What are the economics of the AI industry?
The gold found in the ground is currently far from paying for the shovels.
TLDR Summary
The AI industry is a joint effort of many companies to bring the benefits of deep learning to the masses, both businesses and consumers. This means that problems are being solved through artificial neural networks as opposed through human labor or human written code.
Right now, Jensen Huang is pretty much the only guy making money in this industry. NVIDIA’s annual operating profits have jumped by about $100bn since the launch of ChatGPT. Everyone else is incinerating money with negative cash flows likely exceeding NVIDIA’s profit jump, potentially making the total financial impact of AI negative as of now. Big Cloud will likely generate $40bn in AI revenues this year with opex and capex north of $120bn. And the AI start-up space will likely generate $20bn in revenues with about the same amount in operating losses.
Hence, AI is a zero-sum game at best so far. And it’s not just that. Its operating leverage is incredibly poor which makes is extremely hard for economies of scale eventually make it a positive-sum game.
Why are those losing in that game playing it anyway then? The answer has something to do with corporate incentive structures and investor preferences.
The tale of shovels and nuggets
The gold rush analogy is quite popular to describe and predict the dynamics in the AI industry. You can find my version of it in the article below from two years ago.
Historical precedents suggest that the first to monetize on a gold rush are those selling the equipment. Should the opportunity live up to its expectations, those actually digging for gold will eventually outearn the shovel sellers. Otherwise the shovel industry can’t sustain its revenues.
Almost three years after the launch of ChatGPT seems to be like a good time to check on the financial health of this gold rush. Let’s see how those making and operating the equipment are faring.