A server in ColoTycoon pays for itself when the customer payments it collects exceed the costs of keeping it racked, for long enough to cover what you paid for its parts. Every one of those numbers is simulated hour by hour, on the real clock, even while you are offline. This guide explains the billing clock, walks through the money in and the money out as the game calculates them, works through one complete example and then gives an order of work for reaching payback.

The billing clock

Recurring prices are quoted per month, like real hosting prices, but a billing month passes every 24 real hours. Rent, customer payments, power, licenses, salaries, insurance premiums and IP lease fees all run on this clock, so a "$15 a month" license is paid every real day and a "$20 a month" customer pays you $20 every real day. Hardware, land, clicks and blog posts are one-off prices and are paid once at face value.

That makes payback quick to read: a server that makes $200 a billing month more than it costs earns $200 a day.

How customers pay

Customers pay a month in advance. A new customer pays the first month the moment they order, and the money is in your account at once. After that each subscription renews every billing month, on the hour it signed up. If the server is down when the month is due, the customer holds the payment and pays once it is up again, if they have not left by then. A server is up when it is racked, powered, cabled to the uplink and not shut down; a scheduled maintenance window still counts as up. Nobody gets money back for the rest of a month when they cancel, so downtime costs you mostly through the customers it drives away.

Your monthly recurring revenue (MRR) on the Sales page is the sum of all active subscriptions' monthly prices, which is what they pay per billing month.

Where orders come from

There is no free traffic. Visitors come only from the three channels in the Marketing panel on the Sales page. Every five minutes the game brings in a twelfth of the hour's visitors and turns them into orders:

orders = visitors × 2% × reputation factor × demand

  • The base conversion is 2%.
  • The reputation factor is your reputation divided by 50, held between 0.5 and 1.5. A new company starts at 50, so the factor starts at 1.0.
  • Demand is the average demand this week of the datacenters your sales groups' servers are racked in, between 0.3 and 2.0. The Colocation table shows it as Low, Normal, High or Very high.

Fractions carry over from slice to slice, so a slow trickle of visitors still produces orders eventually. Very happy customers (85 or more) also bring friends, a small number of extra orders that cost nothing. If an order arrives and nothing in stock matches, it is lost and the operations log says so.

What each channel costs per order

At a reputation of 50 and normal demand (1.0):

Channel Price Visitors Orders Cost per order
PPC $0.60 per click, from a daily budget charged every five minutes 1 per click 0.02 per click $30
SEO $150 per level per billing month, levels 0 to 5, charged every five minutes 10 per hour per level about 4.8 per billing month per level about $31
Blog post $1,250 once, one per day none of its own: multiplies PPC and SEO visitors, up to ×1.72 depends on your PPC and SEO falls the more you advertise

Blog posts multiply what PPC and SEO bring: ten posts make every ad and SEO visitor count about 1.35 times, forty about 1.67 times, and the ceiling is 1.72. PPC is the fastest to switch on and off and the most direct way to fill a new server quickly; SEO costs about the same per order but cannot be scaled past level 5. Posts are a long-term investment for a company that already advertises: the more you spend on ads and SEO, the more each post is worth. SEO and blog posts keep sending visitors even when you have nothing in stock, and those orders are lost. If you cannot pay the SEO retainer or the PPC spend in a given five-minute slice, that slice's visitors from it do not arrive. A budget whose daily cost is more than your cash is refused when you save it.

Which offer a buyer picks

Each order comes from a customer persona with a preferred product: bloggers and shops prefer hosting, agencies VPS, software and gaming customers VPS or dedicated, and enterprises dedicated. When you have several plans in stock, the buyer chooses among them with weights for their preference, the demand where each product is racked (squared, so busy datacenters stand out), the memory speed score of its servers (also squared, so DDR5-6400 memory wins about 1.77 times as often as DDR5-4800) and price (a plan at four times the price is picked half as often).

