The 18-Month Head Start: Why Retrofitting Beats Waiting For New Builds.
Everyone's building. AWS, NEXTDC, CDC, AirTrunk—billions of dollars going into greenfield data centers. New is better, right? Purpose-built for liquid cooling, maximum efficiency, perfect layouts. Why retrofit old infrastructure when you could build new? Here's the part nobody's saying out loud: those greenfield builds won't be ready until 2027-2028. And in the meantime, there's an 18-month window where retrofitted facilities own the market.
The Construction Reality Check
Let's walk through what actually happens when you build a new data center from scratch:
Months 1-6: Just getting started
Find and buy land: 2-3 months
Environmental assessments: 2-4 months
Planning approvals: 3-6 months
Grid connection application: 3-6 months
That's half a year before you even break ground. And that's assuming nothing goes sideways - no environmental issues, no community pushback, no grid constraints.
Months 7-12: Design and prep
Detailed engineering: 3-4 months
Order long-lead equipment (switchgear, generators, cooling): 6-9 months
Contractor selection: 2-3 months
Months 13-24: Actual construction
Foundations and structure: 4-6 months
MEP installation: 6-9 months
Electrical and IT: 4-6 months
Testing and commissioning: 2-3 months
That comes to a total of 24-30 months from "let's do this" to "first customer powered on." And that's for experienced operators. First-timers? Add another 6-12 months. Now, let’s compare that to retrofitting existing infrastructure timeline;
Month 1: Assessment and design
Thermal assessment: 2-3 weeks
Engineering design: 4-6 weeks
Permits (usually minor): 2-4 weeks
Months 2-3: Getting equipment
CDUs and cooling gear: 8-12 weeks
Electrical upgrades: 6-8 weeks
Controls: 4-6 weeks
Months 4-6: Installation
Installation (often done in phases while facility stays operational): 8-12 weeks
Testing and commissioning: 3-4 weeks
Customer migration if needed: 2-4 weeks
That comes to a total of 6-9 months from decision to first high-density racks operational. That's a 15-21 month head start. Time-to-market isn't just bragging rights., it's competitive positioning.
Here's what's happening right now:
Today (early 2026):
Enterprise customers are planning AI deployments for Q3-Q4 2026
Hyperscalers are scouting capacity for 2026-2027
Greenfield builds announced today deliver in 2027-2028
If you retrofit now:
You have capacity in Q3-Q4 2026
You capture customers before greenfield competitors launch
Customers deploy, build dependencies, create switching costs
By 2028 when new builds come online, your racks are fully leased with multi-year contracts
Customer Lock-In Is Real, once a customer deploys in your facility, they're sticky. Really sticky.
Why? Because moving is brutal:
Downtime during migration
Reconfiguration and testing
Established network connections (cross-connects, direct links)
Compliance and security certs are facility-specific
Operational procedures built around your systems
Industry churn rates run 5-8% annually for established customers. Win a customer in 2026, you probably keep them 5-10 years - even when "better" facilities open in 2028.
The Premium Pricing Window - supply and demand, mate. Melbourne's pipeline is 97% pre-committed. Sydney has massive demand with limited supply until major builds finish.
Right now, operators with high-density capacity can charge premium prices.
Current market:
Standard density (12-15 kW): $140-$160/kW/month
High density (50-80 kW): $180-$220/kW/month
Ultra-high density (100+ kW): $250-$300/kW/month
That premium won't last forever. Once greenfield capacity floods the market in 2027-2028, pricing normalizes. Quick math: Deploy 1 MW at $220/kW for 18 months before competition arrives = $3.96M. Wait for greenfield and deploy at normalized $180/kW = you left $3.96M on the table. Per MW.
There's another angle here - Capital Efficiency
Greenfield:
$15-25M per MW of capacity
24-30 months to revenue
Capital tied up for 2+ years
Retrofit:
$3-8M per MW
6-9 months to revenue
Generating returns within a year
From an IRR perspective, it's not even close. Even if the greenfield facility is technically "better," the retrofit generates 18-24 months of revenue before the greenfield opens its doors.
Now, let’s look at the risk profile.
Greenfield carries serious risk:
Construction delays
Permitting issues
Grid connection problems
Market demand shifts (will customers still want this in 2.5 years?)
Retrofits have way lower risk:
Building already exists
Operating facility (permits are modifications, not new approvals)
Grid connection exists (upgrades faster than new)
Can pre-lease capacity before retrofitting (validate demand first)
So who’s making the move? Smart operators are recognizing this window:
Established colos retrofitting to compete with greenfield
Enterprise operators modernizing rather than waiting for hyperscaler capacity
Regional operators adding liquid cooling to differentiate
The operators waiting are mostly those without existing facilities or those who don't understand the timing play.
The 2028 Reality
By 2028, the greenfield wave hits. Massive, purpose-built, liquid-cooled facilities launch. At that point, your retrofit advantage disappears.
But if you retrofitted in 2026:
You've had 18-24 months of premium pricing
Racks are fully leased with multi-year contracts
You've generated $10-30M in revenue that paid for the retrofit multiple times over
Now you're evaluating your next move (expand, build your own greenfield, or harvest cash)
The operators who waited until 2028? They're competing with brand-new state-of-the-art facilities on day one, with no installed base, no revenue, and no time advantage. The 18-month window isn't about avoiding greenfield forever. It's about capturing a time-based competitive advantage that exists right now and won't exist in 24 months.
If you have an existing facility with any viable path to retrofit, ask yourself: can you afford to sit idle for two years while competitors capture the market? Because by the time you're ready, the window's closed. And you're competing against operators who already won.
