Ask ten IT providers how much bandwidth a business needs and you’ll get ten different answers — usually followed by a quote for the biggest circuit they sell. The honest answer is that bandwidth needs are calculable, not guessable. They depend on how many people you have, what applications they run, whether you use VoIP and video conferencing, and how much of your work now lives in the cloud.
At COMNEXIA, we’ve been designing business networks in Atlanta since 1991, and we’ve watched bandwidth requirements grow from a shared 56k modem line for an entire office to gigabit fiber circuits that still get saturated during a company-wide Teams call. This guide walks through the actual math, the difference between symmetric and asymmetric connections, and why the businesses that never have internet outages plan for redundancy — not just speed.
How Do You Calculate Bandwidth Needs Per Employee?
A practical baseline is 5–10 Mbps per employee for a modern, cloud-dependent office — but the real number depends on what each person actually does. Bandwidth planning starts by categorizing your users, not counting them.
Here’s a working framework:
- Light users (2–5 Mbps each): Email, web browsing, occasional cloud document editing. Think front-desk staff or roles that live mostly in one line-of-business application.
- Standard users (5–10 Mbps each): Microsoft 365 or Google Workspace all day, VoIP calling, regular file syncing through OneDrive, SharePoint, or Dropbox.
- Heavy users (10–25 Mbps each): Frequent HD video conferencing, large file transfers, CAD or design work, video editing, or developers pulling large repositories and containers.
The critical adjustment most businesses miss is concurrency. Not everyone hits the network at once. In a typical office, 60–80% of staff are actively using bandwidth during peak hours. So a 40-person office with mostly standard users doesn’t need 40 × 10 = 400 Mbps; it needs roughly 40 × 8 Mbps × 70% concurrency ≈ 225 Mbps, plus headroom.
That headroom matters. A good rule of thumb is to add 20–30% above your calculated peak so the connection isn’t running at capacity. Networks that regularly run above 80% utilization exhibit latency spikes and jitter long before they hit a hard ceiling — and latency is what makes VoIP calls sound robotic and video calls freeze.
How Much Bandwidth Do VoIP and Video Conferencing Use?
A single VoIP call uses surprisingly little bandwidth — roughly 100 Kbps in each direction using the common G.711 codec, or as little as 30–40 Kbps with compressed codecs like G.729 or Opus. Video conferencing is a different animal: a single HD video call consumes 1.5–3 Mbps per participant, and group HD meetings on platforms like Zoom or Microsoft Teams can pull 2.5–4 Mbps down and 1–3 Mbps up per user.
The math changes fast when you scale it:
- 10 concurrent VoIP calls: ~1 Mbps each direction. Trivial.
- 10 concurrent HD video calls: 25–40 Mbps down and 10–30 Mbps up. Suddenly significant — especially the upload.
- A company all-hands on video: Potentially your entire upload capacity, all at once.
The trap here isn’t total bandwidth — it’s upload bandwidth and quality of service (QoS). Voice and video are real-time traffic. Unlike a file download, which just takes longer on a busy network, a voice packet that arrives late is worthless. That’s why business networks running VoIP need routers or firewalls with QoS policies that prioritize voice and video traffic ahead of file syncs and backups.
We see this constantly in dealership and professional-services environments: a business buys “plenty” of bandwidth, but their nightly cloud backup runs long into the morning, saturates the upload, and the first hour of phone calls sounds terrible. The fix wasn’t more bandwidth — it was traffic prioritization and backup scheduling. This is exactly the kind of tuning a managed IT services provider handles before it becomes a recurring complaint.
What’s the Difference Between Symmetric and Asymmetric Internet?
Symmetric internet delivers equal download and upload speeds (for example, 500 Mbps down / 500 Mbps up), while asymmetric internet — typical of cable and some fixed wireless — delivers much faster download than upload (for example, 600 Mbps down / 35 Mbps up). For a modern business, that upload number is often the one that hurts.
Twenty years ago, asymmetric made sense: businesses mostly consumed content. Today, businesses produce and push data constantly:
- Cloud backups uploading changed files all day
- VoIP and video conferencing sending your audio and video out
- File sync services (OneDrive, SharePoint, Google Drive) pushing every saved document
- Security cameras streaming footage to cloud storage
- Remote workers connecting in via VPN, which is bounded by the office’s upload speed
A cable circuit advertising “600 Mbps” with 35 Mbps upload can be genuinely worse for a 30-person office than a 200 Mbps symmetric fiber circuit. If your business hosts anything on-premises — a phone system, a server that remote staff access, camera systems — symmetric bandwidth isn’t a luxury, it’s a requirement.
Fiber connections are almost always symmetric and typically come with service-level agreements (SLAs) covering uptime and repair times, which consumer-grade cable does not. Dedicated Internet Access (DIA) fiber also guarantees the bandwidth is yours alone, not shared with the neighborhood. The network solutions conversation we have with most growing businesses eventually lands on the same conclusion: fiber for the primary circuit, with something diverse as backup.
Do You Need Redundant Internet Connections?
If your business loses meaningful revenue or productivity when the internet goes down, then yes — you need a second, diverse connection. Redundancy planning is no longer an enterprise-only concern, because nearly everything a business does now stops when connectivity stops: cloud phones, payment processing, line-of-business SaaS applications, email, even door access systems in some buildings.
Real redundancy has three requirements:
- Diverse carriers. Two circuits from the same provider often share infrastructure. If a fiber cut takes out the provider’s local ring, both of your “redundant” circuits go down together.