The buyer then compares the price with the market reference price for what the plan offers. At or below the reference everyone buys; above it, only (reference ÷ price)² of buyers do, so 1.5 times the reference keeps 44% of the orders, and above that nobody buys. A VPS is referenced at $3 per vCPU, $3 per GB of memory and $0.05 per GB of disk a month, and faster memory raises that reference by half its extra speed (16.5% at DDR5-6400). The full table is in selling hosting.

What a server costs

Rent

Colocation rent is the unit price times the number of units, every hour. Bot datacenters charge from a base of $1.60 per unit per hour, which is $38.40 per unit per billing month, scaled by the building's tier, so 2 units at the base price cost $76.80 a billing month. Optional extras add to it: a redundant uplink adds 50% of the base rent, a B-feed adds 30%, a fiber hand-off costs $300 to $1,200 a month, extra IPv4 addresses cost $0.75 each per month and a cabinet lock $25 a month. See colocation.

Power and heat

Each unit rented from a bot provider includes 100 W of power. A server draws half of its peak power when idle and all of it when its threads or memory are fully sold (a dedicated customer counts as 70% load). Peak power at the wall is what the parts draw divided by the efficiency of the power supplies: 0.85 for Bronze, 0.90 for Gold, 0.94 for Platinum and 0.96 for Titanium.

Power above the included amount is billed at 18¢ per kWh of a billing month's energy. A billing month is 720 hours squeezed into 24, so every real hour stands for 30 hours of draw: 100 W over the included power costs 54¢ an hour, or $12.96 a billing month.

Heat matters as much as the bill. A cabinet's cooling equals the power it includes, and every watt drawn becomes heat. Above a heat-to-cooling ratio of 1.0 the cabinet runs warm and parts fail 1.8 times as often; above 1.25 it overheats and parts fail 4 times as often, and smoke alarms become 3 or 8 times as likely. Renting an extra unit adds 100 W of both power and cooling. Rack cooling units add more cooling but draw power themselves. The details are on the power and cooling page.

Bandwidth

Each unit rented from a bot provider includes 100 Mbps. The game counts average traffic of 12 Mbps per VPS customer, 1 Mbps per hosting account and 60 Mbps per dedicated tenant, plus your website visitors. Traffic above what your leases include costs $0.50 per Mbps per billing month, taken in hourly shares.

Licenses

Software License
Server Linux LTS, Container Host OS free
Hypervisor OS $15 per socket per month
Hypervisor OS (legacy branch) $9 per socket per month
Commercial Server OS $2 per core per month
Hosting control panel and antivirus $55 per server per month

The first month is paid at install (or, for the hosting panel, when the server joins a hosting group), and licenses renew every billing month after that. If a renewal cannot be paid, the license is suspended.

Matching parts to the product

What a server can sell depends on which resource runs out first.

  • VPS: the hypervisor sells two vCPUs per hardware thread. Each plan takes its vCPU from that pool, its RAM from memory minus 2 GB reserved for the hypervisor, its disk from usable disk minus the OS, and one IPv4 address. A cabinet starts with 24 addresses, so the 25th VPS in that cabinet needs a block from the IP market. Memory is usually the limit, so buy it in proportion to the plans you sell. Disk speed is the hidden limit: each VPS uses one unit of disk I/O, and a SAS SSD serves 16, an NVMe drive 40 and a SAS disk only 3, fewer in a mirror. Customers on overloaded disks grow unhappy and leave.
  • Shared hosting: each account uses 0.05 vCPU and 0.1 GB of RAM and shares the server's address. An 8-thread server with 8 GB of memory holds 80 accounts by memory, unless disk runs out first. Memory is usually the limit, so hosting servers want cheap threads and a lot of RAM.
  • Dedicated: one customer takes the whole machine and one IPv4 address at the monthly price you set on the server.

Wear follows load: a busy server's parts age up to 60% faster and fail more often, so a fully sold server needs spares sooner. One real hour ages racked hardware by 48 hours of rated life.

