Skip to content
Research

Provider updates, model releases, and capacity catalysts

Signals that explain price, latency, capacity, and route-quality moves across inference markets.

Provider update

H100 capacity tightens in Johannesburg

Vultr and local GPU pools show lower free capacity after overnight agent runtime launches.

Spot GPU prices up 2.1%
Model release

Open-weight routing lowers token spend

Hosted Llama 4 endpoints are absorbing batch workloads with tighter spreads and better queue depth.

Llama routes down 2.8%
Status change

Edge ARM clears maintenance window

Cape Town edge inventory returned with lower utilization and shorter startup times.

Jetson capacity back online

Market event tape

Latest exchange catalysts

A100 80GB PCIe capacity increase
Vultr · Johannesburg, ZA
2m ago
RTX 4090 price spike
Joburg GPU Pool · Johannesburg, ZA
5m ago
Mac mini M2 Pro added
EdgeLab ZA · Stellenbosch, ZA
7m ago
Low capacity: Jetson Orin Nano
CapeCompute · Cape Town, ZA
12m ago
Claude 3.5 Sonnet demand surge
CapeCompute · Cape Town, ZA
15m ago