What I'm Watching in H2 2026

Coming off the H1 retrospective, here are six things I'll be tracking through the second half of 2026. I've split them by confidence — three calls I'm fairly sure about, three I expect to be at least partly wrong about by Q4.

Calling them now so the December grading post writes itself.

The three I'm fairly sure about

One — 800ZR+ multi-vendor interop converges by Q4

The OIF profile work for 800ZR+ has been the bottleneck on multi-vendor interop. The latest contributions show the major remaining disagreements narrowing. By Q4 I expect either the profile to be finalised or the major vendors to publish a de-facto interop matrix that lets buyers procure across vendors confidently.

The forecast: by end of Q4, you can procure 800ZR+ from two vendors and put them on the same link without doing your own qualification testing.

Why I'm confident: the technical convergence is visible in the public OIF meeting outputs. The political layer (which vendor lets the others catch up) is the variable, and the market pressure for interop is high enough that the political layer resolves.

What would change my mind: if any major DSP vendor announces a single-source 800ZR+ commercial program with hyperscaler customers that doesn't include cross-vendor interop. That would signal that the leader is willing to extend the lock-in window rather than support interop.

Two — RPKI coverage crosses 55% by end of year

Currently 47%. Trajectory continues but at a slower marginal rate. The remaining adoption requires more operational effort per percentage point because the easy adopters have moved.

The forecast: 53-58% by end of year. Most likely landing zone is 55%.

Why I'm confident: the cyber-insurance pressure and the hyperscaler peering policy are still pushing. The pipeline of pending deployments is observable through RIR submissions.

What would change my mind: if a major Tier-1 transit announces it won't deploy ROV at all, signalling a structural rejection of the trajectory. Plausible but unlikely.

Three — ASPA enforcement reaches one major Tier-1 transit by Q4

Currently in pilot at three transits, no production enforcement. By Q4 I expect at least one to flip to production enforcement, which becomes the existence proof other transits cite.

The forecast: at least one Tier-1 announces ASPA enforcement in production by Q4.

Why I'm confident: the technical readiness is there, the customer demand is there, the political will at two of the three pilots is increasing. One of them flips.

What would change my mind: if the first transit's production enforcement causes a publicly visible route-leak miss-detection or false-positive event, the others delay another year.

The three I expect to be wrong about

Four — A second hyperscaler announces CPO as general data center default

Today CPO is hyperscaler AI fabric specific. The transition to "general DC compute" is the broadening signal. I'm betting one hyperscaler announces it by Q4.

The forecast: at least one major hyperscaler publicly commits to CPO outside of AI fabric specific deployments by Q4.

Why I expect to be wrong: my January track record suggests I'm too aggressive on hyperscaler timelines. The cautious deployment pattern observed in H1 means the announcement probably waits until H1 2027.

What I'd need to see to be right: a public announcement that names CPO in a general-DC switching context, with volume figures or deployment scope that's clearly outside the AI fabric.

Five — IEEE 1.6T ratifies cleanly by Q4

I've called this slip already. I expect to be wrong about it slipping further too, which is the meta-prediction.

The forecast: 1.6T primary ratification lands by end of Q4 2026.

Why I expect to be wrong: the slip pattern is consistent. The remaining unresolved issues at the last plenary haven't reduced as fast as a Q4 ratification would require. Most likely actual outcome: ratification slips to Q1 2027.

What I'd need to see to be right: at the next IEEE plenary, the unresolved issue count drops sharply and the working group chair signals confidence in a Q4 close.

Six — A US-headquartered optical transceiver vendor reaches meaningful volume

This is the prediction-11 retread from H1. I keep expecting a new entrant. The market keeps not producing one.

The forecast: by end of 2026, a US-based new entrant ships 5-10K units per quarter at production volume.

Why I expect to be wrong: the capital and qualification timelines work against new entrants. The most likely path is acquisition-of-existing rather than greenfield, and that doesn't fit the "new entrant" framing.

What I'd need to see to be right: any of the half-dozen startup ventures I'm tracking actually ships product at scale, in a quarter that's reported externally.

Three meta-trends I'm watching that aren't predictions

These I'm tracking but won't make specific calls on.

The AI fabric × OLS collision starts producing artefacts. Per the recent post, the political collision happens in 2027. The early signals — carrier announcements, hyperscaler procurement decisions, standards-body documents — should start showing in H2 2026. I'll be reading these as input to a 2027 prediction set.

Cyber-insurance language continues to evolve. Each year the insurance industry writes more specific clauses about routing security, optical infrastructure, and AS-level practices. The next round of clauses lands in late 2026. Whatever's in those clauses signals which operational practices the next year of compliance work will focus on.

Hyperscaler in-house transceiver programs mature or get abandoned. Multiple hyperscalers have explored building their own optical modules. The decisions to either commit or back away land somewhere in 2026-2027. The signal is in capex disclosures and vendor commentary.

What I'd want feedback on

If you're a network operator, optics buyer, or industry observer reading this, the predictions I'd most welcome challenge on are:

• The 800ZR+ multi-vendor interop call. Am I being optimistic about Q4? The political layer might be slower than I assume.
• The CPO general-DC adoption timing. Am I being too aggressive, or too conservative? Both possibilities are real.
• The IEEE 1.6T schedule. I called slip. Is even Q1 2027 too optimistic?

The forecasting discipline only improves with the corrections. The H1 retrospective showed me I'm structurally too optimistic about timing. The H2 calls try to correct for that, which probably means I'm now structurally too pessimistic. The grade comes in December.

See you in six months. Same exercise. Same honesty about what landed and what didn't.

The H2 calls are placed. Time to watch them play out.