Each draft is written in a CTO’s first-person voice and is complete as written. Under each one you will find why it works and the one thing to add to make it unmistakably yours. Where a draft states an opinion, the opinion is labeled as a position, not a fact, so you can sharpen or reverse it.
Post 1 · Architecture
The upgrade is an electronics project, not a fiber project
Nobody is digging up our fiber to get to 10G.
GPON runs at 1490 nm down and 1310 nm up. XGS-PON runs at 1577 nm down and 1270 nm up. Different wavelengths means both systems can live on the same plant. The feeder, the splitters and the drops stay where they are.
What changes is the OLT line cards, the ONTs, and the element that merges the two wavelength sets.
That is why I get nervous when a migration is described as a construction program. It is a headend, inventory and customer-premises program, which changes who needs to be in the planning meeting: central office engineering and logistics, much more than field construction.
Where does your plan spend its first 90 days?
Why it works: it corrects a common mental model with two checkable numbers, and the closing question separates people who have planned a migration from people who have read about one.
Make it yours: add how many PONs you are planning to move first, or what your own first 90 days look like.
Post 2 · Architecture opinion
Combo ports or WDM overlay: how I decide
There are two ways to put XGS-PON on a GPON network. I don’t think either is automatically right.
WDM overlay: keep the GPON cards, add XGS-PON cards, merge them with a WDM element. That element adds roughly 1 to 1.5 dB of insertion loss and usually means reorganizing the headend.
Combo ports: the WDM function is built in. No extra element, no reorganization. The catch is that you pull the GPON cards you already own.
My position: if a PON is already tight on its loss budget, that extra 1 to 1.5 dB is the number I would worry about first. If the chassis is young, I would rather not scrap cards that still have life in them.
What decides it in your network?
Why it works: it gives both options a fair hearing, uses one hard number from the Fiber Optic Association, and states an actual rule of thumb that readers can argue with.
Make it yours: name the tightest loss budget you have measured, without naming the customer or site, and say which way it pushed you.
Post 3 · Business case
The tier you cannot sell is the real reason to upgrade
Orange said it plainly: to offer more than 2 Gbps, you need to move to XGS.
GPON tops out around 2.5 Gbps down and 1.25 Gbps up, shared across the PON. That ceiling, not a spec-sheet race, is the real reason to move. Orange’s stated target was to make 100% of its French customers eligible for XGS-PON by the end of 2026, using combo ports that allow higher splitting and fewer ports overall.
So the useful question for a CTO is not “when do we upgrade?” It is “which tier can’t we sell today?”
If the answer is none, the upgrade can probably wait. If the answer is every business customer who asks for symmetric and anything above 2 Gbps, the clock is already running.
Why it works: it frames the upgrade as a revenue question, and the Orange reference shows a large operator reached the same conclusion.
Make it yours: add how many requests you have turned down in the last year for a tier GPON cannot deliver. Even a rough count is a strong line.
Post 4 · Reality check
A faster PON does not fix a take-rate problem
U.S. fiber now passes over 60% of households, and the average take rate for primary passings is 46.5%. When a second fiber provider enters a market, the combined take rate rises to about 61%. (FBA and RVA, December 2025.)
I think about those numbers every time someone proposes 10G “to win customers.”
A faster PON does not connect a home that has not been sold, scheduled or installed. XGS-PON earns its money in homes you have already passed: premium tiers, business customers and lower oversubscription.
So I test an upgrade like any growth bet. What movement in take rate or ARPU do we expect, and in which footprint?
Where is the fastest place for you to find out?
Why it works: it uses fresh, sourced industry data and then pushes against the hype, which is exactly what a skeptical peer wants from a CTO.
Make it yours: compare your own take rate with the 46.5% U.S. average, if you are comfortable sharing it, or describe the footprint you would test first.
Post 5 · Market view
XGS-PON is not obsolete because 50G exists
Every time 50G PON makes headlines, someone asks whether XGS-PON is already obsolete.
The analysts say no. Dell’Oro pushed its 50G-PON ramp from 2026 to 2028 in early 2024. Its February 2026 forecast still calls the ramp slower than expected, with China, the likely first big market, “progressing slowly.” Growth in PON equipment revenue over the next five years is tied largely to XGS-PON.
The 50G standard, ITU-T G.9804, was also designed to share the same fiber plant as GPON and XGS-PON.
