Great news for another new $1TN dollar company, however in [0]
"Market cap stories (and price fluctuation stories generally) mostly aren't substantive enough to make interesting HN threads. They just turn into generic discussions about the underlying company or currency or whatever, and those are too repetitive to be interesting."
This one is no exception, with such threads should also apply to this one. Just like it did with Tesla and Facebook about them reaching $1TN. Right?
To put this into context with other large tech companies:
P/E MARKET CAP
Salesforce 1,036x $0.2T
AMD 519x $0.2T
NVIDIA 213x $1.0T
Amazon 293x $1.25T
Microsoft 35x $2.4T
Meta/FB 33x $0.7T
Apple 30x $2.7T
Google 27x $1.5T
TSMC 16x $0.4T
Samsung 10x $0.3T
EDIT: "P/E" ratio is the Market Cap "Price" / Earnings the company generates. E.g. Samsung is generating $30B in earnings (not revenue, earnings), investors are valuing Samsung to be a $300B company. That means investors see Samsung is worth 10x P/E
It can be a variety of real reasons combined with the swings of exuberance and pessimism that investors have about various stocks.
On the real front, the tech war with China is intensifying a bit. China doesn't want to be locked out of chipmaking and the US is doing that via ASML. When a company is having profits, one needs to ask whether those profits will continue and for how long. Will China invade Taiwan this year? Unlikely. However, there certainly is a risk now and in the future depending on how things go with US/China relations.
Likewise, TSMC's rise has come during an era when they've out-fab'd everyone else. However, if you go back to 2013, that wasn't really the case. Samsung, GloFlo, SMIC, and Intel arguably had slight leads. It's really been the past 5-7 years that TSMC has really taken off. This could change. Intel's roadmap over the next few years possibly looks better than TSMC's with their 20A process supposedly coming to market before TSMC's 2nm. With Intel also offering to fab chips for others, that could potentially steal some business from TSMC.
Are these likely to happen? I don't know. It could just be market pessimism. Still, there are real risks. TSMC started taking off when Apple decided to commit to using them as their only fab. Apple is known to make long-term commitments to companies so that they can invest with confidence. If Apple makes a new long-term deal with Intel in 2024 so that they can get 20A instead of waiting until 2025 for TSMC's 2nm, will that rob TSMC of their strategic planning advantage? Other fabs didn't have the commitments that Apple brings to the table. If TSMC loses that, how much does that impact their progress? It's easy to invest in the latest stuff when you have commitments. It's harder to go with "if you build it, they will come." TSMC has done a great job, but I have to think part of that is aided by Apple's commitments.
At the same time, TSMC's growth has come during a period when Intel was asleep at the wheel. It looks like that period is ending and Intel is rejuvenated toward their fab business. Even if Intel doesn't surpass TSMC, Intel being equal or barely behind will make for a very different market. We know that Qualcomm and AMD want to replace TSMC. They tried with Samsung and were disappointed at the yields. Maybe Intel will become an alternative (at least for Qualcomm). As Intel's fab gets better, it might dampen demand for ARM server chips and AMD chips in general. AMD has been doing well in part because it had a significant fab advantage via TSMC. If that goes away, AMD processors might go back to being second-string. Likewise, a decent amount of ARM datacenter demand is due to the cost and performance/watt advantage that comes from being able to use TSMC's better fab. If that goes away, do we see ARM datacenter adoption slow?
A lot of people treat TSMC as if their market position is unassailable. I don't think TSMC is going anywhere. I think they're in a decent position. However, if Intel actually pulls off their roadmap, it seems like things could be a bit rough. We already saw Apple forego TSMC's 4nm chips for their non-Pro iPhones last year. We've heard rumors of supply constraints on 3nm parts for this year's iPhone and we still haven't seen a 3nm MacBook. Intel was just at Computex showing off its Meteor Lake processor using Intel 4. When Apple launched their 5nm M1 MacBooks, Intel was making the transition from 14nm to 10nm.
Again, I don't want to make this sound too doom and gloom about TSMC. I just want to note that TSMC has significant risk and that risk has been increasing - both from competitors in the industry shifting and from the shifting political climate and rivalry between the US and China. Will those things come to pass? Who knows. Intel might start slipping on its roadmap like it did for years. China might decide that it's fine with the US cutting them off from EUV/ASML and that it has better things to do. But there is risk.
