Chip industry had worst sales year since dot-com bubble burst | Hacker News Reader

Chip industry had worst sales year since dot-com bubble burst

bloomberg.com

164 pointsbaybal2102 comments

>Revenue fell 12% to $412 billion in 2019, ....That’s the biggest drop since 2001, when industry sales slumped 32% as the dot-com bubble burst.

In 2001 post Dot Com Bubble, that 32% drop was real. The 12% drop in 2019 was only due to NAND and RAM went from Record High and back to "normal". Samsung was the most profitable company in 2018 simply because NAND and DRAM. Almost everyone in Semiconductor industry is making fairly decent amount of profits and increase unit shipment in 2019. Look at look at the leading Foundry ( TSMC ) and Intel all making record profits and shipment.

Bloomberg is quickly becoming BuzzFeed for financial news.

In fact, a significant amount of the 2001 drop was not "real". A lot of computer equipment sold in 2000 was "stuffing the pipe" in some way, such as selling on credit to companies that would never be able to pay, or selling into distributor inventory. While the equipment was never used, the chips inside were still paid for, so the chip companies had a great year. Lucent nearly collapsed as a result [0] and never recovered.

[0] https://www.technologyreview.com/s/403672/how-lucent-lost-it...

I know we're being trained to hate journalists these days, but that quoted sentence is just stating a fact: the biggest drop since dotcom bubble... and the article reads very reasonable. I don't see anything controversial in there - seems to balance the negatives with things like "the rate of decline abated" and "the decline in industry sales didn’t stop investors betting on a future rebound." - how is that sensationalist? Maybe they could have reported it as "19th best year since dot com bubble"?

As someone who watches the financial news, it almost feels more like baseball statistics, where they say something like, "this batter on sunny days in April has an advantage against left-handed pitchers."

While it may be true, it feels more like p-hacking than some kind of useful fact.

I felt like the financial industry was really doing this early in 2019, after the huge drop in stocks, everyone was like the S&P is up 20% this year already (forgetting to mention it was down a huge amount in December and ready for a comeback).

Of course how you frame these numbers and which numbers you pick is part of the reporting. When taking numbers of out of context, I think it can get a little sensationalist.

> As someone who watches the financial news, it almost feels more like baseball statistics...

If you hear a loud banging noise, don’t buy?

GP's point is that someone predicting a bang based on increasingly obscure and cherry-picked data does not in fact reliably predict a real bang in the future.

According to EMH, all information is subsumed into the price. As such, how could the huge 2018 Christmas drop affect future returns?

How could there be a huge temporary drop, with your view of the EMH? The fundamentals can't have changed that much. If they did change that much, and the market absorbed new information, then the price wouldn't have come back. The only way you can retain a belief that the price is "correct" is by imagining there are invisible error bars somewhat bigger than the size of the fluctuation.

Maybe EMH isn't correct?

Statistically it's not incorrect.

While you should be skeptical of strong forms of the EMH, all evidence points to weak forms of the EMH being true.

A corrolary that is good to keep in mind is "markets are more efficient than you're smart" -- prevents you from starting doomed financial enterprises

Weak forms of EMH are almost certainty correct. If they were not, you could beat passive investing with relatively little work, and that is impossible.

The article cannot be distilled to that single sentence. Framing the financial facts as a precursor to a recession by explicitly referencing the last tech bubble that burst carries with it implications that a skimmer may take away from the article if they weren't reading very critically. This is what FUD looks like in the wild. It's much more subtle and devious than the millions of Twitter bots spamming "sell $TSLA".

You don’t like Mike?

By "being trained" do you really mean that we are thinking more critically about the information fed to us?

We've always been trained to think critically about media (I even had classes on it in high school - one called "critical thinking"!). Seems like now it's just "hate journalists" not "think critically". If something doesn't fit your world view, just ad hominem it.

I'm hoping I'm just imagining it though (confirmation bias!), and it's really always been like this.

It might have something to do with social media making it more obvious that journalists and news agencies are usually partisan and agenda-focused rather than neutral and content-focused.

Political journalism (domestic US and international) has completely gone to shit. Tech journalism has slowly fizzled away into just clickbait and financial news has slowly drifted towards interpreting Ouija boards and magic 8-balls.

I think it's nostalgia for an age that never existed, multiplied by a US-based demagogue's campaign against facts.