So here is how I frame it. XGS-PON is the volume decision for today. 50G is a plant-compatibility question for later. Buying XGS-PON now does not wall us in.
Why it works: it answers a question every CTO gets from their board, using two dated analyst data points rather than a vendor roadmap.
Make it yours: say what would make you change your mind, for example a specific price point or ONT availability, so the post reads as a position and not a forecast.
Post 6 · Roadmap
25G is an overlay conversation, not a replacement conversation
25GS-PON is an industry agreement, not an ITU standard. It was designed to coexist with GPON and XGS-PON on the same ODN, with no extra feeder fiber or splitters.
AT&T showed 20 Gbps symmetric on a production PON in June 2022, using wavelength coexistence with XGS-PON. Dell’Oro expects many operators to keep XGS-PON for most residential subscribers and add 25G for premium tiers and enterprise.
My take: design the XGS-PON network so that a 25G port is a future line-card decision, not a plant redesign.
Which of your customers would actually pay for it?
Why it works: it separates standards from industry agreements, which many readers conflate, and it ends on a commercial question rather than a technical flex.
Make it yours: name the type of customer, such as a data center, a campus or a medical group, that you think would be first. No customer names needed.
Post 7 · Budget timing (U.S. operators)
The tax change may matter more than the datasheet
For U.S. operators, a 2026 tax change may matter more to the XGS-PON budget than any optics datasheet.
The Fiber Broadband Association expects 100% bonus depreciation to raise FTTH capex by 5 to 15%. That does not change the engineering. It changes when finance is happiest to see equipment on a purchase order.
If your migration plan has an OLT swap parked in an “if budget allows” column for next year, it may be worth a conversation with your CFO now. Not because the technology moved, but because the money did.
(I am a network person, not an accountant. Check your own tax position.)
Why it works: it speaks to the CFO as well as the CTO, which is who ultimately signs the order, and the closing disclaimer shows judgment.
Make it yours: mention whether your own plan changed, even in general terms such as pulling a phase forward. Skip this post if you operate outside the U.S.
Post 8 · Explainer
Explain “symmetric” with a Tuesday afternoon, not a line rate
GPON gives you about 2.5 Gbps down and 1.25 Gbps up. XGS-PON gives you 10 Gbps in both directions.
The upstream is the hardest part to explain and the most valuable one. Customers rarely call because a download is slow. They call because the video call stutters while the cloud backup runs, or because a large project takes an evening to upload.
Business, remote work and backup traffic all lean upstream.
A CTO who can explain “symmetric” in terms of what a customer does on a Tuesday afternoon wins more internal budget than one who quotes a line rate.
Why it works: it translates a spec into a customer moment, and it positions the CTO as someone who can brief sales and finance, not just engineering.
Make it yours: replace the Tuesday example with a real complaint category from your support data.
Post 9 · Myth buster
“XGS-PON is just GPON, four times faster”
“XGS-PON is just GPON, four times faster.” I hear it in planning meetings and it needs correcting.
A GPON ONT transmits at 1310 nm. An XGS-PON ONT transmits at 1270 nm and receives at 1577 nm. Different wavelengths, different optics, a different ONT.
That difference is what lets the two systems coexist on one fiber. It is also why your installed GPON ONTs are an inventory conversation, not a reuse conversation. Customers move ONT by ONT, as they need the tier.
Plan the logistics, meaning spares, truck stock and install procedures, for two ONT families. Probably for years.
Why it works: a myth, a correction and a practical consequence in under 120 words. This format is easy to read and easy for others to share.
Make it yours: replace “probably for years” with your own estimate of how long you expect to run both ONT families.
Post 10 · Checklist
Five questions before I approve an XGS-PON migration
Five questions I would want answered before approving an XGS-PON migration:
1. What is the loss budget on our tightest PONs, and what does a 1 to 1.5 dB WDM element do to it?
2. Combo ports or overlay, and how much life is left in the GPON cards we would be pulling?
3. Which tier can we not sell today?
4. How many ONT families will field technicians carry, and for how long?
5. Which customers do we migrate first, and how do we tell them?
None of these need a vendor demo. All of them need an honest answer before one.
Why it works: checklists are saved and reshared, and this one doubles as a summary of the whole series, so it makes a good closer for a month of posts.
Make it yours: add a sixth question that only your network would raise, and invite peers to add theirs in the comments.
