diff --git a/benchmarks/eventsource/eventsource-stream.mjs b/benchmarks/eventsource/eventsource-stream.mjs
new file mode 100644
index 00000000000..7d0e5d80685
--- /dev/null
+++ b/benchmarks/eventsource/eventsource-stream.mjs
@@ -0,0 +1,81 @@
+import { bench, group, run } from 'mitata'
+import { EventSourceStream } from '../../lib/web/eventsource/eventsource-stream.js'
+
+const EVENT_COUNT = 250
+
+function buildEventStream (count) {
+  let content = ''
+
+  for (let i = 0; i < count; i++) {
+    content += `id: ${i}\n`
+    content += `event: message-${i % 4}\n`
+    content += `retry: ${1000 + (i % 5)}\n`
+    content += `data: ${'x'.repeat(64)}-${i}\n`
+    content += `data: ${'y'.repeat(64)}-${i}\n\n`
+  }
+
+  return Buffer.from(content, 'utf8')
+}
+
+function splitBuffer (buffer, chunkSize) {
+  const chunks = []
+
+  for (let i = 0; i < buffer.length; i += chunkSize) {
+    chunks.push(buffer.subarray(i, Math.min(i + chunkSize, buffer.length)))
+  }
+
+  return chunks
+}
+
+function parseChunks (chunks, expectedEvents) {
+  let events = 0
+
+  const stream = new EventSourceStream({
+    push () {
+      events++
+      return true
+    }
+  })
+
+  for (let i = 0; i < chunks.length; i++) {
+    stream.write(chunks[i])
+  }
+
+  if (events !== expectedEvents) {
+    throw new Error(`Expected ${expectedEvents} events, got ${events}`)
+  }
+}
+
+const payload = buildEventStream(EVENT_COUNT)
+const payloadWithBOM = Buffer.concat([Buffer.from('\uFEFF', 'utf8'), payload])
+
+const scenarios = {
+  'single chunk': splitBuffer(payload, payload.length),
+  '256-byte chunks': splitBuffer(payload, 256),
+  '64-byte chunks': splitBuffer(payload, 64),
+  '8-byte chunks': splitBuffer(payload, 8),
+  '1-byte chunks': splitBuffer(payload, 1)
+}
+
+const bomScenarios = {
+  'BOM + 8-byte chunks': splitBuffer(payloadWithBOM, 8),
+  'BOM + 1-byte chunks': splitBuffer(payloadWithBOM, 1)
+}
+
+group(`EventSourceStream parsing (${EVENT_COUNT} events)`, () => {
+  for (const [name, chunks] of Object.entries(scenarios)) {
+    bench(name, () => {
+      parseChunks(chunks, EVENT_COUNT)
+    })
+  }
+})
+
+group('EventSourceStream parsing with BOM', () => {
+  for (const [name, chunks] of Object.entries(bomScenarios)) {
+    bench(name, () => {
+      parseChunks(chunks, EVENT_COUNT)
+    })
+  }
+})
+
+await run()
diff --git a/lib/web/eventsource/eventsource-stream.js b/lib/web/eventsource/eventsource-stream.js
index d24e8f6a1b1..7b9e2f8cbba 100644
--- a/lib/web/eventsource/eventsource-stream.js
+++ b/lib/web/eventsource/eventsource-stream.js
@@ -23,6 +23,49 @@ const COLON = 0x3A
  */
 const SPACE = 0x20
 
+const DATA = Buffer.from('data')
+const EVENT = Buffer.from('event')
+const ID = Buffer.from('id')
+const RETRY = Buffer.from('retry')
+
+function isASCIINumberBytes (buffer, start) {
+  if (start >= buffer.length) {
+    return false
+  }
+
+  for (let i = start; i < buffer.length; i++) {
+    if (buffer[i] < 0x30 || buffer[i] > 0x39) {
+      return false
+    }
+  }
+
+  return true
+}
+
+function isValidLastEventIdBytes (buffer, start) {
+  for (let i = start; i < buffer.length; i++) {
+    if (buffer[i] === 0x00) {
+      return false
+    }
+  }
+
+  return true
+}
+
+function isFieldName (line, length, field) {
+  if (length !== field.length) {
+    return false
+  }
+
+  for (let i = 0; i < length; i++) {
+    if (line[i] !== field[i]) {
+      return false
+    }
+  }
+
+  return true
+}
+
 /**
  * @typedef {object} EventSourceStreamEvent
  * @type {object}
@@ -63,11 +106,14 @@ class EventSourceStream extends Transform {
   eventEndCheck = false
 
