Internet Access in 2026: Fiber, Cable, 5G and Satellites

Internet Access in 2026: Fiber, Cable, 5G and Satellites

What you'll learn

• Compare the five access technologies of 2026: fiber (PON/FTTH), cable (DOCSIS 3.1/4.0), the DSL tail, 5G fixed wireless, and LEO satellite
• Explain how a passive optical network shares one fiber — and why upload speeds finally matter
• Tell the satellite reversal story: how the book's fading fallback became a mainstream contender
• Say what MPLS, carrier Ethernet and SD-WAN mean for enterprises — at awareness level, in three sentences
• Choose an access technology for four realistic scenarios, and research the real options at your own address

This lesson builds on Chapter 9 of Dr. Tahseen Al-Doori's Network Essentials — its POTS/DSL/broadband/satellite access section and its wireless-WAN pages — replacing the inventory while keeping the book's evaluation habits: who shares the medium, what limits the distance, what does the guarantee cost, and "it depends" answered by looking up your actual address, which is exactly what the lab does.

Fiber to the home: PON, the new default

Where the previous lesson's technologies rented the phone network's copper, the modern flagship lays its own glass. FTTH (fiber to the home) brings Module 3's optical medium — immune to electrical interference, nearly unlimited in capacity, attenuating so slowly that distance stops mattering at neighborhood scale — all the way to your wall. Across the OECD's fixed-broadband statistics, fiber is now the most common wired access technology — the default that everything else is measured against, a sentence nobody could have written in 2007.

The dominant architecture is the PON — passive optical network — and its cleverness is worth two minutes. From the provider's office, an OLT (optical line terminal) sends light down one fiber, which an unpowered glass splitter — a prism in a street cabinet, no electronics, nothing to fail — fans out to a few dozen homes. At your wall, an ONT (optical network terminal) converts light to Ethernet: the modern heir of the demarcation role the CSU/DSU played in the leased-line era, and of the cable modem's job on coax. Downstream, every home's ONT hears everything and keeps only frames addressed (and encrypted) to it; upstream, the OLT grants each ONT precise time slots so their light never collides — TDM's fairground wheel from the previous lesson, spinning in glass. Widely deployed GPON delivers about 2.5 Gbps down / 1.25 Gbps up per splitter group; its successor XGS-PON (current rollouts) delivers roughly 10 Gbps symmetric — and that word symmetric is fiber's quiet revolution. Symmetry itself is not the new part: the rival EPON family has been symmetric since 2004, and some GPON operators have sold symmetric gigabit plans for years. What changed is that multi-gigabit symmetry became the ordinary residential product rather than a business upsell — ordinary homes can now send as fast as they receive, which matters in the era of cloud backup, video calls and creators uploading their work. Yes, a splitter group shares capacity — but the pool is so deep that contention is rarely felt; the practical constraints are availability on your street and the plan you pay for.

Cable's second youth: DOCSIS 3.1 and 4.0

Module 3 told you coax survived the LAN's death to live on in DOCSIS — the standard that runs data over cable-TV plant. The book's era clocked cable at around 30 Mbps down; DOCSIS 3.1 (current) made gigabit-class downloads routine on the same physical cable, and DOCSIS 4.0 (rolling out) pushes toward multi-gigabit with sharply improved upload — cable's answer to fiber's symmetry argument.

Cable's character flaw is structural and the book taught it with an image worth keeping: the neighborhood shares the segment, so peak evenings can feel like a summer afternoon when every garden hose on the street is running and the water pressure sags for everyone. Two decades of node-splitting and DOCSIS generations have softened that sag enormously — but the asymmetry habit (uploads a small fraction of downloads on most plans) and the shared-medium physics remain cable's identity. Where fiber hasn't arrived, cable is usually the strongest incumbent; where both exist, the comparison typically turns on upload and consistency, not the headline download.

