I wish Money20/20 had made a trip to the zoo before opening its doors last week. Despite occupying four separate halls, there was a notable absence of the elephant in the room, the wolf at the door and the snake in the grass.

Where the hell was Big AI?

The debates were tilted so far in favour of one side that the most uncomfortable questions never quite made it to the stage. The position – held with remarkable confidence – was that speciality providers will maintain their moat. Regulation, complexity, consumer concern: a trinity of protection. Consolidation will happen, but inside the sector. The industry will absorb itself.

Now, technically, there were representatives on site. NVIDIA and Anthropic each appeared on a solitary panel. This was, however, not proportionate to their position and station. Nearly every panel would have justified a subheading: And can Google do this cheaper and faster?  

Last October in Las Vegas, Anthropic was on the main stage. Now the industry seems to have decided to have the AI conversation without the AI companies.

This may not have been a deliberate choice on behalf of Informa. The CEOs, CPOs and CTOs, too busy hyper-growing to show up, were replaced by people whose job is to agree with each other in public. The result was panels thick with phrases like “I’d like to emphasise what she was saying,” “if we could return to the last point,” and “I think there’s a lot of agreement up here.”

There was. That was the problem. Groupthink and tautologies lead to unfounded consensus. And unfounded consensus typically leads to lazy conclusions that melt rapidly in market’s sunbeam.

One confident conclusion from the 2022 show stays with me: “The battleground has moved from the tech stack to the user experience.” That feels inarguable until you look around at the situation four years later. It has never been more about the tech stack.

One way around this, in the absence of Dario Amodei: is to head deeper into the code. After all, attendees now include thousands of vibe coders with a greater understanding of programmes than in any year before. Yet it wasn’t. Someone told me the conference shied away from live demos because the technology was fiddly. But fiddly brings the unexpected. Fiddly is the pathway to actual insight.

Here are three inconvenient truths I kept waiting for someone, anyone, to address from an official pulpit.

Pricing power. Can you genuinely remove 50% of engineering headcount and maintain margin? The large players with nine and ten-figure valuations were flanked at the fringes by nimbler peers from southern and eastern Europe – smaller booths, lower overheads, overlapping claims. The competitive pressure is closer than the main stage suggests.

The partnership tangle. Lawyers are clearly losing the fight against sharing client names – “trusted by” and “our relationships” were plastered on booth sides in greater density than ever. What this actually reveals: European banks and payment companies now carry dozens of individual vendors inside their stack, performing narrow or overlapping functions. The logical endpoint is centralisation through a single provider, or just a few, each working off a common data registry – run by Anthropic itself, or someone like Snowflake. A viable ecosystem on that model would be smaller, with fewer participants.

The interface problem. The growing use of standard LLMs as a primary interface got almost no serious treatment. It was a feature, and not much more (‘you can run this right on your company’s Claude’). I get that: This entry was mostly composed inside of a customised Claude editing project. But by making replication easier, a common interface loosens the grip of individual programmes. Think of how many sources feed into a Bloomberg Terminal – easily added and deleted. The only certainty is the Professional Service itself.

In Hokusai’s 1831 masterpiece, the titular great wave off Kanagawa looms massively: several times the size of Mount Fuji. All seems lost. But in reality, it’s just forced perspective; even the largest wave in Suruga Bay couldn’t wash away the mighty Fuji-san. 

This wave from Silicon Valley may also break harmlessly. The last-mile complications of deploying base models for KYC, payments and deposits are genuinely hard. And the revenue demands of a trillion-dollar public company may redirect spending toward consumer applications, away from the experimental enterprise deals so common right now. It’s possible but not certain. Those arguments deserved a more thorough hearing on one of Europe’s largest fintech stages.

Instead, this is what we got: A press release put out by Money20/20 on the Monday after the event ended with this scintillating conclusion: “Conversations increasingly centered on how AI is transforming financial services from a productivity tool into a core business capability.”

Yawn.

Jon Schubin runs content for Cognito