diff --git a/packages/compiler/src/ml_parser/html_whitespaces.ts b/packages/compiler/src/ml_parser/html_whitespaces.ts
index f2cd01f7e3..73e5b48665 100644
--- a/packages/compiler/src/ml_parser/html_whitespaces.ts
+++ b/packages/compiler/src/ml_parser/html_whitespaces.ts
@@ -18,6 +18,17 @@ function hasPreserveWhitespacesAttr(attrs: html.Attribute[]): boolean {
   return attrs.some((attr: html.Attribute) => attr.name === PRESERVE_WS_ATTR_NAME);
 }
 
+/**
+ * Angular Dart introduced &ngsp; as a placeholder for non-removable space, see:
+ * https://github.com/dart-lang/angular/blob/0bb611387d29d65b5af7f9d2515ab571fd3fbee4/_tests/test/compiler/preserve_whitespace_test.dart#L25-L32
+ * In Angular Dart &ngsp; is converted to the 0xE500 PUA (Private Use Areas) unicode character
+ * and later on replaced by a space. We are re-implementing the same idea here.
+ */
+export function replaceNgsp(value: string): string {
+  // lexer is replacing the &ngsp; pseudo-entity with NGSP_UNICODE
+  return value.replace(new RegExp(NGSP_UNICODE, 'g'), ' ');
+}
+
 /**
  * This visitor can walk HTML parse tree and remove / trim text nodes using the following rules:
  * - consider spaces, tabs and new lines as whitespace characters;
@@ -25,15 +36,9 @@ function hasPreserveWhitespacesAttr(attrs: html.Attribute[]): boolean {
  * - for all other text nodes replace consecutive whitespace characters with one space;
  * - convert &ngsp; pseudo-entity to a single space;
  *
- * The idea of using &ngsp; as a placeholder for non-removable space was originally introduced in
- * Angular Dart, see:
- * https://github.com/dart-lang/angular/blob/0bb611387d29d65b5af7f9d2515ab571fd3fbee4/_tests/test/compiler/preserve_whitespace_test.dart#L25-L32
- * In Angular Dart &ngsp; is converted to the 0xE500 PUA (Private Use Areas) unicode character
- * and later on replaced by a space. We are re-implementing the same idea here.
- *
  * Removal and trimming of whitespaces have positive performance impact (less code to generate
  * while compiling templates, faster view creation). At the same time it can be "destructive"
- * in some cases (whitespaces can influence layout). Becouse of the potential of breaking layout
+ * in some cases (whitespaces can influence layout). Because of the potential of breaking layout
  * this visitor is not activated by default in Angular 5 and people need to explicitly opt-in for
  * whitespace removal. The default option for whitespace removal will be revisited in Angular 6
  * and might be changed to "on" by default.
@@ -61,9 +66,7 @@ class WhitespaceVisitor implements html.Visitor {
     const isBlank = text.value.trim().length === 0;
 
     if (!isBlank) {
-      // lexer is replacing the &ngsp; pseudo-entity with NGSP_UNICODE
-      return new html.Text(
-          text.value.replace(NGSP_UNICODE, ' ').replace(/\s\s+/g, ' '), text.sourceSpan);
+      return new html.Text(replaceNgsp(text.value).replace(/\s\s+/g, ' '), text.sourceSpan);
     }
 
     return null;
diff --git a/packages/compiler/src/template_parser/template_parser.ts b/packages/compiler/src/template_parser/template_parser.ts
index 255e048ca1..c3b53737e8 100644
--- a/packages/compiler/src/template_parser/template_parser.ts
+++ b/packages/compiler/src/template_parser/template_parser.ts
@@ -16,7 +16,7 @@ import {I18NHtmlParser} from '../i18n/i18n_html_parser';
 import {Identifiers, createTokenForExternalReference, createTokenForReference} from '../identifiers';
 import * as html from '../ml_parser/ast';
 import {ParseTreeResult} from '../ml_parser/html_parser';
-import {removeWhitespaces} from '../ml_parser/html_whitespaces';
+import {removeWhitespaces, replaceNgsp} from '../ml_parser/html_whitespaces';
 import {expandNodes} from '../ml_parser/icu_ast_expander';
 import {InterpolationConfig} from '../ml_parser/interpolation_config';
 import {isNgTemplate, splitNsName} from '../ml_parser/tags';
@@ -248,9 +248,10 @@ class TemplateParseVisitor implements html.Visitor {
 
   visitText(text: html.Text, parent: ElementContext): any {
     const ngContentIndex = parent.findNgContentIndex(TEXT_CSS_SELECTOR) !;
-    const expr = this._bindingParser.parseInterpolation(text.value, text.sourceSpan !);
+    const valueNoNgsp = replaceNgsp(text.value);
+    const expr = this._bindingParser.parseInterpolation(valueNoNgsp, text.sourceSpan !);
     return expr ? new BoundTextAst(expr, ngContentIndex, text.sourceSpan !) :
-                  new TextAst(text.value, ngContentIndex, text.sourceSpan !);
+                  new TextAst(valueNoNgsp, ngContentIndex, text.sourceSpan !);
   }
 
   visitAttribute(attribute: html.Attribute, context: any): any {
diff --git a/packages/compiler/test/template_parser/template_parser_spec.ts b/packages/compiler/test/template_parser/template_parser_spec.ts
index a6371c5b1f..f5142322e5 100644
--- a/packages/compiler/test/template_parser/template_parser_spec.ts
+++ b/packages/compiler/test/template_parser/template_parser_spec.ts
@@ -2071,6 +2071,12 @@ The pipe 'test' could not be found ("{{[ERROR ->]a | test}}"): TestComp@0:2`);
 
   describe('whitespaces removal', () => {
 
+    beforeEach(() => {
+      TestBed.configureCompiler({providers: [TEST_COMPILER_PROVIDERS, MOCK_SCHEMA_REGISTRY]});
+    });
+
+    commonBeforeEach();
+
     it('should not remove whitespaces by default', () => {
       expect(humanizeTplAst(parse(' <br>  <br>\t<br>\n<br> ', []))).toEqual([
         [TextAst, ' '],
@@ -2085,6 +2091,18 @@ The pipe 'test' could not be found ("{{[ERROR ->]a | test}}"): TestComp@0:2`);
       ]);
     });
 
+    it('should replace each &ngsp; with a space when preserveWhitespaces is true', () => {
+      expect(humanizeTplAst(parse('foo&ngsp;&ngsp;&ngsp;bar', [], [], [], true))).toEqual([
+        [TextAst, 'foo   bar'],
+      ]);
+    });
+
+    it('should replace every &ngsp; with a single space when preserveWhitespaces is false', () => {
+      expect(humanizeTplAst(parse('foo&ngsp;&ngsp;&ngsp;bar', [], [], [], false))).toEqual([
+        [TextAst, 'foo bar'],
+      ]);
+    });
+
     it('should remove whitespaces when explicitly requested', () => {
       expect(humanizeTplAst(parse(' <br>  <br>\t<br>\n<br> ', [], [], [], false))).toEqual([
         [ElementAst, 'br'],