Lubbock.cloud

Compute tokens

AMD-only tokenized GPU capacity for training, inference, Monte Carlo, and risk grids—priced transparently per compute unit.

AMD token catalogue

Token

GPU: AMD Instinct MI300X · 304 XCDNA 3 CUs
HBM memory: 192 GB HBM3 · 5.3 TB/s
Ideal workloads: Large-model training · Monte Carlo (high path count) · Batch inference

Price

$0.94per GPU-hour

Token

GPU: AMD Instinct MI325X · expanded matrix engines
HBM memory: 256 GB HBM3e · higher effective bandwidth
Ideal workloads: Multi-epoch training · Scenario risk grids · Low-latency inference

Price

$1.18per GPU-hour

Token

GPU: AMD Instinct MI355X · next-gen CDNA architecture
HBM memory: Planned 288+ GB HBM class (roadmap)
Ideal workloads: Frontier backtests · VaR / ES simulation · Strategy research at scale

Price

$1.42per GPU-hour

Institutional table layout—alternating rows for scanability. Figures are representative; final SKUs follow Lubbock.cloud issuance.

Token	GPU	HBM	Workloads	Unit price
LUB-MI300X	AMD Instinct MI300X · 304 XCDNA 3 CUs	192 GB HBM3 · 5.3 TB/s	Large-model training; Monte Carlo (high path count); Batch inference	$0.94per GPU-hour
LUB-MI325X	AMD Instinct MI325X · expanded matrix engines	256 GB HBM3e · higher effective bandwidth	Multi-epoch training; Scenario risk grids; Low-latency inference	$1.18per GPU-hour
LUB-MI355X	AMD Instinct MI355X · next-gen CDNA architecture	Planned 288+ GB HBM class (roadmap)	Frontier backtests; VaR / ES simulation; Strategy research at scale	$1.42per GPU-hour