   /**
-   * @type {Buffer|null}
+   * @type {Buffer[]}
    */
-  buffer = null
+  chunks = []
 
+  chunkIndex = 0
   pos = 0
+  lineChunkIndex = 0
+  linePos = 0
 
   event = {
     data: undefined,
@@ -107,92 +153,20 @@ class EventSourceStream extends Transform {
       return
     }
 
-    // Cache the chunk in the buffer, as the data might not be complete while
-    // processing it
-    // TODO: Investigate if there is a more performant way to handle
-    // incoming chunks
-    // see: https://github.com/nodejs/undici/issues/2630
-    if (this.buffer) {
-      this.buffer = Buffer.concat([this.buffer, chunk])
-    } else {
-      this.buffer = chunk
-    }
+    this.chunks.push(chunk)
 
     // Strip leading byte-order-mark if we opened the stream and started
     // the processing of the incoming data
     if (this.checkBOM) {
-      switch (this.buffer.length) {
-        case 1:
-          // Check if the first byte is the same as the first byte of the BOM
-          if (this.buffer[0] === BOM[0]) {
-            // If it is, we need to wait for more data
-            callback()
-            return
-          }
-          // Set the checkBOM flag to false as we don't need to check for the
-          // BOM anymore
-          this.checkBOM = false
-
-          // The buffer only contains one byte so we need to wait for more data
-          callback()
-          return
-        case 2:
-          // Check if the first two bytes are the same as the first two bytes
-          // of the BOM
-          if (
-            this.buffer[0] === BOM[0] &&
-            this.buffer[1] === BOM[1]
-          ) {
-            // If it is, we need to wait for more data, because the third byte
-            // is needed to determine if it is the BOM or not
-            callback()
-            return
-          }
-
-          // Set the checkBOM flag to false as we don't need to check for the
-          // BOM anymore
-          this.checkBOM = false
-          break
-        case 3:
-          // Check if the first three bytes are the same as the first three
-          // bytes of the BOM
-          if (
-            this.buffer[0] === BOM[0] &&
-            this.buffer[1] === BOM[1] &&
-            this.buffer[2] === BOM[2]
-          ) {
-            // If it is, we can drop the buffered data, as it is only the BOM
-            this.buffer = Buffer.alloc(0)
-            // Set the checkBOM flag to false as we don't need to check for the
-            // BOM anymore
-            this.checkBOM = false
-
-            // Await more data
-            callback()
-            return
-          }
-          // If it is not the BOM, we can start processing the data
-          this.checkBOM = false
-          break
-        default:
-          // The buffer is longer than 3 bytes, so we can drop the BOM if it is
-          // present
-          if (
-            this.buffer[0] === BOM[0] &&
-            this.buffer[1] === BOM[1] &&
-            this.buffer[2] === BOM[2]
-          ) {
-            // Remove the BOM from the buffer
-            this.buffer = this.buffer.subarray(3)
-          }
-
-          // Set the checkBOM flag to false as we don't need to check for the
-          this.checkBOM = false
-          break
+      if (this.handleBOM()) {
+        callback()
+        return
       }
     }
 
-    while (this.pos < this.buffer.length) {
+    while (this.hasCurrentByte()) {
+      const byte = this.currentByte()
+
       // If the previous line ended with an end-of-line, we need to check
       // if the next character is also an end-of-line.
       if (this.eventEndCheck) {
@@ -205,10 +179,9 @@ class EventSourceStream extends Transform {
         if (this.crlfCheck) {
           // If the current character is a line feed, we can remove it
           // from the buffer and reset the crlfCheck flag
-          if (this.buffer[this.pos] === LF) {
-            this.buffer = this.buffer.subarray(this.pos + 1)
-            this.pos = 0
+          if (byte === LF) {
             this.crlfCheck = false
+            this.consumeCurrentByte()
 
             // It is possible that the line feed is not the end of the
             // event. We need to check if the next character is an
@@ -224,19 +197,17 @@ class EventSourceStream extends Transform {
           this.crlfCheck = false
         }
 
-        if (this.buffer[this.pos] === LF || this.buffer[this.pos] === CR) {
+        if (byte === LF || byte === CR) {
           // If the current character is a carriage return, we need to
           // set the crlfCheck flag to true, as we need to check if the
           // next character is a line feed so we can remove it from the
           // buffer
-          if (this.buffer[this.pos] === CR) {
+          if (byte === CR) {
             this.crlfCheck = true
           }
 