The "E" in "PE" is earnings. It all boils down to profit margin. And lately, their input costs have been fairly high, what with COVID supply chain disruption.
Part of what's let NVIDIA perform a bit better (if we're chalking its market cap up to performance rather than hype -- the truth is likely somewhere in between) is that they actually have a bit more of their business tied to software these days, which doesn't typically suffer from raised input costs. Sell more software, and your profits rise a LOT faster, because 100 licenses worth of software has a similar input cost as 1 license.
Essentially, in Enterprise Sales - you have huge upfront costs to win a deal (e.g. paying commissions, pre-sales expenses like T&E, marketing, etc).
And in SaaS, you only collect a small monthly fee (as opposed to On-Prem when you collected a large upfront fee all at once).
So it becomes a cashflow / revenue recognition issue for these SaaS companies and you might not even be profitable until 2-years into the life of the customer contract all due to this cashflow issue.
This gets compounded if you're having high growth in New sales (you go further into debt).
To this point, when the market crashed in 2008 and sales slowed - that was the year that Salesforce experienced higher than normal profits.
GAAP earnings basically consider stock based comp as an operational expense and Salesforce does a lot of stock based comp which makes their GAAP earning numbers very low.
Because of this historically Salesforce has traded more reliably as a multiple of its cashflows and currently it trades at around 26x cashflows which is about right if you comp with similar companies. The argument here being that so long as they can grow their cashflows faster than they're diluting them via stock based comp then GAAP earnings aren't all that important since investor's per share cashflows are increasing. Some disagree with this though.
My opinion on this is that all earnings should be taken with a grain of salt since both GAAP and non-GAAP earnings are arguably sanitised accounting numbers and not necessarily a reflection of actual financial results.
> GAAP earnings basically consider stock based comp as an operational expense and Salesforce does a lot of stock based comp which makes their GAAP earning numbers very low.
Compensation is obviously an operational expense...
TSMC seems real low for a company with a 100% share of the high performance chips market, with a really high moat, and lots of hype that chipmaking will be getting lots of government investment and lots of demand with AI wanting massive amounts more compute...
I think the big thing in this space was Buffet recently pulling out of TSMC saying he loved the company but thought the tensions made it too risky. I don't know what that puts the odds for a full on invasion of Taiwan at but tensions in the area are definitely impacting real investment decisions.
Based on what? This sounds like you just pulling numbers out of the air, in which case 1% vs. 6.5% means comparatively nothing. And what makes you so sure that if China invades Taiwan they would definitely do it by 2033, and there's a 0% chance it'd happen in 2034?
I'm not trying to be [overly] pedantic but HN is littered with comments basically just making shit up in a language of confidence and precision.
I believe Xi Jinping has said something to the effect of "Taiwan should be a part of China by the end of this decade," although I'm having a hard time pinning down the exact quote. Although it should also be noted that Xi also tends to emphasize his desire to see Taiwan unified peacefully rather than force, so it's difficult to assess with accuracy how likely China is to invade Taiwan to compel an unwilling unification.
As you say, 1% versus 6.5% is a level of precision that isn't helpful. The way I prefer to think of it is that a Chinese invasion of Taiwan is sufficiently probable as to warrant making contingency plans for such an event, but not so probable as to warrant making active efforts to avoid Taiwan.
If you believe that Taiwan has a 1% chance of being invaded per year and that this event would remove your ownership then the long term value of purchasing this stock is negative
The image of some western white guy complaining about unfair it is the China invalidates his ownership in the TSMC following a genocidal invasion of Taiwan is pretty good, I’ll give you that.
TSMC is at risk due to geopolitical tensions, particularly between China and the USA. These risks led to Warren Buffett's Berkshire Hathaway selling its stake in TSMC. Investors are worried about that. That s my best guess. Apple also opened three factories in India and other places to mitigate this risk.
Yes and no. If NVIDIA loses access to TSMC due to geopolitical events, presumedly their competitors do too. There might be a couple years where they face increased competition from their own used market if they're unable to produce chis competitive with the previous generations', but at some point Samsung et. al. will catch up on capability and capacity, and NVIDIA will be as well positioned relative to their competitors to take advantage of that as they are today. The only case where this would be significantly different is if one of NVIDIA's major competitors was independent of TSMC and thus could use the lean times for NVIDIA to leapfrog them; but that would require considering Intel a real competitor.
