Tools
Engineering Tools
One tool per engineer question, grouped by the four physical-layer stacks behind every AI training cluster. 10 interactive now; the rest are scoped and on the roadmap. Each tool links back to the topic it informs — open a tool from any topic’s Products tab, or browse the full set here.
energy
Energy
Power, cooling, and grid capacity — the rate limit of every AI buildout.
Datacenter Power Calculator
Estimate MW capacity required for an AI training or inference cluster of a given size.
Live · interactive
PUE Comparison
Compare PUE benchmarks across hyperscale operators, regions, and cooling approaches.
Live · interactive
US Electricity Cost by Region
Per-state industrial electricity rate snapshot + 12-month trend, indexed to datacenter footprint.
Coming soon
Cooling Load Estimator
Estimate the cooling tonnage / liquid-cooling capacity needed for a given rack density.
Live · interactive
optical
Optical
Interconnect bandwidth, reach, and form factor — the fabric of AI training clusters.
QSFP-DD vs OSFP
Side-by-side decision matrix for QSFP-DD vs OSFP form factor across 400G / 800G / 1.6T.
Live · interactive
400G vs 800G Economics
Per-Gbps cost, power, and rack-density tradeoff for 400G vs 800G fabric in mid-2026.
Live · interactive
Coherent vs Lumentum
Two-supplier head-to-head across product line, capacity, customer wins, and recent filings.
Coming soon
DWDM Spacing Calculator
Wavelength count and grid spacing required for a target aggregate bandwidth over fiber.
Live · interactive
hbm
HBM Memory
Capacity and bandwidth bottlenecks for training and inference at scale.
HBM3 vs HBM3e vs HBM4
Bandwidth, capacity, and availability cross-cut for HBM generations across SK hynix / Micron / Samsung.
Live · interactive
HBM Supplier Lead Time
Quarterly lead-time snapshot across the three HBM suppliers + capacity guidance.
Coming soon
HBM Price Index
Rolling $/GB index across HBM2e / HBM3 / HBM3e + 12-month forecast trend.
Coming soon
chips
Chips & Compute
Accelerator TCO, throughput, and supply for AI training and inference.
H100 vs H200 vs B200 TCO
5-year TCO for the three NVIDIA generations on a training workload of a given size.
Live · interactive
Inference Cost Calculator
Per-million-tokens cost for self-hosted inference across H100 / H200 / B200 / MI300.
Live · interactive
Memory Bandwidth Bottleneck Detector
Given a model + accelerator, decide whether you are bandwidth-bound or compute-bound.
Live · interactive