-          this.buffer = this.buffer.subarray(this.pos + 1)
-          this.pos = 0
-          if (
-            this.event.data !== undefined || this.event.event || this.event.id !== undefined || this.event.retry) {
+          this.consumeCurrentByte()
+          if (this.hasPendingEvent()) {
             this.processEvent(this.event)
           }
           this.clearEvent()
@@ -250,22 +221,18 @@ class EventSourceStream extends Transform {
 
       // If the current character is an end-of-line, we can process the
       // line
-      if (this.buffer[this.pos] === LF || this.buffer[this.pos] === CR) {
+      if (byte === LF || byte === CR) {
         // If the current character is a carriage return, we need to
         // set the crlfCheck flag to true, as we need to check if the
         // next character is a line feed
-        if (this.buffer[this.pos] === CR) {
+        if (byte === CR) {
           this.crlfCheck = true
         }
 
         // In any case, we can process the line as we reached an
         // end-of-line character
-        this.parseLine(this.buffer.subarray(0, this.pos), this.event)
-
-        // Remove the processed line from the buffer
-        this.buffer = this.buffer.subarray(this.pos + 1)
-        // Reset the position as we removed the processed line from the buffer
-        this.pos = 0
+        this.parseLine(this.readLine(), this.event)
+        this.consumeCurrentByte()
         // A line was processed and this could be the end of the event. We need
         // to check if the next line is empty to determine if the event is
         // finished.
@@ -273,7 +240,7 @@ class EventSourceStream extends Transform {
         continue
       }
 
-      this.pos++
+      this.advanceCursor()
     }
 
     callback()
@@ -298,64 +265,53 @@ class EventSourceStream extends Transform {
       return
     }
 
-    let field = ''
-    let value = ''
+    let fieldLength = line.length
+    let valueStart = line.length
 
     // If the line contains a U+003A COLON character (:)
     if (colonPosition !== -1) {
-      // Collect the characters on the line before the first U+003A COLON
-      // character (:), and let field be that string.
-      // TODO: Investigate if there is a more performant way to extract the
-      // field
-      // see: https://github.com/nodejs/undici/issues/2630
-      field = line.subarray(0, colonPosition).toString('utf8')
+      fieldLength = colonPosition
 
       // Collect the characters on the line after the first U+003A COLON
       // character (:), and let value be that string.
       // If value starts with a U+0020 SPACE character, remove it from value.
-      let valueStart = colonPosition + 1
+      valueStart = colonPosition + 1
       if (line[valueStart] === SPACE) {
         ++valueStart
       }
-      // TODO: Investigate if there is a more performant way to extract the
-      // value
-      // see: https://github.com/nodejs/undici/issues/2630
-      value = line.subarray(valueStart).toString('utf8')
-
-      // Otherwise, the string is not empty but does not contain a U+003A COLON
-      // character (:)
-    } else {
-      // Process the field using the steps described below, using the whole
-      // line as the field name, and the empty string as the field value.
-      field = line.toString('utf8')
-      value = ''
     }
 
-    // Modify the event with the field name and value. The value is also
-    // decoded as UTF-8
-    switch (field) {
-      case 'data':
-        if (event[field] === undefined) {
-          event[field] = value
-        } else {
-          event[field] += `\n${value}`
-        }
-        break
-      case 'retry':
-        if (isASCIINumber(value)) {
-          event[field] = value
-        }
-        break
-      case 'id':
-        if (isValidLastEventId(value)) {
-          event[field] = value
-        }
-        break
-      case 'event':
-        if (value.length > 0) {
-          event[field] = value
-        }
-        break
+    if (isFieldName(line, fieldLength, DATA)) {
+      const value = line.toString('utf8', valueStart)
+
+      if (event.data === undefined) {
+        event.data = value
+      } else {
+        event.data += `\n${value}`
+      }
+      return
+    }
+
+    if (isFieldName(line, fieldLength, RETRY)) {
+      if (isASCIINumberBytes(line, valueStart)) {
+        event.retry = line.toString('utf8', valueStart)
+      }
+      return
+    }
+
+    if (isFieldName(line, fieldLength, ID)) {
+      if (isValidLastEventIdBytes(line, valueStart)) {
+        event.id = line.toString('utf8', valueStart)
+      }
+      return
+    }
+
+    if (isFieldName(line, fieldLength, EVENT)) {
+      const value = line.toString('utf8', valueStart)
+
+      if (value.length > 0) {
+        event.event = value
+      }
     }
   }
 
