Starting an internet service provider sounds audacious until you learn that thousands of small operators - many started by one technical person who was tired of the local cable monopoly - are running profitable networks right now. It is a real business with real capital costs and real regulatory homework, but it is learnable, and 2026 is arguably the best moment in a decade to start, with federal broadband money still flowing and fixed-wireless hardware better and cheaper than ever.
This is the business playbook, not a router tutorial. Here is what actually determines whether your ISP survives its first two years.
Step 1: find a market that deserves you
Successful small ISPs are almost never founded on "we'll compete with Comcast on price." They are founded on a specific underserved pocket: the rural valley the incumbents will not trench, the RV parks and campgrounds nobody serves properly, the apartment buildings stuck on one bad provider, the lake community with money and DSL.
Validation is unglamorous: drive the area, count rooftops, knock on doors, run a "would you pay $70/month for real internet?" survey. You want to see roughly 30-40% take-rate potential among reachable premises before you buy hardware. A tower that can see 200 rooftops at a plausible 35% take rate is a business; the same tower seeing 40 rooftops is a hobby.
Step 2: pick your technology honestly
Fixed wireless (WISP) remains the low-capital entry: a tower or grain-elevator lease, a few sectors, subscriber radios per rooftop. Modern gear (tarana-class, 60 GHz, or classic 5 GHz from Ubiquiti/Cambium/MikroTik) delivers 100-500 Mbps plans credibly. You can be live in months for $30k-100k.
Fiber (FTTH) is the forever asset with brutal upfront economics: construction routinely runs $15k-40k per mile plus drops. It makes sense where density is high, where you can win grant funding, or as the phase-two overbuild once wireless cash flow proves the market.
Most successful new operators in 2026 start wireless, bank the subscribers, then selectively overbuild fiber - often with BEAD or other grant money covering the construction. If fiber is your path, our primers on the access architectures (PPPoE vs DHCP vs IPoE) apply to both worlds.
Step 3: the regulatory homework (US edition)
Less scary than expected, but non-optional:
- FCC Form 477 / BDC filings - twice-yearly broadband coverage reporting once you have subscribers. The Broadband Data Collection map data also determines grant eligibility, so file accurately.
- Broadband nutrition labels - consumer-facing price/speed labels are now required at point of sale for ISPs of every size.
- CPNI compliance - annual certification that you protect customer proprietary network information.
- CALEA - lawful-intercept capability obligations apply to facilities-based broadband providers; small operators typically satisfy this with a trusted-third-party arrangement.
- Spectrum - unlicensed 5 GHz needs no license; CBRS and licensed microwave links involve coordination. Tower leases and insurance will be your bigger paperwork.
Not in the US? The shape is similar everywhere: a telecom registration, coverage reporting, consumer-protection rules, and lawful-intercept obligations. Budget a few thousand for a telecom-savvy consultant in year one; it is cheap insurance.
Step 4: the money
A realistic opening budget for a small fixed-wireless launch:
- Backhaul/transit: $500-2,000/month for a fiber DIA circuit or licensed link to an IX - your single most important vendor decision
- First tower: $5k-15k (lease, sectors, mounting, power, backhaul radio)
- CPE: $100-300 per subscriber (radio + router), recovered via install fees or amortized
- Truck, tools, test gear, climbing cert or contractor: $10k-30k
- Working capital: 6-12 months of runway, because subscribers arrive slower than the spreadsheet promised
With $70/month ARPU, 60-70% gross margins after transit, and CPE payback inside a year, a single well-placed tower with 70-100 subscribers covers a modest operation. The compounding killer is churn from bad service - which is an operations problem, which brings us to the part founders skip.
Step 5: the software stack from day one
Here is the pattern that kills more young ISPs than bad radios: the founder wires the network beautifully and runs the business on spreadsheets, a personal Venmo, and memory. At 30 subscribers it works. At 150, billing errors, forgotten suspensions, and support chaos eat the margin - and retrofitting systems at 150 subscribers is ten times the pain of starting right at zero.
From subscriber one, you want one system where:
- Signups become services - plan, address, equipment, and PPPoE/RADIUS credentials provisioned together, not five tools glued by copy-paste (what RADIUS does if that layer is new to you)
- Invoices collect themselves - autopay with saved cards, dunning emails, and automatic suspension for non-payment with restore on payment
- Customers self-serve - a portal for bills, payments, plan changes, and tickets, because every call you do not take is margin
- The network is visible - monitoring, a network map, and IP address management that is not a spreadsheet
- Support has a system - tickets tied to the customer record, not a shared inbox
This is precisely what ISPbox packages for new operators: the whole stack at $25/month while you are small, scaling only as you do - with a 7-day trial to have it running before your first install. Start on it at subscriber zero and you never meet the spreadsheet phase at all.
The realistic timeline
Months 1-3: market validation, tower lease, backhaul contract, regulatory filings. Months 3-6: first tower live, first 20-50 subscribers (friends-and-family pricing buys patience and testimonials). Months 6-18: second and third sites, word of mouth compounding, maybe your first grant application. Year two: you either have a real business - or a very expensive lesson in why the take-rate survey mattered.
FAQ
How much does it cost to start an ISP? A small fixed-wireless ISP typically launches on $30k-100k including the first tower, CPE inventory, and working capital. Fiber-first builds start in the hundreds of thousands due to construction costs of $15k-40k per mile.
Do I need a license to start an ISP in the US? There is no general "ISP license," but you have regulatory obligations: FCC broadband data (BDC) filings, broadband labels, CPNI certification, and CALEA lawful-intercept capability. Unlicensed spectrum requires no license; some links need coordination.
Is starting a WISP still profitable in 2026? Yes, in the right market: underserved areas with 30%+ achievable take rates, $60-80 ARPU, and 60-70% gross margins after transit are common. The failures are usually operational (churn, billing chaos) rather than technical.
What software does a new ISP need? Billing with automatic suspension, RADIUS/PPPoE provisioning, a customer portal, ticketing, and network monitoring - ideally as one platform from day one rather than tools glued together later.
