Unlock Editor's Digest free of charge
Roula Khalaf, editor of the FT, picks her favorite stories on this weekly newsletter.
How did Nvidia – at times with Apple – turn out to be the most important publicly traded company on this planet? There are two answers. On the one hand, it’s the leading manufacturer of chips which can be driving the age of artificial intelligence. The other is that it has benefited from the magical mathematics of addressable markets.
Nvidia's founder Jensen Huang presented many recent products and predictions to visitors this week at CES, a serious technology trade fair in Las Vegas. Attendees applauded as he held a collection of unremarkable-looking black boxes and equipment. But the actual drama was reserved for his vision of “physical AI,” the subsequent frontier of artificial intelligence through which complicated models will produce not only words but additionally robotic actions.
Even for a corporation valued at $3.4 trillion, the numbers floating around are high. Only humanoid robots may very well be a $38 billion marketsuggested Goldman Sachs. But that's a small thing in comparison with autonomous vehicles, effectively robots without legs, which Huang sees as a “multi-trillion” industry – a claim echoed by his customer and Tesla boss Elon Musk. Manufacturing, which is ripe for robotization, is a $50 trillion market.
The theory is that each one of this requires data processing chips just like the ones Nvidia sells in bulk. It's plausible. Citigroup analysts recently estimated that by 2035 there can be 1.3 billion AI-powered robots, from vacuum cleaners to drones, moving around us. By 2050, about 648 million, roughly the population of Latin America, will resemble humans to some extent. These depend on models trained to navigate the world, device-level computing power, and fine-tuning as they learn through experience.
This is all grist for the mill of an industry fixated on “Total Addressable Markets” (TAMs). This concept, long utilized in the enterprise capital sector, is now common amongst large listed firms. The phrase appeared in 3,743 filings made with the Securities & Exchanges Commission between 2015 and 2020; In the five years since, almost 20,000 copies have been published. It's easy to see why: Identify a brand new market, take a future share of it, and there you have got it – a future revenue number that analysts and investors can start pricing in.
Nvidia is the poster child for expanding TAMs. Before the AI boom, the market consisted of gaming chips. This expanded dramatically to data centers. Robots – broadly defined – could make it much larger. Bank of America estimates it should have an 81 percent share of a $359 billion marketplace for “accelerator,” or AI, chips by 2030. However, expect a rise in TAMs and subsequently estimates of Nvidia's future value.
Huang adds to the drama that physical AI is reaching its ChatGPT moment. This claim is value considering for 2 reasons. The launch of ChatGPT in 2022 was the moment when recent possibilities, tools and TAMs suddenly got here into view. But it's also the moment when investors collectively lost their minds in regards to the value of future profits that don't yet exist – and maybe never will.
