Infinity, founded by Jeremy Nixon, the creator of hacker network community AGI House, to build an inference library that runs on all chips, raised a $15M seed
Context & Ripple Effects
Infinity enters an inference-software field already attracting both open-source commercialization and multi-hardware deployment efforts. Earlier in 2026, vLLM creators' commercialization effort raised a substantially larger seed round, while Gimlet Labs positioned itself around multi-silicon inference cloud infrastructure.
The contrast with chip-focused inference companies such as d-Matrix's inference-optimized hardware matters: Infinity is targeting the software layer intended to span underlying processors rather than a single processor architecture.
First-order effects
- The $15M seed gives Infinity resources to build and recruit around an inference library designed for cross-chip deployment.
- Developers and infrastructure operators may gain another prospective abstraction layer for running inference workloads without tying software as closely to one hardware target.
Second-order effects
- Inference-library and serving-platform rivals will face added pressure to show broad hardware support, performance portability, or a clearer route to commercial adoption.
- Specialized inference-chip vendors could benefit if a common software layer lowers integration friction, but they may also lose some differentiation if workload portability becomes easier.
Third-order effects
- If cross-chip libraries gain adoption, value in inference could shift toward software compatibility and orchestration layers rather than residing solely in individual accelerators.
- The outcome remains uncertain: portable interfaces must still prove they can accommodate hardware-specific performance trade-offs, leaving room for both broad abstractions and vendor-tuned stacks.
The trend: AI inference is becoming a contest to control the software layer that makes increasingly diverse compute hardware usable at scale.