The DSL tail

DSL — broadband over the telephone pair — was the book's great democratizer, and its physics lesson still holds letter-perfect: bandwidth falls with distance from the exchange, so the farmhouse far from town got little or nothing. The technology's late refinements, VDSL2 (tens to ~100–200 Mbps from a curbside cabinet) and G.fast (fiber to the building's basement, hundreds of megabits over the last meters of copper), squeezed the pair harder than 2007 thought possible — but the story's direction is unmistakable: DSL is the tail (legacy — no new investment beyond keeping it alive). Its subscriber share shrinks every year as fiber overbuilds it, and the PSTN switch-offs from the previous lesson pull the copper out from under it on a schedule. Where DSL remains the only wired option, it works; nobody chooses it where anything else exists. Its honorable epitaph: DSL taught the world always-on broadband, then handed the pair's customers to glass.

5G to the rooftop: fixed wireless access

Here the book's closing question — "wireless: the new replacement for the PSTN?" — gets its answer. FWA (fixed wireless access) uses the cellular network — today meaning 5G, with 4G in supporting roles — to serve a fixed location: a router on your windowsill or an antenna on the eave, aimed at a tower, serving your whole LAN. No trench, no street cabinet: where a tower has capacity, the "installation" is shipping a box, which is why FWA became the fastest-growing access category of the mid-2020s — on the order of 185 million FWA subscriptions worldwide at the end of 2025, per the Ericsson Mobility Report's FWA outlook, with strong continued growth projected. Typical service lands in the tens-to-hundreds of megabits (gigabit-class where mid-band spectrum is generous), latency sits comfortably below satellite, and performance honesty requires the previous module's vocabulary: you share a cell's radio capacity the way cable shares a segment — evenings sag where towers are oversubscribed, and radio physics (distance, walls, weather at the margins) applies. FWA's sweet spots: places wires reach poorly, renters who can't order construction, competitive pressure where the wired incumbent grew complacent, and instant service for pop-ups and temporary sites.

The satellite reversal: the century's best comeback

Now for the reversal this course has been promising since its first module — the clearest then-versus-now story in the book, and arguably in all of networking.

The book's verdict on satellite Internet was written in good faith and was correct as written: a rural fallback — dish for downloads, often a phone line crawling upstream — needed less and less as DSL spread; increasingly rare was the honest 2007 forecast. The physics behind that verdict: traditional satellites sit in geostationary orbit, ~36,000 km up, and radio at light speed needs half a second and more for the round trip — fine for a broadcast, miserable for a conversation, disqualifying for a video call.

The reversal came from changing the orbit, not the radio. LEO — low Earth orbit — constellations fly at a few hundred kilometers instead: latency collapses from geostationary's half-second to a few tens of milliseconds — conversation-grade, videocall-grade, ordinary-Internet-grade. The cost of the low orbit is coverage: each satellite sees only a small, fast-moving patch of Earth, so you need thousands, cheap to build and launch, handing your session from one to the next like an ESS in the sky (lesson 1's roaming, at orbital velocity), with a self-aiming phased-array dish on your lawn doing the tracking (lesson 1's beamforming, pointed at space). Reusable rockets made those thousands affordable — and by 2026 the leading constellation, Starlink, had passed ten million subscribers on a fleet of more than ten thousand satellites in orbit — both figures climbing fast enough that any precise number ages within months, so verify them before you quote them — with competing constellations (Kuiper, OneWeb and others) building behind it. Real-world service lands in the tens-to-hundreds of megabits with latency in the tens of milliseconds.

Tell it as the resurrection it is: a technology correctly written off — the book's reasoning was sound for the machine it described — reborn as a different machine wearing the same name. The dish on the farmhouse in 2007 marked the home that broadband forgot; the dish on the farmhouse in 2026 may enjoy a better connection than the town. For remote regions, ships, aircraft, disaster zones and every address whose economics no trench will ever justify, LEO service turned "increasingly rare" into a mainstream fourth option — and its lesson for your professional judgment is permanent: verdicts on technologies are verdicts on their assumptions; change an assumption (here, the orbit) and the verdict flips.