Journalism has always been deeply flawed and massively biased, yet still useful. The people of the past knew that. You're not smarter than your great grandparents on that front. Tech wasn't some cure-all that opened your eyes but not theirs.

People knew these things.

In my opinion journalism has been destroyed by tech - the tech that drives journalists to optimize for selling online ads. The problem with what's written is not fundamentally political, nor is it the same old same old; it's the fallout of paperclip maximization.

Writing things that leave a false impression, use any and all techniques to inflame, and so on, are becoming the norm due to a runaway optimization process which is massively reshaping society in my lifetime.

Go read some old newspapers. It's clear you haven't.

What do you consider "old"?

100 years or more would be a reasonable start if we're going to have perspective.

No it’s not just nostalgia. The journalism business has totally changed. They once were funded by selling ads and now that business model is failing so they have been pivoting. They have to compete with the free content of blogs, Twitteratti, etc. Totally different now. We used to have serious journalists, they have vanished along with the demand.

So what are suggesting be done?

The “payload “ of this article might be the line that attributes the so called crash to a trade war that damaged free trade. Some partisans consider free trade to be sacred, others, evil.

Disclosure: I lean more towards “free trade is evil”, at least if you are labor, rather than capital. Michael Bloomberg might see it differently.

Edit: and I didn’t go into the article looking for a fight - I wanted to see news affecting an industry I work in, and the trade line jumped out. It’s probably even true, but it’s not hard to to imagine ulterior motives for mentioning it.

> We've always been trained to think critically about media

If that was true, media wouldn't exist because nobody would consume it.

> Seems like now it's just "hate journalists" not "think critically".

The criticism was about a bloomberg article, not a journalist.

> (I even had classes on it in high school - one called "critical thinking"!).

Do you really believe you learned to think critically in a government funded indoctrination center called high school? It's not what high school were designed for. It's certainly not what high schools are equipped for.

> I'm hoping I'm just imagining it though (confirmation bias!), and it's really always been like this.

It's always been like this. The media has always lied. People have always criticized it. Nothing has changed.

"Nothing can now be believed which is seen in a newspaper. Truth itself becomes suspicious by being put into that polluted vehicle. The real extent of this state of misinformation is known only to those who are in situations to confront facts within their knowledge with the lies of the day."

                                                                              -- Thomas Jefferson ( one of the people who gave us free press )

Well just reading the headline, I immediately thought that in terms of absolute sales numbers, they were as bad as since the dot com bubble burst...

Only to learn after that it's talking about a percentage drop?

The headline I'm seeing says "worst sales year", and I would expect that to be measured by the top-line revenue number, not in the y/y drop in revenue.

It was the third best year since the dot com bubble (also the third best year in the entire history of the industry), only worse than 2017 and 2018.

Bloomberg just dropped a pro-fossil-fuels FUD deuce on renewables by implying windmill rotor blades are clogging up landfills everywhere. I guess they can't conceive of multistage FRP recycling because they have to push their yellow journalism clickbait to make money rather than have integrity.

> Bloomberg is quickly becoming BuzzFeed for financial news.

"Bloomberg News Pays Reporters More If Their Stories Move Markets"

https://news.ycombinator.com/item?id=18162440

They've got a lot of authors. I invite you to read some Matt Levine.

Levine is good, but I don't think he'd write something like this. (I'm sure someone will drudge up an example though)

It's like saying Tesla stock had its largest ever price drop this week.

I wrote before that semiconductor industry economists were awaiting a crash for a very long time.

At least since 2016, I saw notion of a coming crash coming from very esteemed industry publications that cost few thousand bucks to access.

Everybody in semi knows of the cyclical nature of the industry, and the non-stop long bull run since 2008 looked very unusual to many.

Why is this so important?

The reason is that few years ago, the industry committed itself to the biggest capital infusion in history with EUV.

Those fabs costs tenths of billions, and take years to build. They are never ever constructed without an extremely thorough research done with top tier economists.

Well, some parts of the industry did. Some just opted out. Every generation of lithography sees a few companies say "forget it" and step off the treadmill, either moving to a fabless model (pay others to make your stuff), or consigning themselves to legacy parts only.

This has resulted in a long term concentration of semi manufacturing among fewer and fewer companies. Even Global Foundries (the company) decided to hold off on EUV.

> tenths of billions

Do you mean tens of billions? The difference is two orders of magnitude :)

I assume OP meant tenths of trillions.

