Snapshot

  • Ticker: WYFI (Nasdaq, IPO August 2025)
  • Bucket: Data-Center Operators (HPC Pivots)
  • Q4 2025 fund position: $27.8 M (1.76 M sh) — 0.50 % of 13F
  • Heritage: Spun out of Bit Digital (BTBT); BTBT retains ~71.5 % post-IPO
  • HQ: New York, NY

Business Overview

WhiteFiber is the AI/HPC infrastructure spin-off from Bit Digital, Inc. — Bit Digital separated its growing HPC business from its legacy Bitcoin mining operations to allow each to be valued cleanly. The IPO in August 2025 raised ~17/share.

WhiteFiber operates two product lines:

  1. Cloud Services (the larger and more mature segment) — owned NVIDIA H100/H200/B200 GPU capacity rented under multi-year contracts and on-demand to foundation-model and AI startup customers.
  2. Colocation (the growth bet) — purpose-built HPC data centers in the US (North Carolina) and Canada (Montreal) hosting customer-owned GPU deployments.

Geographic footprint includes Iceland (legacy Bit Digital site, cheap geothermal/hydro power, EU data residency) and an expanding US/Canada presence.

Financial Trajectory

Metric (USD M)FY24 (precursor, in BTBT)Q2 2025Q3 2025
Total revenue45.718.7 (+48 % YoY)~20.18
— Cloud services16.6 (61 % GM)
— Colocation1.7 (60 % GM)
Adjusted EBITDA3.3
Net loss−8.8

Q2 2025 Adjusted EBITDA was depressed by IPO-related G&A jumping from 15.5 M; ex-IPO costs the underlying business is profitable on an EBITDA basis. The 60 %+ gross margins on both cloud and colocation segments point to genuine operating leverage at scale.

Balance Sheet & Capital

  • Cash post-IPO: ~$183 M gross proceeds before fees
  • CAD 43.8 M USD) undrawn debt facility with RBC for additional growth capital

Operational KPIs

  • GPUs under contract (June 2025): ~4,000
  • NC-1 (North Carolina): 1 M sq ft, 24 MW initial phase, target online Q1 2026
  • MTL-3 (Montreal): 5 MW IT load Cerebras wafer-scale deployment; revenue Q4 2025
  • Iceland sites: legacy capacity for cloud workloads with EU data residency

Why It Fits the Thesis

WhiteFiber is the only pure-play AI infrastructure name in the bucket — no residual Bitcoin mining, no legacy infrastructure to convert. The fund’s small position reflects that it is a smaller, newer, more speculative version of the CoreWeave thesis at much lower scale. Holding it provides exposure to “second-tier” GPU clouds that could consolidate, be acquired, or scale via future contract wins.

The Iceland location is strategically interesting: cheap geothermal/hydro power, naturally cold ambient (free cooling), and EU-jurisdiction data residency for European AI customers — a niche that pure-US operators cannot easily fill.

Position History in the Fund

QuarterPosition
Q4 2025New, 1.76 M sh

Risks

  • Sub-scale. Too small to compete for hyperscaler-class contracts; restricted to mid-market AI startups.
  • NVIDIA allocation risk. Smaller buyers get later access to scarce GPU generations.
  • Iceland power competitive intensity — multiple AI infrastructure buildouts compete for the same Icelandic generation pool.
  • Bit Digital control overhang — 71.5 % parent ownership creates conflicts and float constraints.
  • No marquee anchor contract as of Q4 2025; thesis depends on landing one in 2026.

Sources