Arguably in this sense their only competitors would be screwed even harder. The last non-TSMC GPU that AMD has released is based on GloFo 12nm/14nm, NVIDIA would actually be ahead with Samsung 8nm. Intel's only dGPUs are TSMC as well, they flatly don't have a product without TSMC.
I think Intel's advantage in this hypothetical is access to foundry space. Porting their current CPU designs from TSMC to their own foundry is non-trivial (although it can't be that hard -- their integrated graphics are fabbed on their own process, and share quite a bit with their discrete units), but at least having completed that effort they have somewhere to go, instead of fighting over what will be extremely over-subscribed Samsung fabs.
NVIDIA also depends on not getting the Micron treatment from China. Or even not having China take a particularly hard line on their Micron policy and disallowing NVIDIA to import Micron chips to use in their manufacturing produce for export from China.
It seems to me that they don't use their moat to really drive price. I don't know if it's cultural or what, but they could easily extract higher prices given their position. I seriously doubt Apple, AMD or NVDA would walk to another fab and give up a huge performance edge in their offerings.
While TSMC's gross margins are 60% which is already quite high, their customers also have close to 50% margins... which implies room for higher chip prices
TSMC is capex heavy. They will see margin compression with higher interest rates, at least in perception - their most over intel is only their ability to efficiently deploy capital.
This. The value of a stock should (in theory) be based on discounted cashflow with the emphasis on cash. A company can generate massive earnings but if it needs to invest huge amounts of capital to support those earnings then it should have a lower valuation.
Isn't being CapEx heavy a sign of constant expansion, though?
I would think that selling (or choosing not to buy) a stock because their P/E is low because their CapEx is high is short-sighted unless you don't think their CapEx spending is going to pay off.
It's like...back in 2012, I was talking to one of my wife's relatives at a Christmas party. He said he'd never buy Amazon stock because they've never been profitable. I said they're not profitable because every dollar they make, they put back into R&D. He would have 10x his money now if he had bought their stock.
Software R&D is a special type of capex, in principal - successful R&D execution buys recurring free cash flow in near perpetuity. Contrast this with TSMC who needs to outlay ~50 Billion every 2 years on a new Fab which will deprecate over ~6-7 years. If TSMC stopped buying fabs, then intel would beat them in ~2 years time - and TSMC would be out of business in 6-7 years.
Software R&D may be less efficient then the above math would imply, but there is a big range between perpetuity and some depreciating time horizon. R&D can also boosts growth prospects by expanding markets.
CapEx intensity can boil down heavily to the industry a company is in. If a software company is doing a lot of CapEx, it's likely growing pretty quickly. If a manufacturing company is spending a similar ratio of its earnings, it may well just be treading water.
As usual with these ratio metrics, a full understanding of the industry, or at least understanding that you can't compare apples to oranges, is important.
Price to sales is a more telling comparison here. P/E can appear arbitrarily high/low depending on what the company is optimizing for. A realistic "terminal margin" is pretty similar between like-kind businesses, so you can somewhat assume what it will be in the long run and apply that to projected revenues.
FCF is not great for tech companies because SBC (stock based compensation), of which these companies do a huge amount of, tends to be backed out.
When a tech employee receives shares, it dilutes other shareholders. This is not "free" money to be ignored just because it doesn't take cash to pay it
I don't either. Selim wasn't kind enough to loop us in so I looked it up and I think it means "naked person." I have to say, if that's accurate, I like it. It has an "emperor has no clothes" vibe that I enjoy.
IMO, NVDA is in a similar position to VMware among practitioners--folks hate their pricing and will gladly jump ship to whatever is cheaper with feature parity. They've burned their brand by charging so much and doing licensing fuckery to squeeze as much as possible. Folks are only staying because CUDA has no equivalent (for now).
> folks hate their pricing and will gladly jump ship to whatever is cheaper with feature parity. ... Folks are only staying because CUDA has no equivalent (for now).
ah yes, consistent long-term delivery of realized customer value, the shakiest of moats
When most folks are using CUDA indirectly via library abstractions, swapping it out eventually becomes easy and their moat fades away. CUDA alternatives are innovating pretty fast, and the ever present threat of TPUs and things like Apple's Neural Engine aren't helping them. They should've been building better partnerships and giving more margin away for integration and OEM instead of being greedy. But that wouldn't have juiced their short term profits.
