Join institutional investors, analysts and strategists from the world's largest banks: Subscribe today
Already have an account? log in
Nvidia Scores Historic One-Day Value Gain On ‘iPhone Moment’
"I think you're seeing the beginning of -- call it a 10-year transition to basically recycle or reclaim the world's data centers and build it out as accelerated computing," Nvidia CEO Jensen Huang told analysts, on the company's earnings call.
As discussed here on Wednesday evening after Nvidia wowed Wall Street by guiding 50% above consensus for Q2 revenue, Huang painted a grandiose picture of the future and his company's place in it. "We were in full production when the ChatGPT moment came,"
His point seems at odds to me with the push to public Cloud. If businesses transition fully to the public Cloud, then that means they are actually reducing the data center footprint. With firms consolidating their data centers to public Cloud data centers, that would imply there might be less demand for GPU’s which are superior at running things like generative AI and machine learning.
This also ignores that most firms are still running CPU driven workloads for their production apps.
I also get a sense that AI is going to follow a boom/bust consolidation pattern where one of the players will come out on top similar to Google in the search space. For those of us old enough to remember the period where you had Yahoo, Ask Jeeves, Google, Alta Vista, etc. all vying for the search space until Google developed a superior search algorithm and basically took over; I envision one Generative AI becoming better than the rest and doing the same thing. This would imply significant short term growth similar to Cisco in the build up to the world wide web for Nvidia, followed by significant stagnation as adoption peaks.
I could also see AMD and Intel push for improved GPU production to compete for this market in the next 3-5 years which will impact that 10 year growth trajectory.
I’m not saying he’s wrong in his prognostication but it seems to ignore other factors in tech that might impact future growth for Nvidia.
There won’t be one dominant player, but might be an oligopoly with you-know-who at the top. That being said, I shared a memo allegedly from someone at Google basically saying the open-source models are already better than what Google is producing. I see this market evolving more like the internet where it opens up doors to everyone and is more akin to infrastructure than the tool that search is.
The biggest threat to Nvidia is more efficient models, but it’ll be a while before that slows down this gold rush. Intel is done and will slowly die off. They are too far behind and lack the innovative spirit to catch up. AMD will stick around and play the same role they did when Intel was king: second banana that the industry keeps alive to maintain some semblance of competition.
That reminds me of the chatter from Apple fans when the iPad was released: “No one will ever buy a laptop again!” and “Microsoft missed the boat – they are finished!”
Generated AI will not replace all more general purpose servers.
But it does not pay to stand in front of a stampede.
Yeah… but, there was also the famous “No one will ever want a touchscreen phone” soundbite after the iPhone was released. Apple won that one.
Not sure of that comparison. But I recall an exchange I had with a respected tech analyst who had written something like that. I asked him if he had written his piece on an iPad. To his credit, he answered “No, of course not. That would be a real pain in the butt.”
Folks are really starting to catch onto the deflationary body-blow of GAI. Many happily employed and ‘moated’ people are still not worried, their lovely and safe middle-class jobs as CPA’s and CFA’s haven’t gone to China and are looking forward to seeing 100 more Marvel movies being made for a fraction of the creative and VFX costs in the past. If the MAGA’s felt cheated by the ‘globalizing’ of their jobs to China, the Polo-shirt and khaki-wearing weekend golfers will be introduced to a precarity they thought they were immune to. Don’t worry, AI will steer them to ‘news’ and ‘information’ that keep their anger targeted and distracted by scapegoats.
At least Moore’s Law is still ticking like a time-bomb. As you have mentioned many times, all that matters to anyone in a C-Suite, is making decisions to expand margins at any cost to anyone but themselves. I’ve likened this situation to yeast in a barrel of percolating beer. The yeast will happily do what they do, enjoying the sweet sugared water, until they all die as their God-given right to output methanol makes the beer poison enough to become sterilized. The beer will be wonderful, I’m sure.
It’s been very difficult to stay positive. The gee-whiz, it’s all so amazing comments on LinkedIn are masking what I perceive as a growing discomfort as this deflationary force becomes more understood. I fear our species, much like the yeast in the barrel is actually incapable of stopping what it has always done. For instance, the largest driver of tech has always been weapons. I don’t see that changing until our genome itself is manipulated, which is also as inevitable as the yeasts methanol. Sorry, I’m ranting…
Imagine the world with GAI running on quantum machines. My hope is for a future like Star Trek, but sadly, Star Wars is a far more likely future.
I couldn’t love everything about this comment more. As a former amateur brewer my hope is that we are a high flocculation yeast so that we can at least live long enough to see everything else on the planet die before we do. And as much as I dreamed of a future of humanity like the Next Generation, I fear we are headed more for the Undiscovered Country.
Good rant, might become prophetic
Ethanol, not methanol.
Great comments, I’m inclined to agree with both of them.
Anyone betting against generative AI at this point is begging to lose money. It’s frothy and will remain so, but it’ll be a while before the hype cycle dies down. It’s one thing for wannabes and has-beens to jump on the latest tech trend like blockchain and crypto, but when every major tech company is saying this is revolutionary technology, it might be worth listening.
Another example of an industry that is likely to get disrupted: https://www.bbc.com/news/health-65709834
I don’t know if that’s good or bad for pharma companies, but genAI might actually be the first thing that reverses the healthcare inflation in forever. As Pyrognosis pointed out and I’ve been saying for a while, this will cost people jobs and white collar folks will feel the same precarity that blue collars experienced when we started outsourcing manufacturing. Expect a lot more long-term unemployment in tech and reduced salaries unless you know how to use genAI effectively.
The mega-tech/AI group is sucking the oxygen from the rest of the market, is what it feels like.
Painful for active managers, I’d think: how many managers have 25-30% of their portfolio in 5-6 names? I sure don’t.
Meanwhile I see consumer ex-AMZN rolling, communications ex-GOOG/META weak, industrials rolling, staples stalling, etc etc. The S&P 500 isn’t reflecting economic fears, but the S&P 49X are.
On NVDA, I’m reminded of the company’s heyday in bitcoin mining, before ASICs took over from GPUs. GOOG and AMZN have their own AI chips (TPU, Tranium, Inferentia) that are purpose-built for AI, without the features that GPUs need but AI doesn’t, but since they don’t sell those chips NVDA has the merchant market all to itself for now.
It’s unclear to me how long this phase (many companies scrambling for H100s to build their own massive-scale general-purpose LLMs) will last. It seems plausible to me that there is a finite number of 500 billion token models that will need to be trained, after which continuing training needs will be much lower (can’t be otherwise, or genAI won’t be economically viable) and more companies will use smaller/specialized models and/or use hosted cloud-based models (many running on the equivalent of ASICs) and the demand for H100s will fall off.
NVDA has a deep and powerful software stack, that it’s been building since the early days of GPU computing, and will try to use that to become more economically embedded in AI than merely the silicon provider, with high-margin recurring usage revenue. Remains to be seen if the other players are willing or compelled to share revenue.
It’s sort of a TSLA situation – selling the cars isn’t remotely enough for the valuation, it needs recurring revenue from the installed base of robo-cars.
NVidia is an expression of the old pick-and-shovel play. Deep learning doesn’t need to win the future; it just needs to make a go at it!