@@ -385,12 +341,151 @@ class EventSourceStream extends Transform {
   }
 
   clearEvent () {
-    this.event = {
-      data: undefined,
-      event: undefined,
-      id: undefined,
-      retry: undefined
+    this.event.data = undefined
+    this.event.event = undefined
+    this.event.id = undefined
+    this.event.retry = undefined
+  }
+
+  hasPendingEvent () {
+    return this.event.data !== undefined ||
+      this.event.event !== undefined ||
+      this.event.id !== undefined ||
+      this.event.retry !== undefined
+  }
+
+  hasCurrentByte () {
+    return this.chunkIndex < this.chunks.length &&
+      this.pos < this.chunks[this.chunkIndex].length
+  }
+
+  currentByte () {
+    return this.chunks[this.chunkIndex][this.pos]
+  }
+
+  consumeCurrentByte () {
+    this.advanceCursor()
+    this.syncLineStartToCursor()
+  }
+
+  advanceCursor () {
+    this.pos++
+
+    while (this.chunkIndex < this.chunks.length && this.pos >= this.chunks[this.chunkIndex].length) {
+      this.chunkIndex++
+      this.pos = 0
+    }
+  }
+
+  syncLineStartToCursor () {
+    this.lineChunkIndex = this.chunkIndex
+    this.linePos = this.pos
+    this.dropConsumedChunks()
+  }
+
+  dropConsumedChunks () {
+    while (this.lineChunkIndex > 0) {
+      this.chunks.shift()
+      this.lineChunkIndex--
+      this.chunkIndex--
+    }
+
+    if (this.chunkIndex === this.chunks.length) {
+      this.chunks.length = 0
+      this.chunkIndex = 0
+      this.pos = 0
+      this.lineChunkIndex = 0
+      this.linePos = 0
+    }
+  }
+
+  readLine () {
+    if (this.lineChunkIndex === this.chunkIndex) {
+      return this.chunks[this.chunkIndex].subarray(this.linePos, this.pos)
+    }
+
+    const chunks = []
+    let length = 0
+
+    for (let i = this.lineChunkIndex; i <= this.chunkIndex; i++) {
+      const chunk = this.chunks[i]
+      const start = i === this.lineChunkIndex ? this.linePos : 0
+      const end = i === this.chunkIndex ? this.pos : chunk.length
+      const slice = chunk.subarray(start, end)
+      length += slice.length
+      chunks.push(slice)
+    }
+
+    return Buffer.concat(chunks, length)
+  }
+
+  peekBufferedByte (offset) {
+    let chunkIndex = this.lineChunkIndex
+    let pos = this.linePos
+
+    while (chunkIndex < this.chunks.length) {
+      const chunk = this.chunks[chunkIndex]
+      const remaining = chunk.length - pos
+
+      if (offset < remaining) {
+        return chunk[pos + offset]
+      }
+
+      offset -= remaining
+      chunkIndex++
+      pos = 0
+    }
+  }
+
+  discardLeadingBytes (count) {
+    while (count > 0 && this.lineChunkIndex < this.chunks.length) {
+      const chunk = this.chunks[this.lineChunkIndex]
+      const remaining = chunk.length - this.linePos
+
+      if (count < remaining) {
+        this.linePos += count
+        count = 0
+      } else {
+        count -= remaining
+        this.lineChunkIndex++
+        this.linePos = 0
+      }
+    }
+
+    this.chunkIndex = this.lineChunkIndex
+    this.pos = this.linePos
+    this.dropConsumedChunks()
+  }
+
+  handleBOM () {
+    const first = this.peekBufferedByte(0)
+    const second = this.peekBufferedByte(1)
+    const third = this.peekBufferedByte(2)
+
+    if (second === undefined) {
+      if (first === BOM[0]) {
+        return true
+      }
+
+      this.checkBOM = false
+      return true
+    }
+
+    if (third === undefined) {
+      if (first === BOM[0] && second === BOM[1]) {
+        return true
+      }
+
+      this.checkBOM = false
+      return false
     }
+
+    if (first === BOM[0] && second === BOM[1] && third === BOM[2]) {
+      this.discardLeadingBytes(3)
+    }
+
+    this.checkBOM = false
+    return !this.hasCurrentByte()
   }
 }