Introduction

If you want the CPU version of TensorFlow in a Conda-managed environment, the current practical approach is to use Conda for the environment and pip for the TensorFlow package itself. TensorFlow's official installation guidance recommends pip because official releases are published to PyPI rather than being maintained as the primary Conda install target.

Create a Clean Conda Environment First

Using a fresh environment avoids version conflicts with existing machine learning packages.

conda create -n tf-cpu python=3.11
conda activate tf-cpu

Once the environment is active, upgrade pip:

python -m pip install --upgrade pip

That gives you a clean Python environment managed by Conda while still following the official TensorFlow packaging path.

Using a dedicated environment also makes it easier to remove or recreate the setup later without disturbing unrelated Python projects.

Install the CPU Build Inside the Conda Environment

For CPU-only usage, install TensorFlow from PyPI:

python -m pip install tensorflow

On platforms where the separate CPU wheel is available and you explicitly want it, you may also see:

python -m pip install tensorflow-cpu

The important point is that the installation happens inside the activated Conda environment, even though the package source is pip.

Verify That TensorFlow Sees the CPU

After installation, run a quick import and device check:

python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000]))); print(tf.config.list_physical_devices('CPU'))"

If TensorFlow imports successfully and lists a CPU device, the environment is working.

You can also inspect the installed version:

python -c "import tensorflow as tf; print(tf.__version__)"

That is useful when a project depends on a specific major or minor release.

If the import fails, check the Python version in the Conda environment first. TensorFlow support is version-sensitive, and many installation problems come from choosing a Python version outside the supported range for the release you want.

It is also worth confirming that you are invoking the environment's Python and pip, not a system interpreter left over on PATH. Running which python on Unix-like systems or where python on Windows can make that obvious.

Why Not conda install tensorflow?

The confusing part is that many older tutorials show direct Conda installation. That can still appear in search results, but it is not the recommended official path for current TensorFlow releases.

Using pip inside a Conda environment gives you:

• Conda's environment isolation
• TensorFlow's official package channel
• easier alignment with TensorFlow's published installation docs

That combination is usually the least surprising setup for CPU-only development.

It also makes troubleshooting easier because you can separate environment management from package installation. Conda owns the environment, and pip owns the TensorFlow wheel.

For CPU-only experimentation, that separation is often exactly what you want: a reproducible environment without extra GPU driver or CUDA concerns.

Common Pitfalls

• Assuming "using Conda" means every package must be installed with conda install.
• Reusing an old environment with conflicting package versions.
• Following outdated tutorials that install unsupported or stale TensorFlow builds from Conda channels.
• Forgetting to activate the Conda environment before running pip install.
• Choosing a Python version that the target TensorFlow release does not support.

Summary

• Use Conda to create and manage the environment.
• Install the official TensorFlow CPU package with pip inside that environment.
• Verify the installation with a simple import and device query.
• Prefer a fresh environment to avoid package conflicts.
• Older direct-Conda TensorFlow instructions are often outdated compared with current official guidance.