GoodVision AI Claims New Solution for AI 'Token Shortage'

AI infrastructure company introduces intelligent compute scheduling and distributed edge inference to address rising token consumption, latency, and cost challenges.

Mar. 25, 2026 at 9:22am by Ben Kaplan

Got story updates? Submit your updates here. ›

GoodVision AI, an AI infrastructure company led by former AWS and IBM executives, has introduced an intelligent compute scheduling solution combined with distributed edge inference infrastructure, aimed at addressing rising token consumption, latency, and cost challenges driven by the rapid adoption of AI agents. The company argues that scaling centralized compute infrastructure alone is not sufficient to address the efficiency, cost, and latency challenges observed in real-world AI deployments, and is instead building a distributed and intelligent compute delivery layer to orchestrate inference at scale.

Why it matters

As AI agents and applications become more widely adopted, the demand for compute resources is growing exponentially, leading to issues like rising token consumption, latency, and costs. GoodVision AI's approach of combining intelligent scheduling and distributed edge inference aims to address these challenges and make AI more accessible and scalable for developers, enterprises, and individual users.

The details

GoodVision AI's solution involves an intelligent compute scheduling system that dynamically allocates workloads based on task complexity, cost sensitivity, and latency requirements, routing them across public clouds, private data centers, and a network of edge compute nodes. By owning and controlling the underlying compute resources, the company can stabilize token supply, gain pricing power, and maximize margin capture. The company is also actively deploying edge compute nodes to bring compute closer to end users, reducing latency and improving responsiveness.

  • GoodVision AI has been building out its inference compute footprint across Asia and globally since 2025, securing over 400 MW of power capacity.
  • At full buildout, the company's network is designed to support up to 400,000 inference GPUs, representing a multi-billion-dollar compute asset base.

The players

GoodVision AI

An AI infrastructure company led by former AWS and IBM executives, focused on building a distributed and intelligent compute delivery layer to orchestrate inference at scale.

David Wang

The CEO of GoodVision AI, who has decades of experience in the cloud computing industry, including as a Partner at IBM and a former Senior Director at AWS.

Jensen Huang

The CEO of NVIDIA, who noted at GTC 2026 that AI infrastructure is evolving from traditional 'data centers' into 'token factories,' where inference throughput becomes a key metric.

Got photos? Submit your photos here. ›

What they’re saying

“Inference demand could increase by multiple orders of magnitude, potentially reaching a million-fold growth within the next two years.”

— Jensen Huang, CEO, NVIDIA

“Model training happens once, but inference happens billions of times.”

— David Wang, CEO, GoodVision AI

What’s next

GoodVision AI is already expanding into compute-intensive verticals such as video generation and biotech, where the company's distributed and intelligent compute infrastructure is expected to play a crucial role in supporting the growing demand for AI-powered applications.

The takeaway

GoodVision AI's approach to addressing the challenges of AI compute scaling, by combining intelligent scheduling and distributed edge inference, represents a significant shift in how AI infrastructure is designed and deployed, potentially making AI more accessible and scalable for a wide range of developers, enterprises, and individual users.