Maybe someone can explain. Why hasn’t AMD come out invested $10 billion, poached the necessary engineers from Nvidia, Facebook etc and expanded into the AI space? How much of an advantage does nvidia have when it comes to cuda ?
cuda is the main advantage nvidia has over amd in this space. amd gpus are close enough to nvidia gpus these days that one should be able to buy either. except for the software support and community. so it makes it 10x harder to get something ML started on an amd GPU. so everyone ahs to buy nvidia
I am not an expert in this space but dabble a bit. This comment is spot on. AMD's software is pretty bad and NVDA has captured all the developer attention with CUDA and is the basis for lot of the frameworks people use. It is not a trivial advantage to break. I do wish good luck to AMD.
Speaking of trillion dollar companies, how is Apple the most valuable one? Their devices are dumb as bricks. Apple is do bad at AI that they can't even get autocorrect right on the iPhone. They are completely out of their element in the age of LLM's. Microsoft's business was bulletproof before investing in OpenAI but now it should be more valuable than Apple.
Pretty straightforward. The current value of the company is the Net Present Value of all the future expected cash flows. Basically you can take the money Apple will make in 2024, 2025, 2026 ... and reduce them to today's values by discounting with the interest rates: pretty much divide by (1+r)^n. Since Apple already makes a ton of money each year it is valued pretty high. Nvidia on the other hand has to GROW its earnings a LOT to justify its valuations.
Does your apparent Apple hate blind you to how many devices they are selling at higher margins than other similar companies? Not to mention their huge expansion into services.
Why does Apple have to be good at AI? Why do all of these big tech companies have to have a play for everything?
It makes me a bit sad to hear NVidia is that successful. I'm working with their Jetson SoM product line professionally and was suffering from their bad support on several occasions. I hoped that their bad support would translate into them falling back.
Based on P/E, I don't think so. There's already has a LOT of expected growth priced in.
We're in an AI bubble. AI has a lot of incredible uses, but eventually the bubble will burst and the market will recognize places where AI doesn't really belong.
I wish I had looked at them earlier. PE ratio is still low, but I'm not sure they'll capitalize on the AI trend... ...and it's so hard to buy shares that are up 10x! What do you think?
I've sold my stock, so it will obviously keep going higher. That being said, I don't see how it's a sustainable price, it has massive P/E and Price to Sales ratio. In the past when a stock finally hits the $1 or $2 trillion mark the stock will give it up relatively quickly before it finally reclaims it for good. AI is the new "blockchain" in terms of the level of hype it is getting where every company now can mention it and get a 5% jump in their stock price, to me that means we're getting closer to a saturation in the hype.
That is obvious, I just mean that many companies that don't have, and won't have, any AI competencies are now mentioning AI in their earnings calls. I saw someone had done a chart of earnings calls that mention AI and it has skyrocketed. Also a new trend of stuffing GPT and LLMs into apps that don't need it is already in the pipeline, people comment it on HN all the time.
Sounds like a great opportunity for shorting a stock.
They announce doing something with AI, their stock skyrockets, the new AI features don't noticeably increase revenue while their expenses go up considerably, stock drops. Profit.
It may continue up to 1.25T or so, but the current price is absurdly overvalued at 37x sales. Even with the stellar guidance they gave, at best, the company is worth 0.5T. It may not go down that much, but folks buying here for the long term are in for disappointment.
I did a rough calculation and came up with a current value of $80bn for Nvidia's offer for Arm based on the current Nvidia share price. An expensive regulatory loss for Masa given Arm will probably float for a lot less.
Of course had the takeover gone ahead then its likely that Nvidia would be even more valuable - unless you take the view that Arm would have been a distraction for Nvidia.
Total Real Returns Chart for NVDA: https://totalrealreturns.com/s/NVDA (with dividends reinvested and inflation adjusted). It's been a bumpy ride since 1999! Trendline is quite positive but you'd have to have some real conviction to hang on with that volatility.
Isn't avoiding losing any money the best case scenario for people who buy now?
Even for people within the company, the pie is not getting any bigger so the only game in town is to become the best skilled politician to try and increase one's own slice.
Just a reminder that various AI algorithms have significant influence over markets to the point that it has by proxy fundamentally changed world culture in significant ways.
Everyone looking at the P/E ratio doesn't realize that this isn't simply a speculative bubble. Rather, this is a collective superintelligence bootstrapping itself, the market will just become another data pipeline at that point.
reply