""In the future, those numbers are reversed," he told reporters Wednesday at CES. "Eighty percent will be on the inference and 20% will be on training. That is our forecast.""
""We definitely want to lead the trend.""
""When you train foundational models, you start big, and you put all the capital in up front," he said."
AI investment focus is moving from training large language models toward inference infrastructure. Historically about 80% of AI spending went to training LLMs while 20% went to inference. Forecasts predict that ratio will flip, with roughly 80% of spending on inference and 20% on training. Industry estimates show inference workloads comprising half of AI compute in 2025 and rising to two-thirds in 2026. Training requires large upfront capital, while deployments such as chatbots start small and scale over time. Increased enterprise deployment and scaling will drive substantial inference infrastructure spending in 2026.
Read at Computerworld
Unable to calculate read time
Collection
[
|
...
]