Scaling from 5 to 50 units: how one operator built up over 14 months

This post is a composite. The numbers, mistakes, and lessons are pulled from three separate customer accounts who placed their first orders with us in early 2025 and worked up to fleets of 35 to 60 units by mid-2026. We've anonymised it, but the operational details are accurate to what we actually saw.

Where they started

The starting position looked broadly the same across all three: a mid-sized industrial space (40 to 80 kW capacity), a power rate around $0.045 to 0.055/kWh, and a budget that allowed for an initial 5-unit deployment with room to scale if the operational economics held.

The first orders were 5× Bitmain S21 Pro in every case. Not because the S21 Pro is universally the right unit (it's not), but because at $0.05/kWh it had the cleanest break-even of the current-gen SHA-256 options at the time, and starting on one platform simplifies firmware, monitoring, and pool setup while you're still learning your facility.

The first 30 days

Three things that happened on every first deployment:

1. One unit out of five hit a minor warranty issue within the first three weeks. Hashboard partially offline in one case, fan controller error in another. We replaced both under the 60-day warranty. This is normal, 60-day infant-mortality rates across the industry sit around 5 to 10% on new hardware. Plan for it; don't panic about it.
2. Realised hashrate undershot nameplate by 2 to 4%. Spec says 234 TH/s; real-world averaged 226 to 229 TH/s after firmware-side variance, immersion-vs-air cooling delta, and accepted-share latency. This is also normal. Plan your ROI calculations around realised hashrate, not nameplate.
3. Power draw came in 3 to 6% higher than nameplate. S21 Pro spec is 3,510W; metered draw averaged 3,615 to 3,720W depending on ambient and PSU efficiency. Build this margin into your facility planning.

The scale-up decision

After 60 days of clean operational data, all three operators went back for a second order. The shapes of the orders diverged:

• Operator A doubled down on the same model, another 10× S21 Pro. Simple, predictable, monoculture fleet. Easy to manage.
• Operator B diversified, 7× S21 Pro + 3× Whatsminer M70S+ for direct platform comparison. Slightly more management overhead, but generated useful internal data on which platform delivered better realised efficiency in their specific facility.
• Operator C went sideways, 5 more S21 Pros plus their first 4× refurbished S19j Pros to fill out spare power capacity at lower capex. The refurbs broke even faster than the new units at their specific power rate.

None of these strategies is universally right. They reflect different operator priorities: fleet simplicity vs. comparative data vs. capital efficiency.

The mistakes worth avoiding

Three patterns we've seen scaled operators regret:

• Under-spec'd power infrastructure. If your circuit is rated for 12 kW continuous and you're running 11.2 kW of nominal load, you have no margin for the 5% power-draw overshoot. Operator B had to halt expansion at unit 14 to upgrade their distribution before going further. Plan circuits at 75% utilisation, not 95%.
• Pool concentration risk. Two of three started with 100% of their hashrate on one pool. When that pool had a stratum outage in October, all of them watched their daily output drop to zero for six hours. After that they split across two pools by default.
• Skipping the bulk-procurement conversation too early. All three eventually crossed 25 units in cumulative orders. None of them asked us about bulk pricing until they were ordering a single batch of 10+. Bulk pricing on the Bitcoin platforms kicks in earlier than most operators realise, usually around 5 units. Just ask.

Where they are now

All three operators have fleets between 35 and 60 units as of May 2026. Two are running monoculture S21 Pro fleets; one is running a mixed S21 Pro / Whatsminer M70S+ deployment with about 70/30 split toward the S21 Pro. None of them have switched to a competitor for additional units, which is the honest measure of whether we did right by them.

If this resembles your situation

If you're planning a first deployment, scaling from a small fleet, or trying to validate whether your facility power rate makes ASIC mining work, message us. We'll send you an ROI projection against your specific power rate and stack size, and we'll be honest if the numbers don't support what you're trying to do.

Talk to us about scaling →  ·  See our Bulk Procurement Process