NVIDIA GB300 and GB200 NVL72 systems boost AI efficiency 50x and slash compute costs by 35x

NVIDIA has introduced its latest GB300 NVL72 systems, which deliver 50 times more work per megawatt and cut per-token processing costs by 35 times compared with the earlier Hopper platform. Separate testing on the GB200 NVL72 indicates more than tenfold gains in information processed per watt, while rapid growth in AI coding assistants and a widening global AI skills gap are reshaping hardware demand and enterprise strategies. Projections show the AI agent market climbing from 4.92 billion dollars in 2024 to 44.97 billion dollars by 2035, as businesses increasingly adopt off‑the‑shelf AI solutions.