Cloudflare's AI team gains talent from Ensemble AI to make Workers AI faster and cheaper!

Hi everyone, it's Shiichan! Today's news is a little more grown-up... but if you love AI, it's exciting: Cloudflare's AI team is welcoming some new teammates!

Cloudflare Blog blog.cloudflare.com

What was announced?

On the Cloudflare Blog (dated June 15, 2026), it was announced that key members of a startup called Ensemble AI are joining Cloudflare. The goal is to accelerate AI infrastructure, with a focus on machine learning infrastructure and efficiency.

Ensemble AI is a San Francisco-based company founded in 2023. The team has worked on making large AI models faster, smaller, and cheaper to run — developing model compression and efficient inference approaches that cut the memory, compute, and deployment overhead of large language models and multimodal architectures.

Why it matters

When you try to bring AI to lots of people, one of the biggest walls is the cost of inference. Every time a model runs, GPUs work hard, so scaling up means more money and more operational effort. Cloudflare sees this as one of the biggest barriers to scaling AI applications, which is exactly why it's bringing in the Ensemble AI team.

Here's the mission Ensemble AI has pursued:

making large models faster, smaller, and more cost-effective to serve, without sacrificing quality.

(In other words: run large models faster, smaller, and more affordably — without giving up quality!)

What changes

The joining team will work on improving model efficiency, GPU utilization, and scalable deployment. The aim is to make Workers AI faster, more flexible, and more cost-efficient!

For developers, the direction is to help you deploy AI applications with lower cost, better performance, and less operational overhead. It's not something you can go try today, but it's exciting to think about how Workers AI will evolve from here!

Dive Deep

The post also highlighted some of the technology Ensemble AI has built.

• NdLinear: a drop-in replacement for standard linear layers in transformer models that operates directly on multidimensional activations rather than flattening structure away.
• NdLinear-LoRA: an efficient adaptation method that reduces trainable parameters when fine-tuning large models.
• These approaches can complement quantization and vector quantization techniques.

Cloudflare already has AI infrastructure in place, and this addition layers on top of it.

• Workers AI: serverless, GPU-powered inference running on Cloudflare's global network
• Infire: Cloudflare's inference engine
• Unweight: a tensor compression technique

Adding Ensemble AI's efficiency expertise to this foundation should make the platform for running extra-large language models even stronger!

Wrap-up

• Cloudflare's Blog announced that key members of Ensemble AI are joining Cloudflare.
• The focus is machine learning infrastructure and efficiency, aiming to make Workers AI faster, more flexible, and more cost-efficient.
• Ensemble AI, founded in San Francisco in 2023, specializes in model compression and efficient inference, and built NdLinear and NdLinear-LoRA.
• The goal is to lower inference cost — one of the biggest barriers to scaling AI.

This isn't something you can try hands-on right away, but if you run AI apps on Workers AI or wrestle with inference costs, it's definitely a story worth watching!