The Financial Case for Sustainable GPU Infrastructure: Why Green Pays

Sustainability and Profitability: A False Dichotomy

When I founded AI Green Bytes, I knew that building Europe's sovereign green cloud for AI wasn't just an environmental imperative—it was a financial one. The data centre industry has operated on a flawed assumption for decades: that sustainability comes at a cost premium. We set out to prove otherwise.

Today, I want to share the financial case for sustainable GPU infrastructure, backed by real numbers from our operations and deployment plans.

The Cost of Traditional Air-Cooled Data Centres

Traditional air-cooled GPU data centres operate with a Power Usage Effectiveness (PUE) of 1.67 to 2.0. This means that for every watt of compute power delivered to a GPU, you're consuming 1.67 to 2 watts of total facility power. The rest goes to cooling, power distribution, and overhead.

For a 1 MW GPU cluster running 24/7:

• Annual energy consumption: 8,760 MWh (at PUE 1.85)
• Annual energy cost (at €0.12/kWh): €1,051,200
• Annual water consumption: 2.5 million gallons (for evaporative cooling)
• Water cost and treatment: €50,000–€100,000 annually
• Total OpEx for cooling alone: €1.1–€1.15 million per MW

This is before factoring in real estate, maintenance, staffing, and heat disposal costs.

The AIGB Advantage: Immersion Cooling Economics

Our single-phase immersion cooling technology achieves a PUE of 1.03 to 1.15. For the same 1 MW cluster:

• Annual energy consumption: 9,000–10,000 MWh (at PUE 1.09)
• Annual energy cost (at €0.12/kWh): €108,000–€120,000
• Annual water consumption: Zero (immersion fluid is recirculated)
• Water cost and treatment: €0
• Total OpEx for cooling: €108,000–€120,000 per MW

Direct savings: 90% reduction in cooling costs.

But the story doesn't end there.

Revenue from Heat Reuse

Immersion cooling generates an additional financial advantage that traditional facilities cannot access: heat reuse.

Our systems capture nearly 100% of the waste heat from GPU workloads. A 1 MW GPU cluster generates approximately 900 kW of recoverable heat continuously. In partnership with local district heating networks (as we're implementing in Paris, Oslo, and Frankfurt), this heat can be sold or bartered at €0.08–€0.15 per kWh.

Annual heat revenue per MW: €630,000–€1,050,000

This transforms a cost centre (cooling) into a revenue stream. Over a 10-year facility lifespan, heat reuse alone generates €6.3–€10.5 million in additional revenue per MW of installed capacity.

Space and Density Economics

Immersion cooling enables 65% higher rack density compared to air-cooled facilities. This means:

• Reduced real estate footprint: Fewer square metres of facility required
• Lower construction and maintenance costs: Smaller buildings, simpler infrastructure
• Faster deployment: Modular immersion tank design accelerates time-to-revenue

For a 10 MW facility:

• Air-cooled footprint: 2,000–2,500 m²
• Immersion-cooled footprint: 700–800 m²
• Real estate savings: €1.5–€3 million in construction costs alone

The Total Cost of Ownership: 10-Year Comparison

Let's model a realistic 10 MW facility over 10 years:

This represents a 68% reduction in total cost of ownership compared to traditional air-cooled infrastructure.

Competitive Pricing Without Margin Compression

These cost advantages allow AIGB to offer GPU compute at 30–40% lower prices than traditional cloud providers, while maintaining healthy margins. This creates a powerful market dynamic:

1. Customers get lower costs — making AI workloads more accessible
2. AIGB achieves better unit economics — enabling faster scaling and profitability
3. European AI companies gain competitiveness — reducing reliance on US hyperscalers

The Sovereign Advantage

Beyond the financial metrics, sustainable infrastructure in Europe creates strategic value:

• Data sovereignty: GPU compute stays under European jurisdiction
• GDPR compliance: Built-in from day one, not retrofitted
• Regulatory alignment: ISO 27001 and SOC 2 Type II certified
• Reduced geopolitical risk: No reliance on foreign cloud providers

For European enterprises, this combination of cost, compliance, and sovereignty is unprecedented.

Looking Forward: Paris and Beyond

Our Paris facility, opening in April 2026, will validate these economics at scale. With 282 GPUs and 3 immersion tanks, Paris will demonstrate that sustainable infrastructure is not a niche play—it's the future of competitive GPU compute.

We're already planning 16 sites across Europe. Each one will benefit from the economies of scale we've built into our immersion cooling platform. By 2028, AIGB will operate 80+ MW of immersion-cooled capacity, making us Europe's largest sustainable GPU infrastructure provider.

The Bottom Line

Sustainability and profitability are not opposing forces. They're aligned incentives. By choosing immersion cooling, edge deployment, and heat reuse, we've built a business model where doing the right thing for the planet is also the right thing for our customers and shareholders.

The financial case for sustainable GPU infrastructure is clear. The question is no longer whether it's viable—it's whether traditional air-cooled facilities can survive the transition.

---

Stig Torvund
CEO & Co-founder, AI Green Bytes
February 2026