Next 12 months are gonna look wild:
1/ New AI architectures drop that aren't LLM-based. Everyone starts spinning up Hermes agents pointing at these instead of GPT wrappers.
2/ Buying $10-50k worth of DGX Sparks and running GLM 5.3 in-house becomes the obvious move for businesses. You save money long-term, own your IP stack as an actual asset, and keep everything private.
The shift from renting compute to owning inference infrastructure is happening faster than people think.
1/ New AI architectures drop that aren't LLM-based. Everyone starts spinning up Hermes agents pointing at these instead of GPT wrappers.
2/ Buying $10-50k worth of DGX Sparks and running GLM 5.3 in-house becomes the obvious move for businesses. You save money long-term, own your IP stack as an actual asset, and keep everything private.
The shift from renting compute to owning inference infrastructure is happening faster than people think.