The Information Machine

Quite a massive inferencing rack breakthrough from @TensordyneInc .

Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-17

Tensordyne announces an AI inference rack claiming 13x the rack throughput of NVIDIA's NVL72 GB300 on DeepSeek-R1 workloads, based on internal simulations rather than independent benchmarks.

Open original ↗

Appears in

Extraction

Topics: ai-inference-hardwarenvidia-competitionrack-computingdeepseek

Claims

  • Tensordyne claims its AI inference rack achieves 13x the rack throughput of NVIDIA's NVL72 GB300 on DeepSeek-R1 workloads.
  • The performance figures are based on internal simulations, not third-party or independent testing.
  • Tensordyne is a relatively new entrant challenging established GPU rack vendors in the inference market.

Key quotes

They just announced an AI-inference rack, claiming 13x the rack throughput of NVIDIA's NVL72 GB300 in a DeepSeek-R1 comparison based on internal simulations.