diff --git a/Cargo.lock b/Cargo.lock
index 5d1542a9fb99..23f5306f8a4f 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -220,15 +220,6 @@ dependencies = [
  "syn 2.0.94",
 ]
 
-[[package]]
-name = "atoi"
-version = "2.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f28d99ec8bfea296261ca1af174f24225171fea9664ba9003cbebee704810528"
-dependencies = [
- "num-traits",
-]
-
 [[package]]
 name = "atoi_simd"
 version = "0.16.0"
@@ -2939,7 +2930,6 @@ version = "0.45.1"
 dependencies = [
  "ahash",
  "async-stream",
- "atoi",
  "atoi_simd",
  "avro-schema",
  "bytemuck",
@@ -3505,7 +3495,7 @@ dependencies = [
 name = "polars-time"
 version = "0.45.1"
 dependencies = [
- "atoi",
+ "atoi_simd",
  "bytemuck",
  "chrono",
  "chrono-tz",
diff --git a/Cargo.toml b/Cargo.toml
index 9ba205ecee31..41ff9b368a91 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -21,7 +21,6 @@ repository = "https://github.com/pola-rs/polars"
 ahash = ">=0.8.5"
 aho-corasick = "1.1"
 arboard = { version = "3.4.0", default-features = false }
-atoi = "2"
 atoi_simd = "0.16"
 atomic-waker = "1"
 avro-schema = { version = "0.3" }
diff --git a/crates/polars-arrow/Cargo.toml b/crates/polars-arrow/Cargo.toml
index 8fac057bd48f..8d1678ca5081 100644
--- a/crates/polars-arrow/Cargo.toml
+++ b/crates/polars-arrow/Cargo.toml
@@ -13,7 +13,6 @@ repository = { workspace = true }
 description = "Minimal implementation of the Arrow specification forked from arrow2"
 
 [dependencies]
-atoi = { workspace = true, optional = true }
 bytemuck = { workspace = true, features = ["must_cast"] }
 chrono = { workspace = true }
 # for timezone support
@@ -143,7 +142,7 @@ timezones = [
   "chrono-tz",
 ]
 dtype-array = []
-dtype-decimal = ["atoi", "itoa"]
+dtype-decimal = ["atoi_simd", "itoa"]
 bigidx = ["polars-utils/bigidx"]
 nightly = []
 performant = []
diff --git a/crates/polars-arrow/src/compute/decimal.rs b/crates/polars-arrow/src/compute/decimal.rs
index e5dc464f203b..6145d0985f56 100644
--- a/crates/polars-arrow/src/compute/decimal.rs
+++ b/crates/polars-arrow/src/compute/decimal.rs
@@ -1,6 +1,5 @@
 use std::sync::atomic::{AtomicBool, Ordering};
 
-use atoi::FromRadix10SignedChecked;
 use num_traits::Euclid;
 
 static TRIM_DECIMAL_ZEROS: AtomicBool = AtomicBool::new(false);
@@ -12,103 +11,86 @@ pub fn set_trim_decimal_zeros(trim: Option<bool>) {
     TRIM_DECIMAL_ZEROS.store(trim.unwrap_or(false), Ordering::Relaxed)
 }
 
-/// Count the number of b'0's at the beginning of a slice.
-fn leading_zeros(bytes: &[u8]) -> u8 {
-    bytes.iter().take_while(|byte| **byte == b'0').count() as u8
-}
-
-fn split_decimal_bytes(bytes: &[u8]) -> (Option<&[u8]>, Option<&[u8]>) {
-    let mut a = bytes.splitn(2, |x| *x == b'.');
-    let lhs = a.next();
-    let rhs = a.next();
-    (lhs, rhs)
-}
-
-/// Parse a single i128 from bytes, ensuring the entire slice is read.
-fn parse_integer_checked(bytes: &[u8]) -> Option<i128> {
-    let (n, len) = i128::from_radix_10_signed_checked(bytes);
-    n.filter(|_| len == bytes.len())
-}
-
 /// Assuming bytes are a well-formed decimal number (with or without a separator),
 /// infer the scale of the number.  If no separator is present, the scale is 0.
 pub fn infer_scale(bytes: &[u8]) -> u8 {
-    let (_lhs, rhs) = split_decimal_bytes(bytes);
-    rhs.map_or(0, |x| x.len() as u8)
+    let Some(separator) = bytes.iter().position(|b| *b == b'.') else {
+        return 0;
+    };
+    (bytes.len() - (1 + separator)) as u8
 }
 