- Diverse technology or path. Pair fiber with cable, or fiber with fixed wireless or LTE/5G. Different physical paths into the building matter — we’ve seen a single backhoe take out every wired circuit at a site because they all entered through the same conduit.
- Automatic failover. A backup circuit nobody switches to is decoration. An SD-WAN appliance or a firewall with dual-WAN failover should detect the outage and move traffic in seconds, ideally without dropping active VoIP calls.
For multi-location businesses — something we work on constantly with automotive dealership groups across the Atlanta metro and beyond — redundancy planning extends to the connections between sites. If your DMS, phone system, or file server lives at one location, every other location’s productivity depends on two circuits working: theirs and headquarters’. SD-WAN architectures address this by dynamically routing traffic across whatever connections are healthy.
A useful exercise: calculate your hourly cost of downtime (loaded payroll of affected staff plus lost revenue), then compare it to the monthly cost of a backup circuit — often $100–500 per month for cable or wireless failover. For most businesses over 10 employees, a single avoided half-day outage pays for years of redundancy.
How Do Cloud Applications Change Bandwidth Planning?
Cloud migration fundamentally shifts bandwidth from a convenience to a dependency — every application that used to run on a local server now traverses your internet connection for every click. When email lived on an in-house Exchange server, an internet outage was an inconvenience. With Microsoft 365, Salesforce, cloud accounting, and a hosted phone system, an outage is a full work stoppage.
Cloud-heavy environments change the planning math in specific ways:
- Sustained utilization rises. Instead of bursty browsing traffic, you have constant sync, API, and session traffic all day. Your baseline load is higher even when nobody is “doing anything.”
- Latency matters as much as throughput. Cloud applications are conversational — hundreds of small round-trips. A connection with low latency and low jitter feels faster than a bigger pipe with poor latency. This is another point in fiber’s favor.
- Backup windows compete with business hours. Cloud backup of servers and endpoints can move hundreds of gigabytes. Without scheduling and rate-limiting, backups will happily consume everything.
- Updates arrive in herds. Windows updates, browser updates, and application patches hitting 50 machines at once can briefly saturate any circuit. Caching and staggered deployment policies — standard practice in a managed environment — smooth this out.
A reasonable planning posture for a cloud-first business: treat your internet circuit with the same seriousness you once treated your server room. That means monitoring utilization over time (not guessing), reviewing it quarterly, and upgrading before peak usage crosses 70–80% of capacity.
What Internet Speed Should Different Business Sizes Get?
As a starting framework — before adjusting for your specific applications — these tiers fit most offices in 2026:
- 1–10 employees: 100–250 Mbps symmetric fiber, or the best available cable with an LTE/5G failover. Prioritize symmetric if you use VoIP or host anything locally.
- 10–30 employees: 250–500 Mbps symmetric fiber, dual-WAN firewall, secondary circuit from a diverse carrier.
- 30–100 employees: 500 Mbps–1 Gbps dedicated fiber with an SLA, SD-WAN or robust dual-WAN failover, QoS policies for voice/video, and utilization monitoring.
- Multi-site organizations: Dedicated fiber per site sized to that site’s headcount, SD-WAN connecting locations, and redundant paths at any site hosting shared systems.
These are starting points, not prescriptions. A 15-person engineering firm moving large drawing files can out-consume a 60-person office of email-and-browser users. The right answer comes from measuring: any competent provider should look at your actual utilization data before recommending a circuit — and after 35 years of building networks for Georgia businesses, we’d rather size a connection from evidence than from a sales target.
Frequently Asked Questions
Q: Is 100 Mbps enough for a small business? A: For most offices under 10–15 employees doing email, cloud apps, and VoIP, yes — 100 Mbps symmetric is comfortable. It becomes tight if you add heavy video conferencing, large file transfers, cloud backups during business hours, or on-site systems that remote workers access. Measure your peak utilization before assuming you need more.
Q: Why is my internet slow even though I pay for high speeds? A: The most common causes are upload saturation (backups or sync services consuming your much-smaller upload channel on asymmetric connections), Wi-Fi limitations rather than circuit limitations, aging firewalls that can’t process traffic at the speeds you’re paying for, and missing QoS so real-time traffic competes with bulk transfers. A speed test showing full speed at the router while users suffer points to an internal network issue, not a bandwidth shortage.
Q: Do I need fiber, or is cable internet fine for business? A: Cable is acceptable for small offices with light needs or as a backup circuit. Fiber is the better primary choice when you depend on VoIP, video, cloud applications, or remote access, because it’s symmetric, lower-latency, and typically backed by an SLA with defined repair times. Many businesses run fiber as primary and cable as automatic failover.
Q: How much bandwidth does a VoIP phone system need? A: Roughly 100 Kbps per concurrent call in each direction with standard codecs — so even 20 simultaneous calls need only about 2 Mbps. The challenge isn’t quantity; it’s quality. VoIP needs low latency, low jitter, and QoS prioritization so calls don’t degrade when someone starts a large upload.
Q: What is SD-WAN and do I need it? A: SD-WAN (software-defined wide area networking) is technology that intelligently routes traffic across multiple internet connections — sending real-time traffic over the best-performing path and failing over automatically when a circuit degrades or fails. It’s most valuable for businesses with multiple locations or those that can’t tolerate internet downtime. Single-site businesses can often achieve adequate redundancy with a simpler dual-WAN firewall.
COMNEXIA has designed, built, and managed business networks from our Atlanta-area headquarters since 1991. If you’d like an evidence-based assessment of your bandwidth and redundancy posture, explore our network solutions or managed IT services, or call us at (877) 600-6550.