The enterprise aisle, briefly

Businesses buy from a parallel catalog, and awareness level is all this course needs — three sentences, as promised. MPLS sells the previous lesson's virtual-circuit idea as label-switched paths with performance guarantees between company sites. Carrier Ethernet is the leased line reborn: dedicated point-to-point fiber capacity, sold with familiar Ethernet interfaces and a service-level agreement. SD-WAN is software that steers each application's traffic intelligently across whatever mix of links a site has — MPLS, broadband, FWA — getting circuit-like discipline out of ordinary access, which is why it is often the modern answer where a Frame-Relay-then-MPLS contract used to be.

Choosing access: four scenarios

The book's evaluation habits, applied to the 2026 shelf. For each scenario: the constraint that decides, then the choice.

1 — The apartment streamer-gamer. City apartment; heavy streaming, competitive gaming (latency-sensitive), long work video calls most days. Deciding constraints: latency and evening consistency. Choice: fiber if the building has it — symmetric, uncontended-feeling, lowest stable latency; DOCSIS 3.1 cable as the strong second, accepting the upload asymmetry; FWA only if both wires disappoint, mindful of evening cell load. LEO adds nothing indoors downtown.

2 — The remote farmhouse. Twenty kilometers from town — the book's own far-from-the-exchange farm, updated. No fiber build in any plan; DSL long dead or unusable at that distance. Deciding constraint: what reaches at all. Choice: 5G FWA if a tower serves the area (check coverage honestly — rural cells can be sparse); otherwise LEO satellite, gladly — this address is precisely who the reversal happened for. The 2007 answer (nothing, or a crawling dish) and the 2026 answer (tens to hundreds of megabits either way) measure the era's progress in one household.

3 — The uploading small business. A 12-person design studio: multi-gigabyte client deliveries daily, cloud backup, VoIP phones (Module 8 will teach the QoS side). Deciding constraint: upload and reliability, with a business's tolerance for downtime near zero. Choice: symmetric fiber — a business plan on XGS-PON, or carrier Ethernet if the studio wants a hard SLA; and because "near zero," a cheap diverse backup link — FWA is ideal, since it shares no trench, pole or duct with the fiber. One lesson-four idea resurfaces: paying for guarantees versus paying for capacity are different purchases.

4 — The pop-up and the field site. A three-month retail pop-up, then a construction office in a greenfield lot. Deciding constraint: time-to-service and contract flexibility — no trenching, no twelve-month waits. Choice: FWA first (ship a router, service today); LEO where cellular is thin or the site keeps moving. Wired access is simply the wrong shape for temporary premises — a sentence that would have read as defeat in 2007 and reads as routine now.

From the textbook to 2026

The chapter's access section, run through the pattern one last time. Original: POTS dial-up, DSL, ~30 Mbps shared cable, satellite-as-fading-fallback. What remains true: every evaluation habit — who shares the medium (cable's summer-hose evenings, now FWA's busy cells, now PON's deep-pooled splitter), what limits distance (DSL's exchange, now the cell edge), what a guarantee costs (leased line then, carrier Ethernet and SLAs now). What changed: dial-up is gone (its last major US service closed in 2025); DSL is a managed tail under PSTN switch-offs through ~2027; cable leapt two orders of magnitude via DOCSIS 3.1/4.0; fiber PON became the OECD's dominant wired technology; the cellular network became a fixed-access product with ~185 million subscriptions (end-2025, Ericsson); and satellite executed networking's great reversal via LEO. 2026 practice: the four-scenario judgment above — and the lab below, which replaces the book's "price a T-1 in your area" exercise with its exact modern descendant: find out what your address can actually get, and check the claims against measurement.

Lab: The access audit at your own address

The book sent students to price a T-1 locally — research against the real market. Same spirit, 2026 catalog: you will inventory the access technologies genuinely available where you live, then test the one you have against its advertised number.

Objective. Produce a one-page access audit for your address: every available technology, its advertised speeds, your current service's measured performance, and a verdict on the gap.

Setup.