A worked example

A VPS server in a Tier 2 building at the base rent of $1.60 per unit per hour, with normal demand and a reputation of 50. All prices are vendor list prices.

Part Price
1U Micro Chassis $650
8-Core / 16-Thread Processor $700
2 × 32 GB DDR5-4800 ECC Memory $1,700
2 × 960 GB SAS SSD, as RAID 1 $520
400 W Power Supply (Gold) $110
8-Port Switch $60
Total $3,740

What it can sell with the Hypervisor OS installed:

  • 16 threads give 32 vCPUs.
  • 64 GB of memory minus 2 GB for the hypervisor leaves 62 GB.
  • RAID 1 keeps one disk's 960 GB, minus 20 GB for the hypervisor: 940 GB.

On the VPS M plan (2 vCPU, 4 GB, 40 GB at $20, exactly its reference price), memory runs out first at 15 customers. vCPU would allow 16 and disk 23. The cabinet's 24 included addresses are enough. The two SSDs in RAID 1 serve 2 × 16 × 0.6 ≈ 19 VPS customers, so 15 stay below the 80% mark where disks start to feel busy. A full server brings in 15 × $20 = $300 a billing month.

Running costs per billing month:

Cost Per billing month
Rent, 2 units $76.80
Hypervisor OS, 1 socket $15.00
Total $91.80

Power and bandwidth stay inside what the lease includes. The server peaks at 164 W inside (30 W chassis overhead, 25 W chassis, 95 W processor, 4 W per memory module and 3 W per SSD), which is 182 W at the wall through a Gold supply. With the switch's 12 W the cabinet draws 194 W of its 200 W, so it stays below Warm. Fifteen VPS customers average 180 Mbps of the 200 Mbps included.

Filling the server takes 15 orders, which is 750 PPC clicks or $450. A daily PPC budget of $150 buys about 10 clicks an hour and fills it in about three days. Background churn then takes about 0.5 customers a billing month (3.5% of 15), and replacing them costs about $16 of PPC.

That leaves $300 − $91.80 − $16 ≈ $192 a billing month. The $3,740 of hardware and $450 of marketing are paid back after about 22 billing months, a little over three real weeks, as long as the server stays up and full.

Step 1: Price the hardware against its capacity

Before you buy, write down the most customers the server can hold and multiply by the plan price you intend to charge. That is the server's ceiling MRR. Keep the plan price at or near its market reference, or a share of the buyers will walk away. Divide the cost of the parts by (ceiling MRR minus monthly running costs) to get the best-case payback in billing months, which is the same number in real days.

Step 2: Rent enough units for the power you draw

Count 1 unit for the server, 1 for the switch, and check the Power and Cooling gauges on the Racks page once it is running. If the cabinet shows Warm or Overheating, expand the lease by a unit before you add customers.

Step 3: Put stock on sale before you buy traffic

Create the service group, add the server and add at least one enabled plan with stock. Marketing spent while nothing is in stock is wasted.

Step 4: Start marketing in proportion to free capacity

Start with a PPC budget sized to the free slots. Once ads bring a steady flow, add blog posts: each one makes every ad and SEO visitor go further, for good. Watch Orders so far and Visitors / hour on the Sales page, and watch the operations log for orders lost to missing stock or to price.

Step 5: Protect the reputation you are building

Reputation rises by 0.08 points for every hour no customer's server is down and falls by 0.3 for every hour one is. About 3.5% of customers leave in a billing month for no reason of yours; that rate is 4 times higher while their uplink is saturated, every hour their server is down gives each of them about a 3% chance of leaving, and unhappy customers leave up to 4 times as often. Watch the mood faces in the customers list and the Customer voice feed on the Sales page: complaints about slow disks, a crowded server or the price come before the cancellations. A high reputation keeps more of them. The selling hosting page explains churn in full, and the surviving incidents guide covers the outages that cost the most.