-/// Deserialize bytes to a single i128 representing a decimal, at a specified precision
-/// (optional) and scale (required).  If precision is not specified, it is assumed to be
-/// 38 (the max precision allowed by the i128 representation).  The number is checked to
-/// ensure it fits within the specified precision and scale.  Consistent with float parsing,
-/// no decimal separator is required (eg "500", "500.", and "500.0" are all accepted); this allows
-/// mixed integer/decimal sequences to be parsed as decimals.  All trailing zeros are assumed to
-/// be significant, whether or not a separator is present: 1200 requires precision >= 4, while 1200.200
-/// requires precision >= 7 and scale >= 3.  Returns None if the number is not well-formed, or does not
-/// fit. Only b'.' is allowed as a decimal separator (issue #6698).
+/// Deserialize bytes to a single i128 representing a decimal, at a specified
+/// precision (optional) and scale (required). The number is checked to ensure
+/// it fits within the specified precision and scale.  Consistent with float
+/// parsing, no decimal separator is required (eg "500", "500.", and "500.0" are
+/// all accepted); this allows mixed integer/decimal sequences to be parsed as
+/// decimals.  All trailing zeros are assumed to be significant, whether or not
+/// a separator is present: 1200 requires precision >= 4, while 1200.200
+/// requires precision >= 7 and scale >= 3. Returns None if the number is not
+/// well-formed, or does not fit. Only b'.' is allowed as a decimal separator
+/// (issue #6698).
 #[inline]
-pub fn deserialize_decimal(mut bytes: &[u8], precision: Option<u8>, scale: u8) -> Option<i128> {
-    // While parse_integer_checked will parse positive/negative numbers, we want to
-    // handle the sign ourselves, and so check for it initially, then handle it
-    // at the end.
+pub fn deserialize_decimal(bytes: &[u8], precision: Option<u8>, scale: u8) -> Option<i128> {
+    let precision_digits = precision.unwrap_or(38).min(38) as usize;
+    if scale as usize > precision_digits {
+        return None;
+    }
+
+    let separator = bytes.iter().position(|b| *b == b'.').unwrap_or(bytes.len());
+    let (mut int, mut frac) = bytes.split_at(separator);
+    if frac.len() <= 1 || scale == 0 {
+        // Only integer fast path.
+        let n: i128 = atoi_simd::parse(int).ok()?;
+        let ret = n.checked_mul(POW10[scale as usize] as i128)?;
+        if precision.is_some() && ret >= POW10[precision_digits] as i128 {
+            return None;
+        }
+        return Some(ret);
+    }
+
+    // Skip period.
+    frac = &frac[1..];
+
+    // Skip sign.
     let negative = match bytes.first() {
         Some(s @ (b'+' | b'-')) => {
-            bytes = &bytes[1..];
+            int = &int[1..];
             *s == b'-'
         },
         _ => false,
     };
-    let (lhs, rhs) = split_decimal_bytes(bytes);
-    let precision = precision.unwrap_or(38);
 
-    let lhs_b = lhs?;
+    // Truncate trailing digits that extend beyond the scale.
+    let frac_scale = if scale as usize <= frac.len() {
+        frac = &frac[..scale as usize];
+        0
+    } else {
+        scale as usize - frac.len()
+    };
 
-    // For the purposes of decimal parsing, we assume that all digits other than leading zeros
-    // are significant, eg, 001200 has 4 significant digits, not 2.  The Decimal type does
-    // not allow negative scales, so all trailing zeros on the LHS of any decimal separator
-    // will still take up space in the representation (eg, 1200 requires, at minimum, precision 4
-    // at scale 0; there is no scale -2 where it would only need precision 2).
-    let lhs_s = lhs_b.len() as u8 - leading_zeros(lhs_b);
+    // Parse and combine parts.
+    let pint: u128 = if int.is_empty() {
+        0
+    } else {
+        atoi_simd::parse_pos(int).ok()?
+    };
+    let pfrac: u128 = atoi_simd::parse_pos(frac).ok()?;
 
