Forget about the market reaction for a minutes, which is always difficult to interpret. More holistically, aren't we underestimating their dev ecosystems (CUDA etc.)?
Think of the dev ecosystem effects in all the various permutations, e.g. from Windows or Apple to AWS to more granular like PyTorch.
Those advantages aren't easily eliminated just by more competitors or even 'better' competitors.
But holicistally, if you were to buy the entire Nvidia company today, even if those advantages continue untouched for the next 10 years, and competition doesn't catch up, and the AI boom continues ... you're still not getting your money back. That's how high the valuation is.
Jim Keller calls CUDA NVidia's swamp (not their moat) and I think that's so so so accurate.
Nvidia's fief has a lot of folks in it, who've built sloppy stuff all over the place because they can. Most of that won't get replaced.
But the modern tooling like PyTorch and its underlying JAX doesn't need or care about this legacy. The future can build far more intimately & with deliberation to folks who haven't such large swamps around their ecosystems. We can tune systems more efficiently & carefully when our vendors don't keep us so so far away from the core hardware. Nvidia has to keep maintaining this huge footprint of software, keep it going themselves, supporting a vast legacy, while everyone else gets to innovate forward, with much better access to the hardware.
> Nvidia has to keep maintaining this huge footprint of software, keep it going themselves, supporting a vast legacy, while everyone else gets to innovate forward, with much better access to the hardware.
When I hit an input box on my phone, a keyboard pops up on the screen, with the first row saying QWERTYUIOP, just like the Sholes and Glidden typewriter of 1878. I'm waiting for innovation on this first, then I'll look for innovation beyond CUDA.
Wat. The keyboard you pop up this way will autocomplete your words, autocorrect your typos, and allow you to input whole words by swiping, twirling gestures. It's very much not your legacy typewriter keyboard, even though the order of the letters is the same. Maintaining the order of a couple dozen keys is not expensive or hard.
I thought the opposite: if CUDA were easy to replace, wouldn't it have been replaced already?
PyTorch supports both ROCm and Ibex, but it's support is spotty and incomplete to the point where it becomes a pain to use despite AMD spending billions on improving its software layer.
NVIDIA is reaping a decade of design decisions on CUDA compatibility made less sloppily than its competitors.
Because markets are forward-looking. They know this cycle of the AI hype bubble is coming to an end, so to put it simply, this earnings release is as good as it will get for a while. The market believes there's nowhere to go but down from here.
The market thought Nvidia was toast when cryptocurrency flopped over. The demand for GPGPU compute is pervasive, and even if LLMs go nowhere then there's still applications for CUDA and high-performance server compute. When the dust settles, Nvidia's lifeline is less the AI demand and more their server niche they fill.
Besides Apple, who can't be forced to build server hardware at gunpoint, Nvidia is the only TSMC customer bringing HPC to the consumer. That's an enormous segment they can capitalize on, and AI is really just the tip of the iceberg. With so many other OEMs laser-focused on AI performance, they're entirely glossing over the flexibility that made CUDA successful in the first place. Pour another one out for OpenCL, the industry hardly knew ye.
I should have said "customer" there, to be more clear. Nvidia competes with Apple for fab space on the latest nodes, which is more what I wanted to zoom in on. Apple has zero interest in courting Nvidia's customers, which leads me to believe that Nvidia has a practically captive market.
Everyone already knew they were going to be impressive so the stock price reflected that, and so it fell because they were expecting it to be slightly more impressive than it already was.
The question is if they are impressive for a company worth $3.1 trillion. Annualizing this quarter's earnings it has a P/E of 45. And earnings "only" grew at 60% annualized, slower than in the past. (Revenue growth is also a bit slower as a percentage)
So market likely sees decelerating growth which can't justify such a high P/E. e.g. if growth fell to say 30% YOY in a year, nvda would be seen as overvalued today.
First Time?
Markets are always forward looking. It's always future revenue / profit expectations. Sure current quarter numbers will tell half-the story about the future, but the highest impact is always what the future story is.
Earnings beat is almost always Mainstream, Newbie, Normie focus.
Quarterly Earnings can be (legally) cooked to tell whatever numbers you want. Revenue Growth and Operating Margins always tell the bigger story
Probably not unless there is a lot of retail on one side of the 0D options book. Also, it would be a reverse squeeze, not short ladder, which is largely a mythical beast invented to salve day traders.
