Former Cisco CEO: Nvidia's AI Dominance Mirrors Cisco's Internet Boom, But Market Dynamics Differ (wsj.com) 24
Nvidia has become the U.S.'s most valuable listed company, riding the wave of the AI revolution that brings back memories of one from earlier this century. The last time a big provider of computing infrastructure was the most valuable U.S. company was in March 2000, when networking-equipment company Cisco took that spot at the height of the dot-com boom.
Former Cisco CEO John Chambers, who led the company during the dot-com boom, said the implications of AI are larger than the internet and cloud computing combined, but the dynamics differ. "The implications in terms of the size of the market opportunity is that of the internet and cloud computing combined," he told WSJ. "The speed of change is different, the size of the market is different, the stage when the most valuable company was reached is different." The story adds: Chambers said [Nvidia CEO] Huang was working from a different playbook than Cisco but was facing some similar challenges. Nvidia has a dominant market share, much like Cisco did with its products as the internet grew, and is also fending off rising competition. Also like Nvidia, Cisco benefited from investments before the industry became profitable. "We were absolutely in the right spot at the right time, and we knew it, and we went for it," Chambers said.
Former Cisco CEO John Chambers, who led the company during the dot-com boom, said the implications of AI are larger than the internet and cloud computing combined, but the dynamics differ. "The implications in terms of the size of the market opportunity is that of the internet and cloud computing combined," he told WSJ. "The speed of change is different, the size of the market is different, the stage when the most valuable company was reached is different." The story adds: Chambers said [Nvidia CEO] Huang was working from a different playbook than Cisco but was facing some similar challenges. Nvidia has a dominant market share, much like Cisco did with its products as the internet grew, and is also fending off rising competition. Also like Nvidia, Cisco benefited from investments before the industry became profitable. "We were absolutely in the right spot at the right time, and we knew it, and we went for it," Chambers said.
Déjà vu (Score:2)
No really! This time is different.
Re: (Score:3)
Re: (Score:1)
Their revenue yes, but looking at the same graph of their stock price is a very different story. Peak was $82 in Y2K, and it's never been higher than the mid $50s since. Adjusting for inflation is even worse, that $82 peak is $152 in 2024 dollars.
Yes, there are some dividends in there, but any shares bought at the peak and held for 24 years are still in the red.
The problem I could see would be ASICs (Score:2)
50 years ago the companies would compete but these days CEOs just say "fuck it, let them have all the engineers, competition is for losers". Still with $3 trillion at stake it might move Intel. AMD doesn't have the capital and
Re:The problem I could see would be ASICs (Score:5, Interesting)
The nVidia offering is basically doing purpose built design.
With Bitcoin, nVidia never made an effort to particularly become better at hashing, so there was an opportunity to have some of their crypto boom go away to ASICs (but Ethereum mining was best served by GPUs until proof of work was migrated away from).
AI started as "all we have are CPUs and GPUs, and GPUs are much better for now", and nVidia leaned into it and started putting appropriate hardware design *everywhere*, with parts covering a pretty wide gamut of heavily optimized for the operations AI while limiting the other capabilities a bit to a little bit of AI enabled stuff while doing a lot of other operations.
In short, nVidia benefitted from, but never took seriously the cryptomining (in fact making efforts to cripple their products). They are however taking the AI stuff extremely seriously.
So seriously in fact that there's expectation that nVidia's blackwell datacenter stuff will be outperformed by AMD's counterpart offering for FP64, largely because so much of the Blackwell design is "AI friendly" to the level of compromising other capabilities.
Re: (Score:2)
The nVidia offering is basically doing purpose built design.
Not really. The current crop of GPUs is not particularly optimized for AI, it just happens that GPUs are pretty good at AI as-is.
With Bitcoin, nVidia never made an effort to particularly become better at hashing
It's an understatement. They had a driver that detected and slowed down hashing on purpose.
Re: (Score:2)
Re: (Score:2)
This is not the best overall architecture for neural networks, but it's good enough, and it was easy to add to the existing CUDA architecture.
