From Air to Immersion: How NFN8 Cut CapEx and Stabilized Uptime

How NFN8 is using immersion cooling to reduce ASIC count, lift hashrate, and maximize uptime—with lower CapEx and better TCO.

Introduction

In Bitcoin mining, small swings in reliability compound fast: a few percentage points of downtime translate into permanently lost revenue, emergency repairs, and missed energy windows. For NFN8, where industry veteran JT Gulledge II has led operations since 2017, that math forced a hard pivot. After years of wrestling with the limits of air cooling—even in cold climates—NFN8 transitioned to Rosseau’s immersion cooling platform to regain control of uptime, cost, and growth. This article distills the operator’s view: the specific pain points that pushed the change, the selection process, and the measurable outcomes of immersion—told in practical, verifiable terms for owners, engineers, and finance teams.

Background

NFN8’s path mirrors much of the industry. The team started modestly, scaled through hosting with major providers, and then got caught when counterparties over-levered and entered bankruptcy. With more than 20 megawatts (MW) to energize on a tight timeline, NFN8 moved to self-developed sites and direct operational control. Early buildouts leaned on air-cooled fleets (notably S19-class units). As the market shifted to denser, higher-power devices like the S21 Pro, the air approach that once penciled out became a bottleneck. Hardware failures spiked. Retrofits consumed capital. The “rebuild every generation” cycle was eroding returns.

The Challenge: Air Cooling’s Hidden Taxes

Operators know the pain, but it’s worth making the costs explicit. Hardware instability from thermal cycling. In hot-cold environments or swing seasons, air systems expose hash boards to rapid temperature changes that fatigue solder joints and components. JT tracked days in Texas where 17–20 boards failed in a single session due to thermal cycling. The repair loop was punishing: turnaround times, shipping, and after-repair failure rates that undermined confidence. Operational inefficiencies that compound. Air systems bring labor and variability. In one month, NFN8 replaced ~1,000 fans—an effort JT now avoids by eliminating fans entirely. Power draw also swung widely: a nominal 10 MW setup might drift between 7.5 and 10.2 MW depending on weather, dust loading, and fan behavior. Variability meant inconsistent hashrate, tougher energy management, and more operator “babysitting.” CapEx drag and retrofit risk. Building air infrastructure rang in at roughly $250K–$300K/MW. Next-gen ASICs then demanded costly retrofits: redoing power distribution units (PDUs), slot spacing, and airflow. Warranties often failed to protect cash-flow, given shipping delays and dead-on-arrival replacements that pushed revenue out by months. In short: air lacked the adaptability and stability required for long-term, multi-generation growth.

The Decision: A Different Equation with Immersion

Hydro cooling and next-gen air were evaluated. Immersion had long been on NFN8’s radar, but historical perceptions of high infrastructure cost created hesitation. Rosseau’s system—and the way it is deployed—changed the equation on three fronts:

1. Effective cost per TH (terahash) drops. By enabling stable overclocking, Rosseau immersion reduces the number of ASICs per MW. NFN8 runs 117–120 units per MW in immersion versus 250–270 in air to hit comparable (or better) hashrate output. Fewer ASICs purchased up front = lower effective $/TH and a materially smaller fleet to maintain.
2. Stable overclocking on modern high-density ASICs. With WhatsMiner M66S+, NFN8 runs devices at ~8,000 W versus a ~5,500 W stock setting—yielding ~40–50% hashrate uplift per device (from ~294 TH/s stock into the ~422–430 TH/s range). The key is stable uplift—no boom-bust thermals, no hashboard failures, tighter power draw, and consistent output.
3. Operational simplicity and predictability. Rosseau’s immersion keeps power draw within ~3% of target, dramatically reducing the day-to-day variability that plagues air. In practice, this means smoother energy scheduling, steadier revenue, and fewer unexpected operational challenges.

Outcomes: The Numbers That Matter

1) Upfront CapEx savings—fewer machines, less money. On a 3 MW build, NFN8’s fully built and filled immersion came in at more than 20% below the capex for air. The delta is dominated by eliminating excess ASIC purchases: because immersion lets each device contribute more hashrate reliably, you buy fewer to reach the same (or higher) output. With ASICs at $4,000–$5,000 each, this becomes millions saved on multi-MW rollouts. 2) Operational stability and uptime. With Rosseau immersion, NFN8 reports 99%+ uptime without the “mental gymnastics” that air required to manage heat spikes, fan failures, and dust. That steadiness converts directly to revenue: downtime isn’t just delayed cash—it’s lost cash. The combination of higher sustained hashrate and fewer interruptions drives better realized revenue per MW. As JT put it: “Uptime is everything. Downtime means lost revenue that can never be recovered, and you’re out of pocket for parts, labor, and shipping.” 3) Power predictability within ~3%. Where air sites wandered (e.g., 7.5–10.2 MW on a “10 MW” block), immersion holds a tighter band around the plan. That’s invaluable for curtailment programs, day-ahead planning, and minimizing imbalance penalties. 4) Long-term CapEx avoidance: extended ASIC life. By removing the main failure accelerants—thermal cycling, dust/debris, and fan failures—NFN8 expects 5–6 years of useful life from immersion-operated miners, versus 2–3 years typical in air. Longer life extends replacement cycles, cushions halving periods, and improves TCO. 5) Scalability without serial retrofits. The leading power density of Rosseau’s immersion infrastructure provides a solution that handles today’s most powerful miners and the increasingly powerful next-generation miners. NFN8 eliminates the “rip-and-replace” tax each time OEMs release a higher-power flagship.  

The Partnership Factor

NFN8 is blunt: the technology matters, but so does the partner. Rosseau supported site design, deployment, operator training, and ongoing optimization. That collaboration is what converted “on paper” benefits into day-one stability and week-over-week repeatability. For operators and financiers, that’s the difference between a promise and a plan.

Results in Context

When you reduce ASIC count per MW, you also shrink your future risk surface: fewer PSUs, fewer boards, fewer connectors, less cumulative failure probability. Combine that with steadier power draw, and a mining site becomes easier to forecast and finance. The result is not just a faster payback on the initial project; it’s a more credible platform for the next project.

Conclusion

NFN8’s move to Rosseau immersion transformed mining from a high-variance grind into a stable, scalable operation. The case for immersion isn’t marketing—it’s math: fewer machines to buy, higher stable output per machine, tighter power control, and far less downtime. For operators facing the same constraints—next-gen ASIC heat, curtailment, and widening maintenance loops—the upgrade path is clear. As NFN8 puts it: “We’re tired of solutions that cost more and deliver less. We want confidence in the outcome.” Immersion—executed with the right partner—delivers that confidence.