diff --git a/doc/languages-frameworks/python.section.md b/doc/languages-frameworks/python.section.md
index 3e4e12665bd0..acd2bf769b0d 100644
--- a/doc/languages-frameworks/python.section.md
+++ b/doc/languages-frameworks/python.section.md
@@ -1104,7 +1104,7 @@ on `numpy` will be built with `mkl`.
 The following is an overlay that configures `numpy` to use `mkl`:
 ```nix
 self: super: {
-  python36 = super.python36.override {
+  python37 = super.python37.override {
     packageOverrides = python-self: python-super: {
       numpy = python-super.numpy.override {
         blas = super.pkgs.mkl;
@@ -1114,6 +1114,15 @@ self: super: {
 }
 ```
 
+`mkl` requires an `openmp` implementation when running with multiple processors.
+By default, `mkl` will use Intel's `iomp` implementation if no other is
+specified, but this is a runtime-only dependency and binary compatible with the
+LLVM implementation. To use that one instead, Intel recommends users set it with
+`LD_PRELOAD`.
+
+Note that `mkl` is only available on `x86_64-{linux,darwin}` platforms;
+moreover, Hydra is not building and distributing pre-compiled binaries using it.
+
 ## Contributing
 
 ### Contributing guidelines
diff --git a/pkgs/development/libraries/science/math/mkl/default.nix b/pkgs/development/libraries/science/math/mkl/default.nix
index 37814047f975..0eecb2012f1e 100644
--- a/pkgs/development/libraries/science/math/mkl/default.nix
+++ b/pkgs/development/libraries/science/math/mkl/default.nix
@@ -1,21 +1,8 @@
 { stdenvNoCC, writeText, fetchurl, rpmextract, undmg }:
 /*
-  Some (but not all) mkl functions require openmp, but Intel does not add these
-  to SO_NEEDED and instructs users to put openmp on their LD_LIBRARY_PATH. If
-  you are using mkl and your library/application is using some of the functions
-  that require openmp, add a setupHook like this to your package:
-
-  setupHook = writeText "setup-hook.sh" ''
-    addOpenmp() {
-        addToSearchPath LD_LIBRARY_PATH ${openmp}/lib
-    }
-    addEnvHooks "$targetOffset" addOpenmp
-  '';
-
-  We do not add the setup hook here, because avoiding it allows this large
-  package to be a fixed-output derivation with better cache efficiency.
- */
-
+  For details on using mkl as a blas provider for python packages such as numpy,
+  numexpr, scipy, etc., see the Python section of the NixPkgs manual.
+*/
 stdenvNoCC.mkDerivation rec {
   name = "mkl-${version}";
   version = "${date}.${rel}";
@@ -43,16 +30,23 @@ stdenvNoCC.mkDerivation rec {
   '' else ''
     rpmextract rpm/intel-mkl-common-c-${date}-${rel}-${date}-${rel}.noarch.rpm
     rpmextract rpm/intel-mkl-core-rt-${date}-${rel}-${date}-${rel}.x86_64.rpm
+    rpmextract rpm/intel-openmp-19.0.0-${rel}-19.0.0-${rel}.x86_64.rpm
   '';
 
   installPhase = if stdenvNoCC.isDarwin then ''
       mkdir -p $out/lib
+
       cp -r compilers_and_libraries_${version}/mac/mkl/include $out/
-      cp -r compilers_and_libraries_${version}/mac/mkl/lib/* $out/lib/
+
       cp -r compilers_and_libraries_${version}/licensing/mkl/en/license.txt $out/lib/
+      cp -r compilers_and_libraries_${version}/mac/compiler/lib/* $out/lib/
+      cp -r compilers_and_libraries_${version}/mac/mkl/lib/* $out/lib/
   '' else ''
       mkdir -p $out/lib
+
       cp -r opt/intel/compilers_and_libraries_${version}/linux/mkl/include $out/
+
+      cp -r opt/intel/compilers_and_libraries_${version}/linux/compiler/lib/intel64_lin/* $out/lib/
       cp -r opt/intel/compilers_and_libraries_${version}/linux/mkl/lib/intel64_lin/* $out/lib/
       cp license.txt $out/lib/
   '';
@@ -66,8 +60,8 @@ stdenvNoCC.mkDerivation rec {
   outputHashAlgo = "sha256";
   outputHashMode = "recursive";
   outputHash = if stdenvNoCC.isDarwin
-    then "1224dln7n8px1rk8biiggf77wjhxh8mzw0hd8zlyjm8i6j8w7i12"
-    else "0d8ai0wi8drp071acqkm1wv6vyg12010y843y56zzi1pql81xqvx";
+    then "0000000000000000000000000000000000000000000000000000"
+    else "1amagcaan0hk3x9v7gg03gkw02n066v4kmjb32yyzsy5rfrivb1a";
 
   meta = with stdenvNoCC.lib; {
     description = "Intel Math Kernel Library";
@@ -78,7 +72,7 @@ stdenvNoCC.mkDerivation rec {
       threading models.
     '';
     homepage = https://software.intel.com/en-us/mkl;
-    license = [ licenses.issl licenses.unfreeRedistributable ];
+    license = licenses.issl;
     platforms = [ "x86_64-linux" "x86_64-darwin" ];
     maintainers = [ maintainers.bhipple ];
   };
diff --git a/pkgs/development/python-modules/numexpr/default.nix b/pkgs/development/python-modules/numexpr/default.nix
index acf41fb539e3..280b11706a17 100644
--- a/pkgs/development/python-modules/numexpr/default.nix
+++ b/pkgs/development/python-modules/numexpr/default.nix
@@ -3,7 +3,6 @@
 , fetchPypi
 , python
 , numpy
-, llvmPackages ? null
 }:
 
 buildPythonPackage rec {
@@ -16,16 +15,11 @@ buildPythonPackage rec {
   };
 
   # Remove existing site.cfg, use the one we built for numpy.
-  # Somehow openmp needs to be added to LD_LIBRARY_PATH
-  # https://software.intel.com/en-us/forums/intel-system-studio/topic/611682
   preBuild = ''
     rm site.cfg
     ln -s ${numpy.cfg} site.cfg
-    export LD_LIBRARY_PATH=${llvmPackages.openmp}/lib
   '';
 
-  buildInputs = [] ++ lib.optional (numpy.blasImplementation == "mkl") llvmPackages.openmp;
-
   propagatedBuildInputs = [ numpy ];
 
   # Run the test suite.
@@ -47,4 +41,4 @@ buildPythonPackage rec {
     homepage = "https://github.com/pydata/numexpr";
     license = lib.licenses.mit;
   };
-}
\ No newline at end of file
+}
diff --git a/pkgs/development/python-modules/numpy/default.nix b/pkgs/development/python-modules/numpy/default.nix
index 8a3879f19822..d66cbd77bbc5 100644
--- a/pkgs/development/python-modules/numpy/default.nix
+++ b/pkgs/development/python-modules/numpy/default.nix
@@ -71,8 +71,6 @@ in buildPythonPackage rec {
     inherit blasImplementation cfg;
   };
 
-  doCheck = blasImplementation != "mkl";
-
   # Disable two tests
   # - test_f2py: f2py isn't yet on path.
   # - test_large_file_support: takes a long time and can cause the machine to run out of disk space