval
+# again should yield 2.
+query I
+SELECT nextval('s108297')
+----
+2
+
+subtest end
diff --git a/pkg/sql/opt/optbuilder/scalar.go b/pkg/sql/opt/optbuilder/scalar.go
index 40ed863989d7..30dfe170a363 100644
--- a/pkg/sql/opt/optbuilder/scalar.go
+++ b/pkg/sql/opt/optbuilder/scalar.go
@@ -24,6 +24,7 @@ import (
 	"github.com/cockroachdb/cockroach/pkg/sql/opt/props"
 	"github.com/cockroachdb/cockroach/pkg/sql/opt/props/physical"
 	"github.com/cockroachdb/cockroach/pkg/sql/parser"
+	"github.com/cockroachdb/cockroach/pkg/sql/parser/statements"
 	"github.com/cockroachdb/cockroach/pkg/sql/pgwire/pgcode"
 	"github.com/cockroachdb/cockroach/pkg/sql/pgwire/pgerror"
 	plpgsql "github.com/cockroachdb/cockroach/pkg/sql/plpgsql/parser"
@@ -728,6 +729,21 @@ func (b *Builder) buildUDF(
 		if err != nil {
 			panic(err)
 		}
+		// Add a VALUES (NULL) statement if the return type of the function is
+		// VOID. We cant simply project NULL from the last statement because all
+		// column would be pruned and the contents of last statement would not
+		// be executed.
+		// TODO(mgartner): This will add some planning overhead for every
+		// invocation of the function. Is there a more efficient way to do this?
+		if rtyp.Family() == types.VoidFamily {
+			stmts = append(stmts, statements.Statement[tree.Statement]{
+				AST: &tree.Select{
+					Select: &tree.ValuesClause{
+						Rows: []tree.Exprs{{tree.DNull}},
+					},
+				},
+			})
+		}
 		body = make([]memo.RelExpr, len(stmts))
 		bodyProps = make([]*physical.Required, len(stmts))
 
@@ -750,6 +766,8 @@ func (b *Builder) buildUDF(
 		if err != nil {
 			panic(err)
 		}
+		// TODO(#108298): Figure out how to handle PLpgSQL functions with VOID
+		// return types.
 		var plBuilder plpgsqlBuilder
 		plBuilder.init(b, colRefs, o.Types.(tree.ParamTypes), stmt.AST, rtyp)
 		stmtScope := plBuilder.build(stmt.AST, bodyScope)
diff --git a/pkg/sql/opt/optbuilder/testdata/udf b/pkg/sql/opt/optbuilder/testdata/udf
index e73a283147f0..3f9cb89abbff 100644
--- a/pkg/sql/opt/optbuilder/testdata/udf
+++ b/pkg/sql/opt/optbuilder/testdata/udf
@@ -1031,6 +1031,47 @@ project
                                └── variable: s:2
 
 
+# --------------------------------------------------
+# UDFs with VOID return type.
+# --------------------------------------------------
+
+exec-ddl
+CREATE FUNCTION retvoid(i INT) RETURNS VOID LANGUAGE SQL AS 'SELECT i'
+----
+
+build format=show-scalars
+SELECT retvoid(1)
+----
+project
+ ├── columns: retvoid:5
+ ├── values
+ │    └── tuple
+ └── projections
+      └── udf: retvoid [as=retvoid:5]
+           ├── args
+           │    └── const: 1
+           ├── params: i:1
+           └── body
+                ├── project
+                │    ├── columns: i:2
+                │    ├── values
+                │    │    └── tuple
+                │    └── projections
+                │         └── variable: i:1 [as=i:2]
+                └── project
+                     ├── columns: column4:4
+                     ├── limit
+                     │    ├── columns: column1:3
+                     │    ├── values
+                     │    │    ├── columns: column1:3
+                     │    │    └── tuple
+                     │    │         └── null
+                     │    └── const: 1
+                     └── projections
+                          └── assignment-cast: VOID [as=column4:4]
+                               └── variable: column1:3
+
+
 # --------------------------------------------------
 # UDFs that are STRICT/RETURNS NULL ON NULL INPUT.
 # --------------------------------------------------
@@ -1699,11 +1740,11 @@ build set=enable_multiple_modifications_of_table=true
 SELECT upd_ups(1, 2, 3)
 ----
 project
- ├── columns: upd_ups:27
+ ├── columns: upd_ups:29
  ├── values
  │    └── ()
  └── projections
-      └── upd_ups(1, 2, 3) [as=upd_ups:27]
+      └── upd_ups(1, 2, 3) [as=upd_ups:29]
 
 exec-ddl
 CREATE FUNCTION ups2(a1 INT, b1 INT, c1 INT, a2 INT, b2 INT, c2 INT) RETURNS BOOL LANGUAGE SQL AS $$