Artificial intelligence companies may need to make good on promises to develop artificial general intelligence (AGI) in the near term in order to make up for the disparity between investments and profits in the industry.
Unfortunately, there’s still no scientific evidence that AGI — machines capable of human-level or greater reasoning — is even possible.
A growth market
According to analysts, the current AI market is largely anticipatory. OpenAI is one of the few highly profitable generative AI enterprises, and the difference between its revenue (about $3.4 billion, according to The Information) and the next closest competitors is massive.
What this amounts to is a capital dearth, or a negative flow, somewhere in the area of $600 billion, at least according to recent analysis from Sequoia Capital.
It bears mention that Sequoia’s figures are based on estimates of Nvidia GPU usage. With this in mind, it’s probable that the figures above are slightly deflated where global industry expenditures are concerned.
Essentially, the analysis indicates that AI companies need to make up more than half a trillion dollars in revenue to justify current expenditures — a figure expected to grow year over year.
Where’s the product?
While the current surge in investor and corporate interest in generative AI technology may have driven the market to all-time highs, including Nvidia’s brief tenure as the world’s most valuable company by market cap, many analysts are asking when the actual AI products or services that are going to sustain this level of growth are going to arrive.
As of yet, it’s hard to argue that generative AI has found a legitimate use case that will lead to exponentially expanding profits for those invested.
ChatGPT might be the sector’s flagship product, but there’s little reason to believe it’s going to suddenly explode into the mainstream.
Simply put, that $600 billion revenue mark will take decades to reach if OpenAI’s 10-digit profit margin is carrying the bulk of the market. Generative AI has yet to find the same kind of value proposition that machine learning has, yet investments continue to ramp up at the VC, government and corporate levels.
This could very well indicate that the AI market will soon enter an “AGI or bust” era where the viability of companies such as OpenAI and Anthropic will depend on whether they’ve made the right bets when it comes to delivering machines that can reason as well as humans.
On the negative side of things, those companies at the heart of the generative AI sector may be coming up on crunch time for revenues. If the market can’t justify Nvidia’s position at or near $3 trillion soon enough to avoid a potential drawback, then that $600 billion dearth for the industry could widen to the point of no return.
However, on the positive side of things, the point of no return won’t exist if the industry actually invents AGI. And Nvidia is also the key to this scenario.
As Sequoia Capital also pointed out, Nvidia is gearing up to launch its new Blackwell-based chipset (called the “B100”) for training generative AI. The B100 is purported to outperform the current industry standard for training models (Nvidia’s H100) by as much as 2.5x and will reportedly only cost 25% more.
If the experts believe it was possible to realize AGI before Nvidia’s latest and greatest chip was released, it should follow that it’ll be even easier with hardware featuring a 150% increase in power and efficiency.