Insights / Hosting
How to Start a Hosting Company: Product First, Servers Second
Write the offer, the buyer, and the prices before you buy servers. Buying a rack first is how you fund a catalog you cannot sell.
People ask how to start a hosting company as if the first purchase is a server. It is not. The first work is an offer a specific buyer will pay for, at a price that still covers support. A VPS (virtual private server) is a virtual machine with a set amount of CPU, RAM, and disk that a customer rents, usually by the month. I learned this order operating VPSGrid — U.S. Linux and Windows VPS — not from a forum thread about which switch to buy.
Hardware is easy to shop for. Salespeople will help you spend. The hard part is the paragraph on the website that says what you sell, who it is for, what happens when something breaks, and what you will not host. If that paragraph is weak, more RAM will not save you.
This is the order that has worked: offer, ideal customer, pricing, support model, then infrastructure direction, then billing. Legal and operations sit beside that path. They are not a reason to delay the offer. They are also not a reason to skip counsel.
Write the offer this week
Block a morning. Write one page, not a brand deck.
What do you sell in one sentence? Linux VPS, Windows VPS, reseller, managed WordPress on a CMS stack — those are different businesses. Pick one for version one. “We do everything” is how you inherit every ticket type before you have a process.
Who is it for? A freelance developer who needs a lab box is not a 20-person company that needs Windows, backups, and a human. Write the buyer as if you will speak to them on a call. If you cannot name where they spend time or what they already pay, you do not have an ICP (ideal customer profile — who you sell to). You have a guess.
What is not in version one? Write it down. No “unlimited,” no 14 locations, no white-label if you cannot support the downstream customer. Constraints are the product.
What happens when they open a ticket? Hours, channel, what you will actually do. “24/7” from a one-person company is a promise you will resent. Publish a window you can keep.
What will you refuse? Mining, spam, phishing, anything that turns your IP space into a problem. Put it in the AUP (acceptable use policy) before the first order, not after the first complaint.
Read the paragraph out loud. If you would not buy it, neither will a stranger. Fix the sentence before you request a quote for hardware.
Pick a buyer you can actually support
Hosting looks like a volume game from the outside. For a new company it is a support game. Your ideal customer is the person whose tickets you can close without destroying the week.
Developers who want cheap KVM and who know how to use a panel are one ICP. Small businesses who installed a CMS and need you to be the IT department are another. Agencies who resell are a third. Mixing them on one “unlimited support” line is how evenings disappear.
Geography is part of the ICP. U.S. VPS is a commercial choice: latency, data residency stories, payment rails, and who trusts the region. If your buyer is not in that story, do not spend extra for the location as decoration.
Write three names of companies or roles you could actually close this quarter — not “startups” in the abstract. If you cannot, your marketing will be generic and your pricing will chase comparison charts. Generic hosting is a race you will not win against incumbents.
A useful test: would you take a 30-minute sales call with this buyer without dreading the ticket that follows? If no, they are not your ICP yet. Either change the offer (managed vs unmanaged) or change the buyer.
Set prices before you buy infrastructure
Set three prices before you know your exact wholesale. Use conservative cost assumptions. You can refine when a quote arrives. You cannot refine after you have advertised a number that cannot fund support.
The method is in VPS pricing for new hosting providers: cost stack (IP, bandwidth, support hours, fraud, payment fees), then packaging. Do not skip it because you are “just starting.” Starting is when underpricing does the most damage.
Your first catalog should assume you will touch tickets. Price so one bad client does not erase the month. Cheap grids attract the customers a giant can serve with automation you do not have.
Windows vs Linux is a price and a product, not a toggle. Reseller is a different margin and a different abuse pattern. Do not launch all of them in week one to look complete.
| Decision | Do this first | Do not do this first |
|---|---|---|
| Offer | One sentence, one primary product | Twelve SKUs copied from a wholesale portal |
| ICP | A buyer whose tickets you can survive | “Anyone who needs a VPS” |
| Price | Three plans on a cost stack | A dollar under the loudest competitor |
| Support | A published window and a suspend story | “We’ll figure it out when tickets appear” |
| Infrastructure | A direction that fits the offer | A rack because it looks like a company |
Choose infrastructure after the offer is clear
Once the offer is real, pick a path for version one. Proxmox on rented dedicated hardware or colocation is a sane path for many new VPS providers. Wholesale you rebrand is another. They are different risks: control vs speed, capital you spend vs someone else’s outage.
You still need IP addresses, backups, and an abuse process on any path. A hypervisor without a restore test is a demo. A server without IP space is not a VPS product.
U.S. locations are a commercial choice. Dallas, for example, is a region I operate from because the product is U.S. VPS — not because a blog said “you must colo.” Match region to buyer.
Do not buy a full rack to feel serious. Buy or rent the minimum that can deliver the three plans. Empty units do not impress customers. They impress nobody except the person collecting the lease.
Automation comes after the happy path: order, provision, welcome, suspend. If you cannot suspend, you cannot enforce the acceptable use policy or non-payment. That is operations, not a feature request for later.
Panel choice — including WHMCS (the billing panel most hosting companies use to take orders, invoice, provision, and suspend) — comes after the catalog; what to decide before you install WHMCS is its own checklist. Catalog first, then the licence.
Write support as a product, not a personality
Support is not “we are friendly.” It is hours, channel, severity, and what you will not do.
