Four AI-Chip Giants Challenge NVIDIA in a Single Ten-Day Stretch

In a ten-day burst this August, four of the world's largest technology companies unveiled new AI silicon, intensifying competition with NVIDIA. NVIDIA opened its Rubin platform; Google split its eighth-generation TPU into two variants — the TPU 8i for fast inference and the TPU 8t for training; Intel detailed a 3-nanometer, efficiency-focused AI processor claiming 40% better power efficiency; and Microsoft prepared its Maia 300 accelerator, with reports of a 300,000-unit TSMC order for 2027. The coordinated moves come as cloud revenue surges — Google Cloud's backlog reached $514 billion — and hyperscalers seek to reduce dependence on a single supplier.