Op-Ed: AI revenue numbers aren’t stacking up: The markets are responding to every breeze

This new approach to market behaviourism analysis definitely won’t reassure anyone. The markets are fighting their own numbers and the lack of numbers in key areas.

The resulting chaotic mess is patiently explained by The Economist in some depth, explaining the tides of “unreliable numbers” and market responses. The problem is that market people are getting tired of the chaos. Investors certainly are.

The numbers are “unreliable” in too many extremely expensive ways. After all this time promising rivers of gold, a large and rather pretentious dripping tap seems to be the result. Markets are overreacting as they always do, but they’re hitting direct market investments hard in the process.

The Economist points out Annual Recurring Revenue (ARR) as a case in point. This is a key metric, and one of the big issues is reporting quality. The ARR and many other core reporting topics are what’s worrying the markets.

If you’re not an accountant:

Comparing revenues over set periods is standard practice. Anthropic reported revenue of $16.8bn in Q3, after revenue of $65bn in Q2. Any questions? That’s a big bandwidth.

EBITDA (earnings before interest, tax, depreciation, and amortisation) is another contentious issue in every sense. EBITDA can be manipulated too look good , feel good, and be the basis of higher-value loans. It can also be the basis of massive hype. That’s not exactly popular.

Sources of revenue matter a lot. Pulling numbers out of a blue sky doesn’t cut the mustard with hard money investors. Breaking down and analysing revenue can’t be done with a few cooperative tea leaves and goodwill alone.

Depreciation and shelf life are definitely major things in AI market realities. The value of 2023 tech vs 2026 tech, for example, needs to be in the numbers. What are the real current asset values? A fossilized inventory isn’t necessarily an asset, even in the addled mental state of the US markets. Particularly in the AI sector where you can be blown out of the water by a new chip, LLM, and practically anything else.

The markets are right to be unimpressed and scared of themselves

Markets are often their own worst enemies. The sub-primes roped in every sort of numbers manipulation in 2008 and took the whole US domestic economy with them. Idiotic fraudulent loans were rated triple A. The house of cards never had a chance of standing up, and it didn’t.

Quite a lot of very sincere and often totally correct scatology goes into describing market behaviour. This time, positive and negative arguments are pretty strong in many ways:

First, the negatives.

This is also a frontier investment sector. It’s a whole new area of investment. Throwing money at things is what the market does. Hype is the adrenaline switch. Market mindsets are contagious. Does that sound like the last two years or so of frothing, flailing finances to you? Because that’s exactly what it is.

All of which leads to what, you may well ask. It leads to a jumpy market handling trillions of dollars’ worth of capital while walking a tightrope over the inferno of its own finances.  

A market panic in the trillions of dollars range can’t be anything but a colossal train wreck. The markets know that very well. Shaky numbers are the expected dead canaries in the coal mine. Meandering numbers are a terrible look. There’s plenty to worry about.

Now, the positives.

There’s just a bit of realism in iffy numbers at the moment. The AI sector and related businesses haven’t really taken hold. Revenue in the bricks-and-mortar sense isn’t yet a realistic expectation.

The macro data environment and collateral cashflow aren’t in any sort of normal business modes. The numbers are all over the shop, but that’s to be expected. Lack of regulation doesn’t help stability (or trust) in any business environment, but everyone wants regulation, and it does have to happen.

There’s reason to believe that AI operational costs like energy can be drastically reduced, improving profitability. The sector is responding to these major issues, not very efficiently and with stunningly lousy PR, but it’s at least pointed at solutions, not just more problems.

So it’s not all bad. Just inarticulate and insensitive.

Accountancy vs the AI sector

The fundamentals of accountancy do have every right to be there in AI businesses. Depreciation alone should have 24/7 care and due diligence, and that’s not an overstatement. Reporting needs to be responsive, not at a “whatever” level.

Let’s get a little bit terse, shall we?

AI has been promising the world for so long, and it’s not delivering much more than hiccups in far too many ways.

In many ways, it’s just a much more expensive way of doing the same things.

“Rehiring syndrome” has undercut one of the more repulsive tenets of AI, and has been a very unnecessary real disaster in progress.  

Chatbots don’t impress anyone anymore.

AI agents aren’t and can’t be trusted on any level. Nice going, geniuses, killing your own gimmick.

AI security issues alone could sink big balance sheets, vast ranges of infrastructure, and anything else at the event horizon without a trace, and that issue’s not even debatable.

Unreliable numbers are the last straw. Pretend to be competent and produce some real numbers.