Stop us developing AI to save humanity…and our balance sheets

Being an AI company is expensive. According to leaked financial documents, OpenAI is sitting on a staggering $20bn+ operating loss, driven by $19bn in R&D alone. Anthropic is supposedly faring better, but is also sloshing cash into aggressively advancing the technology.

They have little choice. To fall behind now means surrendering ground in what’s likely to be a winner-takes-most market.

But, as it stands, continuing to push the technology to new heights is costing far more than it’s generating. And most use cases for the tool today work just as well with older models. A huge push from enterprise to manage costs through harnesses and FinOps, both intrinsically linked to diverting workloads to the most appropriate model for cost reasons, will weigh even heavier on the business logic of ongoing investment at the frontier.

In some cases, the cash is tied to an ideological commitment to AI. Altman famously craves AGI. He’s declared reaching it almost as often as he’s launched new versions of ChatGPT. Which is to say, a lot. But in reality, pushing at the frontier serves little more than bragging rights over competitors. Case in point, in recent weeks Anthropic, OpenAI, and Meta have all insisted what they’re working on is so dangerously good it should be stopped.

A marketing ploy? Perhaps. But the answer may be as simple as a desperate bid to find fiscal breathing space. By convincing policymakers to enforce a moratorium on development, cash-strapped companies can hit pause on R&D spend and start building a viable balance sheet, particularly before they IPO. If they unilaterally hit pause, they surrender ground. But if everyone has to stop, happy days. 

International developers complicate matters. Which is why the language has shifted in recent months to global restraint. Lest humanity be overcome. China is unlikely to do much. But that doesn’t matter in the medium-term—businesses squeamish about tapping into Chinese hyperscale cloud over the last decade won’t suddenly warm to their AI. And average consumers aren’t about to host and run raw weights locally. Just getting US regulators to slow down national AI companies should do the trick.

So, as the bills continue running up, we can expect more outcry from the AI giants: Regulate us, hold us back, for goodness sakes. We can’t afford to continue as we are.