The last big surprise were the May 2023 earnings, it went from closing at $31.26 to closing at $38.95 the next day (up 24% in one day) after the earnings announcement.
Now it is the company with the second largest market cap in the world and enormous growth, so analysts are watching it more and talking about it more. Results are more baked in.
Also, it was announced the next-gen Blackwell chips had a flaw found late in the design and will probably be delayed three months, which will also affect revenues.
It's manufacturing capability and there is no absolute slowdown. Nvidia has been increasing data center revenue by exactly $4b for many quarters QoQ. The only difference now is that the base is higher so it looks like growth is slowing down on % view.
Nvidia is totally supply constraint as it can easily be seen. They also underguide every quarter for 2 years and beating it. But if you look closely then you can easily see the $4b QoQ growth. That is the constant there since 2022.
Blackwell will add additional supply next year. I don't think Hopper wills top shipping and I don't think demand will fall. So next year QoQ absolute growth could go up which will probably surprise to the upside.
It depends on the market sentiment, it can go both ways. On one hand a stock buyback can be considered a bullish signal, the company is strong and doing well, on the other it can be a bearish signal, the company is out of useful things to do with all the cash it has laying around so it's buying back stock.
What is the case for continued growth of Nvidia? Every single sizeable entity (Google, Microsoft, Intel, AMD, etc.) is working furiously to replace them. I can imagine plenty of reasons that Nvidia sales would remain strong, but I understand less where more growth would come from.
Simple, while they are working to replace them, Nvidia is working to replace them as well. Nvidia AI Enterprise has been availble for 2 years and growth is exponentially. It's $4500 per GPU license and SaaS business for Nvidia. Last year Nvidia SW run rate was $1b, this year it's $2b according to CFO. SW revenue is 100% YoY and expect that number to rise.
If Nvidia can utilize $4500 for very 2nd GPU they ship in then next 10 years then in 10 years Nvidia will make more revenue annually with SW only than today total. And that is just the beginning and shows the power of a platform.
Do you think Apple dominates because of the great iPhone HW or iOS? Think about that and you'll understand why Nvidia will continue to grow and surprise.
Intel market cap plus AMD market cap is less than one ninth of Nvidia market cap, so the market opportunity you describe can't be what is keeping Nvidia stock price as high as it is. I.e., the non-AI server hardware market opportunity is just not that lucrative.
None of those things has successfully built a platform in the way nvidia has with CUDA, although Microsoft is well-positioned to do it eventually.
The reason you’re confused is you are thinking of the peers of NVIDIA as being AMD and Intel and not Apple and Nintendo and Sony, or mainframe vendors, etc. The reason they are different from AMD or Intel (or I think we can throw Xbox into this pile too too) uses AMD or intel because they want to use AMD or intel, they use them because they are a convenient compile target for code that you’ve gone to the effort of making very portable. And that’s a commodity market, where the other things are platforms with an organic userbase that’s not just there because it’s the cheapest thing.
(this is true of gaming gpus too btw, other than AMD APUs being very compelling for handheld and laptop gaming, and maybe Lunar Lake. whereas people want to use nvidia - that’s the actual meaning of “mindshare” and not the “mind control field” AMD fans assume mindshare to be. People who use the product for the product’s sake and not just pure cost expediency.)
The other thing I think people are sleeping on is the mediatek partnerships. They already announced three different markets (some are already launched) - automotive, smartphone SOCs, now monitor controllers too. They will almost certainly be the vessel for Nvidia’s desktop chip next year too - functionally they are partnered with /oem for nvidia now. Nvidia already has PNY to be their Foxconn/Sapphire, now they have Mediatek to do the boring plumbing work in their designs. Nvidia simply doesn’t have the staffing to tap these markets and staffing up with a bunch of boring suit and tie workers would destroy what makes nvidia tick, now they can go tap these other markets.
I think people really really overstated Nvidia’s unwillingness to license their IP etc. Buying ARM was a play for getting GeForce as the default graphics IP for basic smartphones and shit, and giving CUDA access to that market. Well, getting your IP into mediatek is the next best thing. Thinking Jensen wanted to buy arm to sell more tegras is pleb tier thinking and ambition. Nvidia didn’t pony up $40 billion to burn it all down. That’s fucking stupid.