• A browser, plus the connection you already have.
• A privacy note before you type anything: availability checkers require your address, and speed-test sites observe your IP address and provider — normal, but data you're handing over. Prefer official regulator/ISP checkers over ad-heavy "comparison" portals, skip any site demanding an account or phone number for basic availability, and when you keep notes for this lab, record street-level results without writing your full address into files you might later share (your module scorecards travel with you into the capstone).

Steps.

1. Inventory the wires. Check your national regulator's broadband map if one exists (many countries publish official coverage maps), then the availability pages of your likely fiber and cable providers. Record for each: technology (from this lesson's catalog — the marketing name and your identification: "FTTH — XGS-PON", "cable — DOCSIS 3.1", "VDSL2"), advertised down/up, and advertised price tier if shown.
2. Inventory the air. Check the FWA offering of your mobile carriers (coverage checkers again — address-level), and the LEO provider's availability for your area. Record the same fields; note where the FWA page hedges with "up to" and "typical" figures — that distinction is this course's bandwidth-versus-throughput lesson printed as marketing.
3. Identify what you have. From your router's WAN status page, your bill, or the box on your wall (an ONT means fiber; a coax modem means DOCSIS; a phone-jack modem means DSL; a SIM slot or window-mounted unit means FWA): name your current technology and its advertised plan speed.
4. Measure it. Run a reputable browser speed test three times: once mid-morning, once mid-evening (the shared-medium hour), once late night — from a device wired to the router if at all possible, so you measure the access link rather than your Wi-Fi (this module's first two lessons explain exactly why the radio hop would contaminate the experiment; if wireless is unavoidable, sit next to the router on 5 or 6 GHz). Record down, up, and latency each time.
5. Analyze. Compute measured-versus-advertised as a percentage for your best and worst runs. Note the evening dip, if any — and name which section of this lesson predicted it for your technology.
6. Write the audit. One page: available technologies ranked, your current service, the measurement table, and a three-sentence verdict — is your household on the right technology of those available, and what would you switch to if one thing changed (price, a fiber build reaching your street, a new tower)?

Expected result. Most addresses in 2026 list two to four real options (a wire or two, FWA, LEO); measurements typically land between 60% and 100% of advertised on a healthy wired service, with evening dips most visible on cable and FWA; latency separates the catalog sharply (fiber lowest, LEO respectable-but-distinct, any DSL tail showing its age).

Verify. Your audit page answers, with your own data: which catalog technologies serve your address; what you actually measured against what is advertised; and one sentence of evidence that either confirms a shared-medium evening sag or honestly reports its absence.

Questions.

1. Your evening download dipped 30% but latency barely moved. Which technologies make that signature likely, and why?
2. The FWA checker says "available" but quotes only "typical 50–180 Mbps" while your cable ISP advertises "up to 1 Gbps." Frame the honest comparison a professional would make before choosing.
3. Why did this lab insist on a wired device for the speed test — state the exact lesson-1/lesson-2 machinery that would otherwise contaminate the numbers.
4. Your measured upload is 5% of your download. Which technology are you almost certainly on, and which two catalog entries would fix that ratio?

(Answers: 1 — shared-segment technologies under evening load: cable/DOCSIS or FWA; capacity contention throttles throughput while the path—and hence base latency—stays the same. 2 — compare typical to typical: run-or-request real measurements, weigh evening figures, upload, and latency against price; "up to" is a ceiling, not a forecast — bandwidth versus throughput, in marketing form. 3 — the Wi-Fi hop is a shared half-duplex medium with CSMA/CA overhead, its link rate varies with distance/walls/interference, and neighbors contend for the channel — any of which can cap measurements below the access link's real capacity. 4 — an asymmetric plan, most likely legacy cable or ADSL/VDSL; fiber PON (symmetric by design, especially XGS-PON) fixes it outright, and DOCSIS 4.0's improved upstream is the cable path to respectability.)

If it goes wrong.