Re: (Score:2)
Re: (Score:2)
sustainable income (Score:3)
AI companies need sustainable income. Right now there are huge investments in the future, bets on what is happening in the future. At some point, those costs need to be recovered and produce a profit. The current environment is running on investment and is unsustainable as-is. How many reading this are ready to pay $/month for AI services? There is a shake-out coming at some point.
Re:sustainable income (Score:4, Interesting)
How many reading this are ready to pay $/month for AI services? There is a shake-out coming at some point.
How many assume AI is being built specifically for those reading this?
AI-enhanced (even mildly) analysis is a benefit to corporations, not individuals. And if corporations find even a 10% benefit to take advantage of in that analysis, they'll spend that 10% no matter how big that number is.
I do see a shake-out coming for this AI personal assistant nonsense. But then again the individuals being targeted for that product already waste a lot per month on many other forms of entertainment addiction anyway. Par for the course.
Re: (Score:2)
I assume there are lots of IT decision makers that read Slashdot, we are not just consumers.
Different this time (Score:4, Insightful)
One constant of stock market bubbles is people riding the euphoria always say "it's different this time".
Re: (Score:2)
It's just to keep the stupid money coming in.
My Prediction (Score:5, Interesting)
Right now, everyone is rushing to develop the best AI. This involves, at the very least, coming up with an algorithm to train the AI and another algorithm to query the AI. Both of these tasks require extremely large amounts of computing power and Nvidia has benefited immensely from companies using their hardware for these tasks. However, capitalism will eventually do what it always does and whittle the widespread competition down to just a few major players. That will lower the demand for Nvidia hardware. And then, the AI companies that are left will have to deal with an enormous number of queries against their systems which will inevitably cause them to turn their attention to developing special-purpose chips that can process their proprietary algorithms as efficiently as possible. That will be another huge loss of demand for Nvidia which will likely cut them out from most of the AI market and relegate them back to making products solely for gamers, a few scientists, and whatever is left of the cryptocurrency market. This is obviated by the fact that many companies which tangentially use AI have already started designing their own chips, such as Apple and Tesla, and the fact that companies solely focused on AI have already expressed interests in creating their own silicon, such as OpenAI. And the fact that it's easier than ever to design your own chip and have it fabbed by a third-party just makes this path extremely obvious for whoever wins the market.
No matter how I look at this, I just can't imagine a future where Nvidia lives up to the current expectations of the stock market. Perhaps if they buy a major AI company and get an early start on developing chips that process that specific algorithm, but then it comes down to which company to buy and that's a total crapshoot at this point. They're in a weird spot where they're fortunate enough to be on top of the world but not fortunate enough to have many reasonable possibilities available to maintain their position.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
A few questions about your analysis:
However, capitalism will eventually do what it always does and whittle the widespread competition down to just a few major players. That will lower the demand for Nvidia hardware
Why would convergence reduce nVidia's market share? During the PC revolution, it didn't matter if IBM or Dell or Compaq won - Intel wins either way. nVidia is in Intel's position. It doesn't matter if Anthropic, or Meta, or OpenAI, or Alphabet wins. The chip suppliers win no matter what. Going further up the chain, TSMC wins regardless too. Did I misunderstand?
then, the AI companies that are left will have to deal with an enormous number of queries against their systems which will inevitably cause them to turn their attention to developing special-purpose chips that can process their proprietary algorithms as efficiently as possible.
Amazon, Microsoft, and Google are dealing with "an enormous number of queries" but didn't develop their own
sadly with ai /. comments are \. :( (Score:2)
Re: (Score:1)
Re: (Score:2)
Or they're just being paid obscenely to develop AI-somethings by the MBAs.
The Difference is... competence (Score:3)
Cisco rested on their laurels and became largely incompetent, with security holes left and right and embarrassing problems like switches that had to be updated for Y2K because the BACKPLANE was somehow incompatible with four digit dates. (The hallowed and venerable Cat5k among them.)
Nvidia has reinvested and built the best hardware AND the best software, to the point that their profits are now over 50% of their earnings.