diff --git a/src/dotenv/env_loader.rs b/src/dotenv/env_loader.rs
index 50a321b0d8f7..931b5699dc65 100644
--- a/src/dotenv/env_loader.rs
+++ b/src/dotenv/env_loader.rs
@@ -994,7 +994,9 @@ struct Parser<'a> {
     value_buffer: &'a mut Vec<u8>,
 }
 
-const WHITESPACE_CHARS: &[u8] = b"\t\x0B\x0C \xA0\n\r";
+// Input is UTF-8, so this set must be ASCII-only: 0xA0 is a continuation byte
+// there (NBSP is C2 A0), and trimming it byte-wise corrupts multi-byte sequences.
+const WHITESPACE_CHARS: &[u8] = b"\t\x0B\x0C \n\r";
 
 impl<'a> Parser<'a> {
     fn skip_line(&mut self) {
diff --git a/test/cli/run/env.test.ts b/test/cli/run/env.test.ts
index 23038a50c7ba..ac7362ab30f2 100644
--- a/test/cli/run/env.test.ts
+++ b/test/cli/run/env.test.ts
@@ -12,6 +12,7 @@ import {
   isWindows,
   tempDirWithFiles,
 } from "harness";
+import { parseEnv } from "node:util";
 import path from "path";
 
 function bunRunWithoutTrim(file: string, env?: Record<string, string>) {
@@ -372,6 +373,24 @@ test(".env space edgecase (issue #411)", () => {
   expect(stdout).toBe("[A B]");
 });
 
+test(".env does not byte-trim 0xA0 out of UTF-8 values", () => {
+  // U+0920 DEVANAGARI LETTER TTHA encodes as E0 A4 A0 (trailing 0xA0)
+  // U+00A0 NO-BREAK SPACE encodes as C2 A0; Node.js preserves it verbatim.
+  expect(parseEnv("A=x\u0920\nB=\u00A0x\u00A0\nC=\u00A0\nD=  x  \n")).toEqual({
+    A: "x\u0920",
+    B: "\u00A0x\u00A0",
+    C: "\u00A0",
+    D: "x",
+  });
+
+  const dir = tempDirWithFiles("dotenv-utf8-nbsp", {
+    ".env": "A=x\u0920\nB=\u00A0x\u00A0\n",
+    "index.ts": "console.log(JSON.stringify({ A: process.env.A, B: process.env.B }));",
+  });
+  const { stdout } = bunRun(`${dir}/index.ts`);
+  expect(JSON.parse(stdout)).toEqual({ A: "x\u0920", B: "\u00A0x\u00A0" });
+});
+
 test(".env special characters 1 (issue #2823)", () => {
   const dir = tempDirWithFiles("dotenv-issue-2823", {
     ".env": 'A="a$t"\nC=`c\\$v`',