Unmanaged VPS means the hypervisor and the network, not the customer’s CMS, plugin conflicts, or application code. Say so. If you want to sell help for WordPress or another CMS, that is a managed line with a different price — not a free extra on the cheap plan.
Publish the window. Email-only 10:00–18:00 in one timezone is honest. Slack-from-bed 24/7 is how founders burn out and start hating their own customers.
Write the suspend and terminate path. Non-payment, AUP breach, fraud. If you cannot bring yourself to suspend, you do not have a business. You have a hobby that hosts other people’s risk.
Staffing can be you at the start. That is fine. It is not fine to sell a support level that requires a team you do not have. Hire or contract when the working plan’s volume justifies it — not when a cheap plan’s tickets drown you.
Measure tickets per plan from week one. If the cheap plan consumes the week, raise it or cut what is included. Data from your own queue beats a competitor’s marketing site.
Legal and operations at a high level — not legal advice
I am not your lawyer. What follows is a map of topics founders trip on, not a substitute for counsel in the place you incorporate and the places you sell.
You will need an entity appropriate to where you operate. You will need terms of service, a privacy policy, and an acceptable use policy you are willing to enforce. You will need a way to take payment that can handle cards or invoices, chargebacks, and the customer you must refund or not refund according to your terms — not according to a mood.
Tax, VAT, and sales tax are fact-specific. Do not copy another host’s tax plugin settings and call it compliance. Ask someone licensed.
Abuse and copyright complaints will arrive. Know who receives them and how fast you can act. “We’ll read it later” is how IP space gets a reputation you cannot buy your way out of.
Data handling: what you log, how long you keep it, what you give law enforcement when required. Again, counsel — not a forum post.
Ops beside the legal stack: backups you have restored, a documented provision path, a documented suspend path, off-hours coverage that matches what you sold, and a record of who has access to the hypervisor. That is boring. Boring is how you still have a company after a bad weekend.
Why buying a rack first fails
A rack is a monthly number that starts immediately. Customers are not. You will pay for power, remote hands, and IPs while you still do not know which plan sells.
Founders buy racks because hardware feels like progress. A signed colo agreement is visible. An offer paragraph is not. Investors and friends may nod at a photo of gear. None of that is demand.
The failure pattern is predictable. Capacity sits half empty. The founder cuts prices to fill it. Cheap customers arrive with tickets and abuse. Support eats the week. The founder concludes “hosting is not profitable.” They never sold the working plan to the buyer they never wrote down.
A smaller start — even a single well-chosen dedicated or a modest wholesale pool — lets you test the paragraph, the price, and the ticket load. Scale hardware when the working plan is repeating, not when you are afraid of looking small.
There is a time to colo and a time to own more metal. That time is after you can describe utilization, ticket minutes, and which SKU prints money. Not after a weekend of reading hardware threads.
A one-week sequence you can actually run
- Day 1: Offer paragraph, ICP, three things you will not sell.
- Day 2: Three plans and a cost stack. Kill any plan you would resent supporting.
- Day 3: Support window, AUP draft, suspend rules. Send the legal topics to counsel; do not freeze the product while you wait for a perfect document.
- Day 4: Infrastructure direction — one path, minimum capacity, IP and backup story.
- Day 5: Billing path. What the customer sees after they pay. Provision and suspend on paper. Panel install only if the catalog is stable.
- Day 6: Website copy that matches the paragraph. No twelve-location fiction.
- Day 7: One test order, one test provision, one test suspend. Then you may talk about ads.
If you skip to ads before step 7, you are paying to find out that suspend does not work. That lesson is more expensive than a quiet week of tests.
I advise on this path because I operate a host, not because a rack is romantic. Hosting Launch Call is $199 for 60 minutes. Hosting Business Blueprint is $499 — the featured package — a 90-minute session plus a concise written plan. Hosting Launch Advisory is $999: that blueprint plus three 60-minute sessions within 45 days. None of these include implementation.
Book through the consultation page. Fees are non-refundable once paid, except if I cancel and we cannot agree a new date.
Start with the paragraph. The servers can wait until the paragraph can sell.
Questions people ask
How do I start a hosting company without a huge budget?
Write the offer, the buyer, and three prices you can support. Rent or wholesale a small amount of capacity. Do not buy a rack to feel legitimate.
Is a hosting business profitable?
It can be if you do not underprice support and abuse. It is a poor business if you compete only on the cheapest VPS against companies with more automation.
What should I decide before buying servers?
What you sell, to whom, at what price, what support window you will honor, and what you will refuse to host. Infrastructure is a direction after that.
Do I need WHMCS on day one?
You need billing, provisioning, and suspend. WHMCS is common. Install it after the catalog exists.
Linux or Windows VPS first?
Pick the buyer you can actually support. Windows is a different product: licences, images, tickets. Do not launch both as a checkbox in week one.
What legal and ops work is required?
You will need an entity, terms, privacy, AUP, a way to take payment, and a way to suspend. That is not a substitute for a lawyer. This article is not legal advice.
Why does buying a rack first fail?
Hardware does not invent a customer. You still have to price, support, and refuse abuse. Empty capacity with no offer is a lease, not a company.
What does hosting consulting include?
Strategy and, on Blueprint or Advisory, a concise written plan. Not server builds, WHMCS configuration, or procurement. Fees are non-refundable once paid except if I cancel and we cannot agree a new date.
Not legal advice. Company formation, contracts, tax, and acceptable-use enforcement are your counsel’s work in your jurisdiction.