IBC 2026: AI May Have Won the Bullshit Bingo, Economics Will Decide the Real Winners

IBC published its final numbers for 2026: 41,225 attendees from 170 countries and more than 1,300 exhibitors.

That is 2,633 fewer visitors than last year, a decline of roughly 6% and probably a relative growth when compared to the layoffs throughout the industry The exhibitor count remained broadly stable. But more strikingly, 256 companies exhibited for the first time, 36% more newcomers than in 2025. A staggering rate of renewal.

You can interpret that as evidence of a vibrant market. You can equally interpret it as evidence of enormous fallout. Hundreds of new companies are arriving because hundreds of established propositions have become irrelevant, being absorbed or quietly disappearing. I would argue that both interpretations are probably correct.

AI Won the Bingo

Nowhere was this churn more visible than in Hall 7. That was where the AI bullshit bingo card filled up fastest.

Almost every product had become “AI-powered”. Many vendors seemed having discovered “agents”. A surprising number had apparently built an “agentic platform” since NAB. Yet when you asked what those agents could actually perceive, decide and execute independently, the answers were usually less convincing. In one embodiment, although one of the best known tech vendors in our industry, by simply bolting on Google Gemini.

A language model connected to a product interface is not an agent. A chatbot that launches a predefined process is not agentic merely because the button has been replaced by a prompt. A sequence of API calls is still bare workflow automation unless the system can reason about objectives, context, exceptions and consequences. This is not an argument against agents – rather the opposite and an appreciation in favour of precision

Avid presented Content Core as the intelligence layer for “agentic media production”. Its new Assist framework is intended to let organisations build governed agents that operate across production systems. That is directionally interesting, but much of what was shown remains an early demonstration. The video-understanding component relies substantially on Google Gemini. The browser editor comes from CuttingRoom. Calling this a coherent agentic ecosystem may eventually be justified, but it will require some serious further integration.

Adobe’s approach is more tangible and, arguably, more honest. Premiere can organise assets, rename clips, add markers and generate video, music or sound effects directly in the timeline. After Effects now has an AI Assistant that can inspect a project and execute multi-step technical operations. These are useful capabilities that remove friction from a craftsperson’s working day.

Blackmagic Design appears less interested in winning the vocabulary contest. DaVinci Resolve continues to add tightly scoped neural capabilities for editing, transcription, audio isolation, masking, reframing and grading. Its advantage is not a more animated chatbot. Its advantage is control over an unusually broad, integrated post-production environment, combined with a commercial model that does not make every innovation another recurring subscription.

The industry’s inbreeding crisis

Evan Shapīro rightfully described media as suffering from a “groupthink epidemic”. He compared its leadership class to the Habsburg dynasty: too much inbreeding, intermarriage and cronyism, producing a leadership cohort with similar backgrounds, incentives and blind spots.

The comparison was designed to provoke, but it explains something important. Since key decision takers are attending the same conferences, hiring from the same competitors and are briefed by the same consultants, the industry has become very good at repeating consensus, and less at recognising when that consensus is increasingly getting economically absurd.

One such absurdity is the assumption that more technology spending automatically produces a healthier media-technology market. At the Devoncroft Partners Executive Summit, Joshua Stinehour estimated that media companies spend approximately $67 billion on technology (although a large proportion of this goes to public cloud, general purpose IT, and services). Devoncroft’s published analysis shows that aggregate revenue of traditional MediaTech actually declined by 2.4% in 2025, with only slight growth expected in 2026. Managed services now represent more than half of the market, while many traditional product categories continue to contract.

Meanwhile, a growing share of the real technology bill has become invisible to the traditional media-technology sector. A good chunk of it is captured by hyperscalers, general-purpose SaaS platforms, AI model providers, consultancies and infrastructure vendors.

The money has not disappeared. It has migrated. This migration illustrates a peculiar form of collusion between media and #BigTech. Media companies depend on technology platforms for infrastructure, distribution, audience discovery, advertising and increasingly production intelligence. Those same platforms compete for rights, viewers and advertising revenue.

Amazon is simultaneously a cloud supplier, technology partner, retailer, advertising platform, distributor and studio. The proposed Paramount acquisition of Warner Bros. Discovery is another expression of the same search for scale: combine more studios, networks, streaming services, data and negotiating power, then promise that consolidation will create efficiency.

Such “efficiencies” are typically announced in boardrooms, but experienced as redundancy elsewhere in the organisation. It may reduce costs and strengthen global distribution, yet scale can also increase complexity, slow decision-making and widen the distance between commissioners and audiences. The future will tell whether these economies of scale will be able to compete with more nimble content producers competing for localised attention.

Industrialisation, finally

The answer cannot be to reject automation. Industrialisation in production and post-production is imperative to withstand cost pressure.

By industrialisation, we don’t mean turning creative decisions over to machines. It is about eliminating repetitive technical work, broken handovers, duplicate data entry, unnecessary file movements and the endless reconstruction of information that was known earlier in the process.

Unpopular opinion: the essential ingredient is not AI, it is metadata control. AI becomes valuable where reliable metadata is not available or can’t be captured deterministically: transcribing speech, interpreting images, identifying story elements, translating dialogue or suggesting useful classifications. Everything else should preferably be based on controlled data, business rules and transparent automation.

Many conversations we had at IBC supported this view. Customers were interested in semantic search, brand detection, automated delivery, story building, subtitle condensation and production wizards. They also asked for metadata to be exposed inside Avid and Adobe, and for headless integration with Resolve.

The cloud correction

Joshua Stinehour argued that cloud adoption is real but incomplete. I agree with the first part, but “incomplete” may imply that every organisation is still travelling towards the same destination. We increasingly see selective cloud repatriation instead.

Customers are not abandoning cloud services. They are becoming more deliberate about what belongs where. Compute-intensive, temporary or highly collaborative processes remain excellent cloud use cases. Permanent custody of irreplaceable high-resolution material is an completely different ball game.

Producers have discovered that when a subscription ends, access to workflow context, proxies or even assets can become complicated. Egress costs, unpredictable consumption bills and vendor dependency are now board-level concerns. Nobody wants to discover that leaving a supplier also means losing practical control of their own catalogue.

The future is therefore not “all cloud”. It is intelligently hybrid: cloud where it creates measurable operational value, customer-controlled storage where ownership, continuity and economics require it.

A prediction that did not age very well

In our 2021 technology outlook, we predicted that AI and cloud automation would contribute to a macroeconomic shock in media. That prediction did not age particularly well, at least not on our timetable.

As an industry, we overestimated the speed at which individual AI capabilities would become dependable production systems. We underestimated organisational inertia, integration work, quality control and the stubborn importance of metadata. A model can produce a remarkable answer in seconds. Building an accountable process around that answer can take years.

On this point I fully agree with Joshua Stinehour: the winners of the AI arms race, producers and media-technology vendors alike, will be those that connect AI to real economics. Not the largest model. Not the most agents. Not the most impressive prompt.

As heard on last year’s DPP leader’s briefing, AI doesn’t fix chaos. It scales it. Control the metadata, understand the workflow and measure the economic outcome. Only then add intelligence.