Not too many years ago, people used to queue outside the shops to get their hands on the latest Apple iPhone. These days the razzmatazz associated with a new release has become the exclusive domain of Nvidia (NASDAQ: NVDA). But unlike previous new chip launches, the stock price didn’t surge after Nvidia’s latest debut. The question now is whether it can get its mojo back.
Super Bowl of AI
Last Tuesday (18 March), at a packed developer conference in Silicon Valley, Jensen Huang unveiled a new line of chips. Dubbed the Rubin, so named after the astronomer who discovered dark matter, the AI superchip will be up to 14 times more powerful than the Blackwell.
His strategy is obvious. Despite the threat brought on by the release of DeepSeek, he very much sees more computational power as the future of AI. Indeed, he claimed that the world will need 100 times as much computing power than was envisaged just a year ago.
Echoes of the past
I’m very much a firmer believer that one can learn a lot from how previous technological breakthroughs shaped the development of an industry.
Back in the 1970s, mainframe computing was king, and IBM was at the centre of that universe. With the invention of the PC in the early 80s, the company failed to appreciate its threat. It even described them as an “entry system”. The expectation was that anyone buying a PC would move up to more powerful mainframe computers.
Doubling down
IBMs motto back in the day was ‘Think!’. Think, yes; but not like a bunch of clones. In the face of the threat posed by the emergence of the PC, the company doubled down, sinking even more money into main frame computing.
As computers became commodities, the company’s margins eventually collapsed, and it nearly went bankrupt. Even the great Warren Buffett was invested in IBM and got burnt. Maybe that explains why he has stayed away from the hype surrounding Nvidia.
Future of AI
I am not trying to make any direct comparisons between then and now. What I am cautioning investors to do is to stop and think.
DeepSeek has changed the narrative around AI, of that there is no doubt. OpenAI, by far the most popular and widely used large language model (LLM) has effectively been commoditised. That includes its reasoning model, o1.
DeepSeek may fade in obscurity. But that doesn’t matter. What they have shown is that you don’t need the latest and greatest chips, which only Nvidia has, to make a workable LLM.
It’s not just Nvidia that is equating greater power with better. The hyperscalers are too. Last month, the Wall Street Journal reported that Microsoft’s total capital expenditure in 2025 would top $90bn. Meta has already committed $65bn this year. These are truly eye watering sums of money.
To me, it feels like the entire tech industry is stuck in a delusional epoch. How the AI industry evolves in the years ahead is anyone’s guess. If these bets spectacularly backfire, then the tech giants of today could suffer the same fate as the dinosaurs did 65m years ago. Therefore, until I see a clear path for the AI industry, I won’t be investing in Nvidia.