-    if lhs_s + scale > precision {
-        // the integer already exceeds the precision
+    let ret = pint
+        .checked_mul(POW10[scale as usize])?
+        .checked_add(pfrac.checked_mul(POW10[frac_scale])?)?;
+    if precision.is_some() && ret >= POW10[precision_digits] {
         return None;
     }
-
-    let abs = parse_integer_checked(lhs_b).and_then(|x| match rhs {
-        // A decimal separator was found, so LHS and RHS need to be combined.
-        Some(mut rhs) => {
-            if matches!(rhs.first(), Some(b'+' | b'-')) {
-                // RHS starts with a '+'/'-' sign and the number is not well-formed.
-                return None;
-            }
-            let scale_adjust = if (scale as usize) <= rhs.len() {
-                // Truncate trailing digits that extend beyond the scale
-                rhs = &rhs[..scale as usize];
-                None
-            } else {
-                Some(scale as u32 - rhs.len() as u32)
-            };
-
-            parse_integer_checked(rhs).map(|y| {
-                let lhs = x * 10i128.pow(scale as u32);
-                let rhs = scale_adjust.map_or(y, |s| y * 10i128.pow(s));
-                lhs + rhs
-            })
-        },
-        // No decimal separator was found; we have an integer / LHS only.
-        None => {
-            if lhs_b.is_empty() {
-                // we simply have no number at all / an empty string.
-                return None;
-            }
-            Some(x * 10i128.pow(scale as u32))
-        },
-    });
     if negative {
-        Some(-abs?)
+        if ret > (1 << 127) {
+            None
+        } else {
+            Some(ret.wrapping_neg() as i128)
+        }
     } else {
-        abs
+        ret.try_into().ok()
     }
 }
 
@@ -322,6 +304,11 @@ mod test {
         assert_eq!(deserialize_decimal(val, Some(5), 6), None); // insufficient precision, excess scale
         assert_eq!(deserialize_decimal(val, Some(5), 3), None); // insufficient precision, exact scale
         assert_eq!(deserialize_decimal(val, Some(12), 5), Some(120001000)); // excess precision, excess scale
-        assert_eq!(deserialize_decimal(val, None, 35), None); // scale causes insufficient precision
+        assert_eq!(
+            deserialize_decimal(val, None, 35),
+            Some(120001000000000000000000000000000000000)
+        );
+        assert_eq!(deserialize_decimal(val, None, 36), None);
+        assert_eq!(deserialize_decimal(val, Some(38), 35), None); // scale causes insufficient precision
     }
 }
diff --git a/crates/polars-time/Cargo.toml b/crates/polars-time/Cargo.toml
index d144b177174b..62577bdb41cf 100644
--- a/crates/polars-time/Cargo.toml
+++ b/crates/polars-time/Cargo.toml
@@ -16,7 +16,7 @@ polars-error = { workspace = true }
 polars-ops = { workspace = true }
 polars-utils = { workspace = true }
 
-atoi = { workspace = true }
+atoi_simd = { workspace = true }
 bytemuck = { workspace = true }
 chrono = { workspace = true }
 chrono-tz = { workspace = true, optional = true }
diff --git a/crates/polars-time/src/chunkedarray/string/strptime.rs b/crates/polars-time/src/chunkedarray/string/strptime.rs
index 04b6f4ebd315..229281a30dfa 100644
--- a/crates/polars-time/src/chunkedarray/string/strptime.rs
+++ b/crates/polars-time/src/chunkedarray/string/strptime.rs
@@ -1,6 +1,5 @@
 //! Much more opinionated, but also much faster strptrime than the one given in Chrono.
 //!
-use atoi::FromRadix10;
 use chrono::{NaiveDate, NaiveDateTime};
 use once_cell::sync::Lazy;
 use regex::Regex;
@@ -14,7 +13,7 @@ static TWELVE_HOUR_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"%[_-]?[Il]")
 static MERIDIEM_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"%[_-]?[pP]").unwrap());
 
 #[inline]
-fn update_and_parse<T: atoi::FromRadix10>(
+fn update_and_parse<T: atoi_simd::Parse>(
     incr: usize,
     offset: usize,
     vals: &[u8],
@@ -22,7 +21,7 @@ fn update_and_parse<T: atoi::FromRadix10>(
     // this maybe oob because we cannot entirely sure about fmt lengths
     let new_offset = offset + incr;
     let bytes = vals.get(offset..new_offset)?;
-    let (val, parsed) = T::from_radix_10(bytes);
+    let (val, parsed) = atoi_simd::parse_any(bytes).ok()?;
     if parsed == 0 {
         None
     } else {
@@ -154,7 +153,7 @@ impl StrpTimeState {
                         let new_offset = offset + 2;
                         let bytes = val.get_unchecked(offset..new_offset);
 
-                        let (decade, parsed) = i32::from_radix_10(bytes);
+                        let (decade, parsed) = atoi_simd::parse_any::<i32>(bytes).ok()?;
                         if parsed == 0 {
                             return None;
                         }