There’s obviously a point where it gets stupid, but nvidia is genuinely still a company with a huge amount of growth potential and thinking they can sustainably lock in most of their current earnings isn’t too crazy. The real counterweight is that it’s a one man show and Jensen is getting old. But it seems like maybe Catanzaro is being positioned as a potential heir, what with him being the public face of dlss as well as the AI effort as a whole there. He seems like a logical choice of heir.
Fiscal years are unaligned with regular years and are usually offset by a few months. Different companies may choose different start months for their fiscal year.
It doesn’t mean much beyond their accounting processes.
OpenAI, Nvidia, and friends have been leaders in pushing compensation packages past $1m/yr. This uplift will trickle down into tech roles in any workplace.
Most people in the world are not American. We have differing retirement plans and invest in local markets, or all sorts of international markets. The NASDAQ is one of many.
Most people reading this will indeed have zero investment in Nvidia.
"One of many"... Nasdaq & Nyse have capitalization equivalent to pretty much the rest of the world.
Most people reading this, if they have any money in any stock exchange, they are directly or indirectly exposed to Nvidia, whether they like it or not.
Absolutely, but capitalisation does not necessarily reflect the quantity of individual investors.
If you're talking about the indirect impact of foreign stock markets on a local markets... Sure, I guess. But that's hardly in the spirit here. Any large company in many countries has an impact on markets local markets when they sway, but rarely in any significant way.
The majority of people on HN are not impacted in any meaningful way by Nvidia's stock price and have no direct (brokerage, retirement fund, etc) investment in them.
The assessments in further up comments that give context here are that we all either invest in Nvida unknowingly or have our employment/compensation tied to them. I highly doubt this is true for the majority of HN users.
Nvidia is buyable by many international retirement plans, and/or affects international companies. It’s really hard to not be invested indirectly in them right now, even internationally unless your country has restrictions
tech is a large component of the overall economy though, so like it or not, we're all attached to this. I'm not a banker investing into mortgage-backed-securities either, but the 2008 GFC still affected me.
it really doesn’t matter, if nvidia pops hard it would almost certainly take the rest of the market with it, just on the basis of that being such a bearish leading signal on the market enthusiasm for AI. Nvidia popping would mean the AI party is over and a gut blow to the tech market as a whole.
this isn’t to say a down quarter that’s 25% down or whatever would be fatal. Obviously it’s going to have to correct a little at some point etc. Or if everyone else is obviously doing fine and it’s some nvidia-specific thing. But if Jensen got on the phone and said they were down 75% in next quarter orders on the basis of secular market trends and expected further cancellations, the tech market as a whole would probably crumple.
Most of the gpus nvidia sells aren’t DGX, they’re going into dell or HP or supermicro servers. Even the ones that aren’t officially partnered benefit from this ecosystem momentum etc. AI buildout has sold a lot of computers, a lot of memory, a lot of racks and networking and optics and electrical, etc.
As what is still functionally the only company making money on gpgpu sales right now, and with them functionally being near-100% datacenter and related (automotive etc) they are a massive leading signal on where the market is headed. The next best one is like… powercolor (TUL Corp).
(and nvidia conveniently does break it out anyway, but a lot of companies like AMD tend to submerge it in another group to deliberately make the revenue difficult to break out.)
The entire Tech sector was propped up Nvidia and GenAI since Nov 2022.
The bloodbath that would have ensued in the Tech sector (depressing salary and accelerating layoffs) would have been unfathomable.
If you think the $4 Trillion created in the Tech sector since Nov 2022 didn't affect non-AI people positively, you are naive about how the economy works
I think the naivity may be thinking the primary demographics of HN users are American.
It's been shown multiple times[0][1] that the majority of HN users are not American, certainly not 100% (as you assessed further up). It would also be reasonable to assume that not all American HN users work in a graphics card relevant space, or even tech.
Think of the dev ecosystem effects in all the various permutations, e.g. from Windows or Apple to AWS to more granular like PyTorch.
Those advantages aren't easily eliminated just by more competitors or even 'better' competitors.