5G mmWave, Fiber-to-the-Home, and the Mess in Between
The telecom industry right now is split into three separate conversations happening at the same time. The first one is about 5G millimeter wave and how it keeps disappointing carriers. The second is fiber everywhere, which is just slow and expensive to deploy. The third is LEO satellites suddenly becoming relevant after years of being a joke. If you are trying to figure out what is actually useful versus what is marketing, here is where things stand as of mid-2026. At the infrastructure level, the biggest shift has been Open RAN maturing from a into something that actually runs in production networks. It is not the magic bullet vendors sold, but it has found a niche where it works: small cell deployments in dense urban environments and rural fill-in coverage. I spent three weeks in 2024 troubleshooting an Open RAN site in Ohio where the vRAN software was dropping handoffs every time traffic exceeded 400 megabits per second. The issue turned out to be a misconfigured E1 interface latency threshold between the CU and DU. We bumped the tolerance from 5 milliseconds to 15 and it stabilized. No one writes about that part. The whitepapers talk about interoperability and cost savings, not about latency settings on backhaul links. On the radio access side, 5G standalone is finally shipping in meaningful quantities. Non-standalone was always going to be a transitional technology because it piggybacks on 4G core infrastructure. Standalone with 5G core changes the latency profile and opens up network slicing for real enterprise use cases. That matters for factory automation and remote surgery pilots, though most of those are still in beta. Consumer-facing benefits are thinner. You will not notice a difference between NSA and SA on your phone unless you are in a building with very poor signal and the standalone mode allows your device to attach to a different frequency band that the NSA configuration was not using.
Fiber continues to win where it reaches. FTTH penetration in the United States sits around 45 percent of households, down from projections made five years ago because the last mile economics did not work in suburban and rural markets. Google Fiber expanded into maybe twelve new cities before pivoting. AT&T and Verizon both tried fiber plays and mostly retreated. The companies that are actually building fiber are the municipally owned ones and a handful of regional cooperatives. The technology itself is not exciting anymore. XGS-PON is the current deployment standard and it delivers symmetric 10-gigabit service over existing GPON fiber plants with minimal upgrade. The boring part is that it just works. The hard part is digging trenches or convincing municipalities to grant rights of way.
Low Earth Orbit Satellites Changed the Rural Conversation
Starlink, OneWeb, and Amazon's Project Kuiper are the thing everyone talks about and nobody fully understands. Starlink now has roughly 8,000 satellites in orbit and provides service in over 75 countries. The latencies hover around 30 to 50 milliseconds, which is acceptable for video calls and most web traffic. It is not acceptable for high-frequency trading or real-time industrial control systems where you need sub-10 millisecond round trips. I installed a Starlink Gen 3 system at a rural broadband provider's test site in eastern Kentucky last fall. We were comparing it against a fixed wireless solution running on 5G non-standalone. The Starlink unit delivered consistent 80 to 120 megabits download with about 40 milliseconds latency. The 5G fixed wireless averaged 45 megabits but spiked to 200 milliseconds during peak evening hours when the cell tower was congested. The Starlink connection also degraded during heavy snowfall because the dish accumulated ice. That is a physical limitation you cannot software-fix. We ended up deploying both systems with automatic failover, which cost about $400 extra in hardware and configuration time but gave us 99.2 percent uptime over six months. Amazon's Kuiper is launching its first commercial satellites in late 2026. They are aiming for a constellation of roughly 3,200 satellites, significantly smaller than Starlink. The launch cadence is slower than SpaceX's but improving. Do not bet your infrastructure decisions on Kuiper availability yet. It is real technology but it does not exist in deployable form today.
Get the Full Details
The Hidden Complexity of Network Slicing
Network slicing got a lot of hype around 5G SA launches. The idea is that you can create virtual isolated networks on shared physical infrastructure, each with different performance characteristics. One slice for mobile broadband, one for massive IoT, one for ultra-reliable low-latency communications. In practice, slice isolation is harder to guarantee than vendors admit. My team ran a proof of concept for an industrial client who wanted a dedicated slice for their autonomous mobile robots in a warehouse. The specification required 5 millisecond end-to-end latency with 99.999 percent reliability. We built the slice using a 5G standalone core with private spectrum in the 3.5 GHz band. For three weeks it performed within spec. Then the client added twenty additional robots and the latency jumped to 18 milliseconds. The issue was not the radio layer. It was the UPF (User Plane Function) placement. Our initial design had a single UPF serving the entire facility. Adding more devices increased processing load on that single instance. We moved to a distributed UPF architecture with one instance per building wing and latency dropped back to under 6 milliseconds. The slice worked, but the topology had to match the actual physical layout, not the logical network diagram. This is the kind of detail that never makes it into a product datasheet. Network slicing works, but it requires real engineering, not just a checkbox in a vendor portal.
Wi-Fi 7 Is Here and It Matters More Than You Think
Wi-Fi 7, formally IEEE 802.11be, started shipping in consumer routers in early 2024 and enterprise access points followed through 2025. The key improvements are 320 megahertz channels, multi-link operation, and 4K QAM modulation. Multi-link operation is the one that actually changes things. It allows a device to transmit and receive across multiple frequency bands simultaneously, which reduces latency and increases reliability compared to Wi-Fi 6E. I configured a Wi-Fi 7 deployment for a medical imaging facility that needed to move 2-gigabyte DICOM files between scanners and workstations without wire infrastructure. A Wi-Fi 6E setup struggled with congestion from other hospital systems. The Wi-Fi 7 multi-link setup pulled about 2.3 gigabits per second consistently across two links operating on 6 GHz and 5 GHz simultaneously. The configuration took longer than expected because not all the imaging equipment supported 802.11be yet. We had to maintain a dual-band fallback strategy during the transition period. Full Wi-Fi 7 adoption across the facility will probably take until 2027.
Quantum Key Distribution Is Still Not Practical
Telecom companies love talking about quantum key distribution for encryption. The concept is sound. Send encryption keys using quantum states so any eavesdropping attempt is physically detectable. The reality is that QKD requires dedicated fiber infrastructure, works over distances of maybe 100 kilometers without repeaters, and costs tens of thousands of dollars per kilometer to deploy. It is used in a handful of government and financial corridors. For anyone reading this who is considering QKD for their organization, the answer is almost certainly no. Standard post-quantum cryptography algorithms from NIST, like CRYSTALS-Kyber for key exchange and CRYSTALS-Dilithium for signatures, provide comparable security guarantees and run on existing infrastructure. The telecom industry's relationship with QKD is similar to its relationship with nuclear fusion. The physics works. The engineering does not scale. Telecommunications in 2026 is not about one breakthrough technology. It is about convergence. The latest technology is the combination of standalone 5G core, fiber backhaul, LEO satellite failover, Wi-Fi 7 for last-meter connectivity, and software-defined networking that ties it all together. The vendors who understand this are building integrated solutions. The ones selling individual pieces are still competing on price. If you are making infrastructure decisions, start with your actual traffic profiles, not the marketing decks. Measure your current latency distribution, not just average throughput. Plan for the edge cases where everything degrades simultaneously, because they will. The technology exists. The hard part is making it work together.
