Thesis

Our Bet

Robotics is about to go through the same jump that generative AI went through between 2020 and 2024. The pattern is always the same:

Once someone turns a hard technical problem into a simple API, usage goes vertical.

We believe CodecFlow will be that API for robotics.

The GenAI Arc

What happened to GenAI in under five years:

• AI infrastructure grew from ~$57B in 2024 to ~$72B in 2025 (+26%). On track for ~$182B by 2029.
• Replicate turned “run any model on a GPU” into a simple API. $5.3M revenue by 2024. Acquired by Cloudflare in 2025.
• fal.ai built a serverless engine for GenAI and raised $140M in 2025 on fast growth.
• AI coding tools went from nothing to $4.7B in 2025, with seven companies already over $100M ARR.
• ChatGPT went from 0 to 100M monthly users in two months, and 800M+ weekly users by late 2025.

Robotics is now sitting where GenAI infrastructure was in 2020–21. Physical AI is the next wave. Every major AI lab is moving toward robots that can act in the real world. And the money is flowing:

• Robotics market: ~$50B in 2025 → ~$111B by 2030 (~14%/yr)
• Industrial robotics: ~$34B in 2024 → ~$70.6B by 2030 (~13%/yr)
• Robot simulation software is growing +20%/yr through 2033 — faster than the hardware market itself

The Analogy

GenAI example	What they did	Our version
fal.ai	”No GPUs to set up, no autoscaler, global low-latency inference for GenAI.”	SimArena + Fabric — no MuJoCo or Isaac setup, no cloud wiring. Build a robot and a scene in the browser. Click to run sim or training in the cloud.
Replicate	Made “run any model on a GPU” a simple API. Acquired by Cloudflare.	Fabric — serverless GPU and CPU for robots and agents. Pay per second with x402.
Coding agents	Turned code writing into a utility. $4B+ in five years.	optr — programmable See → Think → Act runtime across edge, cloud, and browser.

GenAI jumped from “hard research” to “billion-dollar infrastructure” the moment someone made “run a model” feel easy. Robotics is at the same edge. CodecFlow is that stack.