diff --git a/.changeset/easy-points-tan.md b/.changeset/easy-points-tan.md
new file mode 100644
index 0000000000..60edbf94eb
--- /dev/null
+++ b/.changeset/easy-points-tan.md
@@ -0,0 +1,5 @@
+---
+'svelte': patch
+---
+
+perf: speed up parser interactions with Acorn or avoid them where possible
diff --git a/packages/svelte/src/compiler/phases/1-parse/acorn.js b/packages/svelte/src/compiler/phases/1-parse/acorn.js
index 3c3ad2bc8a..add61ea19f 100644
--- a/packages/svelte/src/compiler/phases/1-parse/acorn.js
+++ b/packages/svelte/src/compiler/phases/1-parse/acorn.js
@@ -142,17 +142,24 @@ let last_template = '';
 let lf_only = true;
 
 /**
- * Without `startLocation`, acorn counts the lines before `index` on every call.
- * It also breaks lines on bare `\r`, `\u2028` and `\u2029`, which the locator doesn't, so those templates are left to acorn
+ * Without `startLocation`, acorn counts the lines before `index` on every call
  * @param {Parser} parser
  * @param {number} index
  */
 function start_location(parser, index) {
+	return has_lf_line_breaks_only(parser) ? locator(index) : undefined;
+}
+
+/**
+ * acorn breaks lines on bare `\r`, `\u2028` and `\u2029`, which the locator doesn't
+ * @param {Parser} parser
+ */
+export function has_lf_line_breaks_only(parser) {
 	if (parser.template !== last_template) {
 		last_template = parser.template;
 		lf_only = !regex_non_lf_line_break.test(last_template);
 	}
-	return lf_only ? locator(index) : undefined;
+	return lf_only;
 }
 
 const regex_position_indicator = / \(\d+:\d+\)$/;
diff --git a/packages/svelte/src/compiler/phases/1-parse/read/expression.js b/packages/svelte/src/compiler/phases/1-parse/read/expression.js
index 1c8f097c2f..b63ad443a8 100644
--- a/packages/svelte/src/compiler/phases/1-parse/read/expression.js
+++ b/packages/svelte/src/compiler/phases/1-parse/read/expression.js
@@ -1,9 +1,13 @@
-/** @import { Expression } from 'estree' */
+/** @import { Expression, Identifier } from 'estree' */
 /** @import { Parser } from '../index.js' */
-import { parse_expression_at, remove_parens } from '../acorn.js';
+// @ts-expect-error acorn type definitions are borked in the release we use
+import { isIdentifierStart, isIdentifierChar } from 'acorn';
+import { has_lf_line_breaks_only, parse_expression_at, remove_parens } from '../acorn.js';
 import { regex_whitespace } from '../../patterns.js';
 import * as e from '../../../errors.js';
 import { find_matching_bracket } from '../utils/bracket.js';
+import { is_reserved } from '../../../../utils.js';
+import { locator } from '../../../state.js';
 
 /**
  * @param {Parser} parser
@@ -33,6 +37,9 @@ export function get_loose_identifier(parser, opening_token) {
  * @returns {Expression}
  */
 export default function read_expression(parser, opening_token, disallow_loose) {
+	const simple = read_simple_expression(parser);
+	if (simple) return simple;
+
 	try {
 		const node = parse_expression_at(parser, parser.template, parser.index);
 
@@ -57,3 +64,104 @@ export default function read_expression(parser, opening_token, disallow_loose) {
 		throw err;
 	}
 }
+
+/**
+ * Most template expressions are an identifier or a `a.b.c` member chain followed by `}`.
+ * Those are built directly for better parse performance, with the same shape acorn would produce; anything else goes to acorn
+ * @param {Parser} parser
+ * @returns {Expression | null}
+ */
+function read_simple_expression(parser) {
+	if (!has_lf_line_breaks_only(parser)) return null;
+
+	const template = parser.template;
+	const index = parser.index;
+
+	parser.allow_whitespace();
+	const start = parser.index;
+
+	let end = read_word(template, start);
+	if (end === -1 || is_reserved(template.slice(start, end))) {
+		parser.index = index;
+		return null;
+	}
+
+	/** @type {Expression} */
+	let node = identifier(template, start, end);
+
+	while (template[end] === '.') {
+		const property_end = read_word(template, end + 1);
+		if (property_end === -1) {
+			parser.index = index;
+			return null;
+		}
+
+		node = {
+			type: 'MemberExpression',
+			start,
+			end: property_end,
+			loc: { start: position(start), end: position(property_end) },
+			object: node,
+			property: identifier(template, end + 1, property_end),
+			computed: false,
+			optional: false
+		};
+
+		end = property_end;
+	}
+
+	parser.index = end;
+	parser.allow_whitespace();
+
+	if (!parser.match('}')) {
+		parser.index = index;
+		return null;
+	}
+
+	parser.index = end;
+	return node;
+}
+
+/**
+ * @param {string} template
+ * @param {number} start
+ * @returns {number} the end of the identifier starting at `start`, or -1
+ */
+function read_word(template, start) {
+	if (start >= template.length) return -1;
+
+	const code = /** @type {number} */ (template.codePointAt(start));
+	if (!isIdentifierStart(code, true)) return -1;
+
+	let end = start + (code <= 0xffff ? 1 : 2);
+
+	while (end < template.length) {
+		const code = /** @type {number} */ (template.codePointAt(end));
+		if (!isIdentifierChar(code, true)) break;
+		end += code <= 0xffff ? 1 : 2;
+	}
+
+	return end;
+}
+
+/**
+ * @param {string} template
+ * @param {number} start
+ * @param {number} end
+ * @returns {Identifier}
+ */
+function identifier(template, start, end) {
+	return {
+		type: 'Identifier',
+		start,
+		end,
+		loc: { start: position(start), end: position(end) },
+		name: template.slice(start, end)
+	};
+}
+
+/** @param {number} index */
+function position(index) {
+	const { line, column } = locator(index);
+	return { line, column };
+}