• Availability sites disagree: trust the provider's own checker over aggregators, and the regulator's map over both for "does the infrastructure exist"; record disagreements as findings — messy data is real data.
• Speed test wildly below plan on every run: re-test wired if you weren't; reboot the modem/ONT-side router once; test at a different hour. If it persists wired at night, your finding is genuine — keep it for Module 8, where the troubleshooting method turns findings like this into diagnoses.
• No wired device available: note "measured over Wi-Fi" on the audit and treat your numbers as lower bounds — honest labeling beats false precision.
• You can't determine your own technology: photograph the box the provider installed and match it against step 3's field guide; the bill's plan name usually settles ties.

Reset/cleanup. Nothing was changed. Keep the audit page: the capstone's requirements-and-design lesson asks you to choose WAN access for a fictional firm, and you will do it faster with a real audit under your belt. Delete any notes containing your address or IP details from shared folders — the privacy habit is part of the lab.

Check yourself

1. A splitter group's fiber is shared by 32 homes, yet PON users rarely complain about evening sag while cable users historically did. Reconcile, using the capacity arithmetic implied by this lesson.
2. Map each to its catalog entry: (a) an unpowered glass prism in a street cabinet; (b) a self-aiming flat antenna tracking objects crossing the sky; (c) a windowsill router with a SIM; (d) hundreds of megabits over the last thirty meters of telephone copper from the basement.
3. The book called satellite "increasingly rare" and this course calls that verdict correct as written. Defend both statements in the same paragraph — what assumption changed, and what is the transferable professional lesson?
4. A rural clinic must choose between FWA (tower 8 km away, "typical 40–150 Mbps") and LEO (flat-terrain sky view). Its workload: telemedicine video and cloud health records. List the three questions you would answer before deciding, tied to this lesson's constraints.
5. An enterprise migrates its last Frame Relay contract. In one sentence each: what MPLS offers it, what carrier Ethernet offers it, and what SD-WAN would add on top.
6. Your audit shows fiber advertised at your address "from the pole," cable active in the building, FWA strong, LEO available. Rank the four for a household of remote workers with heavy video calls, and defend the top pick in two sentences.

Answers

1. Both share, but the pools differ by orders of magnitude: a PON group shares multi-gigabit (GPON ~2.5, XGS-PON ~10 Gbps) among a few dozen homes, while legacy cable segments shared far less among far more — so contention exists on both but is rarely felt on PON. The lesson: "shared" is not a verdict; shared-capacity-per-subscriber is.
2. (a) The PON splitter. (b) A LEO satellite terminal's phased-array dish. (c) 5G FWA equipment. (d) G.fast — the DSL tail's last clever squeeze.
3. The 2007 verdict correctly described geostationary satellite: half-second physics, crawling uplinks, a fallback shrinking as DSL spread — sound reasoning about the machine that existed. LEO changed the assumption the verdict rested on (the orbit, hence the latency, hence the economics), and thousands of cheap satellites plus reusable launch built a different machine with the same name — over ten million subscribers on the leading constellation. Transferable lesson: technology verdicts are verdicts on assumptions; audit the assumption before repeating the verdict.
4. First, real coverage and load at the clinic's exact location (a coverage map's "available" versus measured evening throughput at 8 km from the tower). Second, latency and jitter under load for telemedicine video (both candidates can serve it; measure, don't assume). Third, reliability and failure modes (tower congestion and radio path versus sky obstruction and weather margins) — ideally concluding with the business answer: primary plus the other as diverse backup, the scenario-3 pattern.
5. MPLS: label-switched paths with performance guarantees between sites — the virtual-circuit promise its Frame Relay contract made, kept with modern machinery. Carrier Ethernet: dedicated fiber capacity with familiar interfaces and an SLA — the leased line reborn. SD-WAN: software steering each application across MPLS-plus-broadband-plus-FWA, buying circuit-like discipline from ordinary links — often shrinking how much MPLS it must keep buying.
6. Fiber first — symmetric upload and the lowest, steadiest latency are precisely a video-call household's constraints, and "from the pole" means a routine install. Cable second (strong download, weaker upload), FWA third (fine service, evening-cell variance), LEO last here — not because it is weak, but because at a fiber-served address its advantages are all solving problems this household doesn't have.

