Anti-NVIDIA data center start-up reaches $1.5B valuation

Anti-NVIDIA data center start-up reaches $1.5B valuation

In a major development, TensorWave, an AI-focused cloud computing start-up, has raised $350 million in its Series B funding round.

The investment was led by Advanced Micro Devices (AMD) and hedge fund Magnetar Capital.

With this investment, TensorWave's post-money valuation has soared to $1.55 billion, nearly quadrupling its value from the last funding round held about a year ago when it was valued at around $400 million.

The company's focus on AMD's hardware

Unlike most companies in the AI space, TensorWave doesn't use NVIDIA's GPUs or any of its products.

The company's co-founder and CEO, Darrick Horton, believes that NVIDIA has too much control over the AI infrastructure market.

To promote competition, TensorWave exclusively uses hardware and software from AMD.

The company currently operates three data centers in Pennsylvania, Arizona, and Florida with computing capacity equivalent to 10,000 megawatts (MW) of electrical power.

Plans for expansion and new data centers

TensorWave plans to use the fresh capital from its Series B funding round to expand its operations.

This includes adding more data centers and purchasing generators, chillers, critical power-supply gear, transformers, and other equipment.

The company has signed leases for a total of 500MW capacity, mostly in still-to-be-built data centers.

In the coming year, it hopes to raise that number to 2GW.

Partnership with AMD to enhance ROCm software

Over the past year, TensorWave has partnered with AMD to improve ROCm, its custom-software platform.

The software was earlier criticized for being buggy and hard to use compared to NVIDIA's rival programming library, CUDA.

However, Horton claims that the software has improved significantly and is now "pretty much plug-and-play."

This makes AMD's chips well-suited for handling inference computing workloads, which allow AI models to respond to queries.