IBC 2026: AI gets practical, control matters, and the gaps are becoming the problem

IBC 2026 may have seen a 6% drop in footfall overall, but for Hypersolid it was our busiest show in years. Back-to-back meetings made it hard to get away from our stand and get a sense of what's moving in the industry. A nice problem to have, obviously!

Thankfully, I did manage to escape on the final day and I've pulled together this summary of the tech trends I spotted across the show floor.

Monetizing an archive is harder than it looks

I’ll start with what I heard on our own stand. Across broadcasters, media companies and sports organizations, the same questions kept coming up: where can AI genuinely help, how do you innovate without rebuilding everything, and how do you bring in outside expertise without giving up control?

One thing that generated a lot of discussion was something new from our Hyperlab innovation team: the Stream Companion Analyzer.

Several broadcasters told us they know they're sitting on a lot of untapped value in their archives, but unlocking it is difficult. We showed them how the AI analysis can help them do that across a range of different content types, including football matches, reality shows, and even weather forecasts.

Here's what struck me elsewhere at the show:

AI has stopped being the pitch

For the last couple of years, it has sometimes felt as though adding “AI” to a product description was enough to create an IBC proposition. This year, there were far more examples of AI doing useful jobs inside real workflows: localization, metadata, search, QC, clipping and subtitling. That is progress. But many still solve only one part of the process. You can optimize ten individual steps and still have an inefficient workflow if the systems around them don’t join up.

One project already looking beyond those individual steps was IBC’s Smart Stories Accelerator, involving organizations including the BBC, Channel 4, ITV, Sky, Reuters, AP and ITN. It looks across the content chain from news gathering to transmission.

Increasingly, I think the cost will sit in the gaps between systems. By next year, I expect much more focus on the complete workflow rather than individual AI features.

Human-in-the-loop is still the preferred approach

The discussion around AI also felt more mature. The strongest examples were not about removing humans altogether. One buyer I spoke to had spent a full day looking at dubbing solutions and told me synthetic voices still aren't quite there for certain languages – including Norwegian and Portuguese. The issue wasn't whether the words were understandable. It was whether the voice carried the right emotion. He could also see a clear difference between workflows that retained human review and those trying to automate everything.

Open standards are becoming real infrastructure

TAMS is another good example of something that felt noticeably more mature this year. What started as a BBC project around storing and addressing media by timeline, Time Addressable Media Stores (TAMS) is developing into a much broader open ecosystem. There is now an open specification, reference implementation, tooling, and a growing set of integrations around it.

The underlying principle is powerful: one stored version of media that multiple downstream tools can work with, rather than repeatedly creating, moving and managing copies through separate silos. 

Hypersolid has championed the "one copy" principle for some time. But as that becomes an open standard, it stops being a differentiator and starts becoming something buyers expect. The engineering challenge moves upward: how intelligently can we search, enrich, and build workflows around that common timeline?

Buyers are setting the innovation agenda

It was also interesting to note who was being recognized for innovation this year. Major IBC Innovation Award winners included projects led by the EBU/NABA, Reuters and Globo, with technology companies appearing as partners in solving real operational problems.

I think that distinction matters. Increasingly, the organizations running the workflows are defining what they need and bringing in specialist partners to help build it.

That mirrored many conversations on our stand. Large broadcasters and streamers often don’t want to outsource their whole platform, but they don’t want to build absolutely everything themselves either. They want outside experts who can bring ideas, technology, and engineering capacity while working alongside their own teams.

That middle ground between a rigid off-the-shelf product and a complete bespoke build feels increasingly important and is a close match to our approach at Hypersolid.

Efficiency is becoming an architecture question again

Last, but not least, infrastructure efficiency was also a recurring theme. Our partner Scalstrm, for example, was demonstrating 110 HD channels in a single 1U server using NETINT VPUs, with significant reductions in TCO, power use and rack space.

The exact benchmark matters less to me than the principle. For a while, heavier workloads seemed automatically to mean more generic compute, more cloud resources and more GPUs. Video is a specialized workload, and purpose-built processors deserve another look. At scale, efficiency is not just a sustainability issue – although that's a big part of it. Efficiency also affects costs, capacity, and architecture, and we heard a lot about how important those are to our customers over the four days of the show!

The gaps are becoming the interesting bit

So, IBC is over for another year. If I had to reduce IBC2026 to one takeaway, it would be this: the individual technologies are getting very good. The harder problem is making them work together.

That is where a lot of the hidden cost still sits, and where I think some of the most interesting engineering work is going to happen next.

Which is good news for people like us. There is still plenty left to build.