Key terms

• FTTH / PON — fiber to the home over a passive optical network: OLT at the provider, unpowered splitter in the street, ONT on your wall.
• OLT / splitter / ONT — the PON cast; the ONT is the fiber era's demarcation box (the CSU/DSU's modern heir).
• GPON / XGS-PON — the deployed PON generations: ~2.5/1.25 Gbps shared vs. ~10 Gbps symmetric.
• Symmetric service — equal upload and download; fiber's quiet revolution for the cloud-backup, video-call era.
• DOCSIS 3.1 / 4.0 — cable broadband's current and emerging generations: gigabit-class down, then multi-gig with serious upstream.
• VDSL2 / G.fast — the DSL tail's refinements: cabinet-fed tens-to-hundreds of megabits; basement-fed hundreds over the last meters.
• FWA (fixed wireless access) — cellular (5G-era) service to a fixed location; ~185 million subscriptions at end-2025 (Ericsson Mobility Report).
• GEO vs. LEO — geostationary's ~36,000 km half-second physics vs. low-orbit constellations' tens of milliseconds and thousands of satellites.
• LEO constellation — the satellite reversal: 10,000+ satellites, 10M+ subscribers on the leading system; the mainstream fourth access option.
• MPLS / carrier Ethernet / SD-WAN — the enterprise aisle: guaranteed label-switched paths; the leased line reborn on fiber; software steering across any mix of links.
• Access audit — the lab's deliverable: available technologies, advertised claims, measured truth — the book's price-a-trunk-line exercise, modernized.

Summary

• The 2026 access catalog: fiber PON (the OECD-dominant wired default, symmetric on XGS-PON), DOCSIS 3.1/4.0 cable (multi-gig, shared-segment character), a shrinking DSL tail (VDSL2/G.fast under PSTN sunset), 5G FWA (~185 million subscriptions end-2025, per Ericsson), and LEO satellite (the reversal: 10M+ subscribers where 2007 saw a fading fallback).
• The book's evaluation habits transfer intact: who shares the medium, what limits distance, what guarantees cost — only the inventory changed.
• PON's machinery is the old ideas in glass: TDM time slots upstream, a demarcation box (ONT) on the wall, shared capacity with a pool deep enough to feel private.
• The satellite reversal is the course's cleanest lesson in technology judgment: sound verdicts die when their assumptions (here, the orbit) change.
• Enterprises shop a parallel aisle — MPLS guarantees, carrier-Ethernet dedicated capacity, SD-WAN software discipline — three sentences of awareness that Module 9's capstone will spend.
• Choosing access is constraint-matching, not spec-chasing: latency and symmetry for some, reach and deployment speed for others, diverse backup for businesses.
• You audited your own address and tested claims against measurement — the module ends the way this course always ends a topic: with evidence.

Next lesson

Module 6 is complete — radio, generations, security, history and the modern edge. Module 7 turns to the discipline that ran beneath half of this module's stories: security proper — threats, defense in depth, firewalls, encryption and authentication, from the CIA triad to passkeys.

Sources and further study

• Al-Doori, T., Network Essentials, Chapter 9 — the access-technology section and wireless-WAN pages this lesson modernizes, and the evaluation habits it keeps.
• Ericsson Mobility Report, "Fixed Wireless Access outlook" — https://www.ericsson.com/en/reports-and-papers/mobility-report/dataforecasts/fwa-outlook — the FWA subscription figures cited in this lesson (end-2025 data).
• Cloudflare, "The 2025 Cloudflare Radar Year in Review" — https://blog.cloudflare.com/radar-2025-year-in-review/ — measured Internet traffic and connectivity trends from a global vantage point.
• Cloudflare Learning Center, "How does the Internet work?" — https://www.cloudflare.com/learning/network-layer/how-does-the-internet-work/ — how access networks join the routed core you studied in Module 5.