> They are never ever constructed without an extremely thorough research done with top tier economists.

EUV is necessary to keep up with Moore's Law expectations.

Yes, as I understood, the industry people thought that it's now or never.

From one side, there was a long held reservation to poor money waiting for the cycle to restart, but the end of the cycle was not seemed coming despite everybody showing that the conditions for it were there.

I run this site, where we track public companies revenues, and Q3 2019 shows -8% drop in revenue for semiconductors segment, but it can be because Q3 2018 had +15% Y/Y: https://finintelligence.com/app/segment/2873229239

Repost from: https://news.ycombinator.com/item?id=20630279

Semiconductor industry is a very reliable canary for the long term economic situation. Semi industry employs top economists. It is critically important for them to have extremely thorough economics research when they decide on their financials. Fabs cost billions, and whole fab complexes close to $10B. They take many years to construct and make running.

Financing plans for them have to be done impeccably, and account for way more than capital forecasts.

The financial outlook for a fab changes dramatically depending on how the world changes during its construction. The few fragments of such report done for TSMC I saw included everything down to political, cultural, and technological black swan events.

One of the most bizarre one that they actually managed to predict was the current regulatory and sociocultural "dotcom backlash".

For example, they see big server chip buyers like Facebook, Amazon, Google greatly scaling down their ambitions, and therefore they did not include the option for big reticle sizes in their lithography equipment buy list.

I used to work in the semiconductor equipment industry in the 90s and it was quite cyclical. One year the demand was great and they were hiring like crazy and the next year things were looking dismal and they had to do layoffs. This repeated many times though the overall trend was of growth. I managed to stick around for some time since I was in R&D and with a bit of luck. Things smoothed out considerably over the decades.

Does this follow companies not buying new equipment annually, but wait a couple of years inbetween before the next upgrade?

I'm not sure why all companies would be on the same cycle of upgrading. You would assume they would all be offset and it wouldn't make a difference in sales.

Could be that Intel chip supply shortages, and decreased average sale prices (due to AMD offerings) could be the reason for this. Not as many people upgrading as inter generational gains start to level off.

I think embedded systems account for the vast majority of chip production, so any dynamics involving Intel/AMD would be very minor.

In units, perhaps, but not in $$, far from it.

Embedded chips are generally cheap. CPU’s sell for ~1,000x the price of most embedded chips. X00$ CPU vs X0c 32 bit microprocessors / microcontroller etc.

10c 32-bit microcontroller from Intel/AMD? Have I been buying the wrong chips? Care to share some links?

EDIT: commenter above edited their post, mentioned $.10 MCUs from Intel/AMD

I edited the comment for clarity, but never mentioned Intel or AMD.

All right, X0 c 32-bit microcontrollers from whom?

Cypress, Microchip, STM... Really, take your pick.

Inter generational gains leveled off 5-6 years ago.

I have an intel chip from 6 years ago and I still feel there is hardly any reason to upgrade. It still does everything I need it to do very well. I upgraded my gpu that was about 5 years old and I saw about 3x the performance.

The last few years have been mostly about performance per watt. I changed my notebook a couple of years ago, jumped from an i3 3200 to an i5 7200U. I got 80% better performance (on numerical code) at 15W, instead of 35W for the i3. Unfortunately manufacturers cheaped out on the battery, so I went from 6 cell to 3 cell, for about-ish the same battery life I had on my previous notebook.

Agreed, this news doesn't square with market performance. For instance, FSELX (Fidelity Select Semiconductors portfolio) mutual fund is up 43% over one year, doubling the performance of the S&P500. Top-10 holdings (68%) include Intel, Qualcomm, Broadcom, Nvidia, Micron, Marvell, ON, NXP, FLex, Commscope, so it's a fairly broad measure. For this cross-section to be performing so strongly, others must be circling the drain if indeed the industry is turning down steeply. I'm not seeing that.

In saying that there cannot be a turndown because stocks are rallying, you are not comparing apples to apples. Companies can circle the drain while their stock rallies. Companies can do good business while their stock gets crushed.

First, people are buying equities in general because interest rates are being held low artificially by the Fed, which has been pressured by the President to do this.

Second, few traders and investors understand how to differentiate semiconductor companies, so when it blows, it will all blow. People get out of crowded trades chaotically. Investors and traders, who are largely unsophisticated about the specific technologies that each company develops and markets, will punish the sector rather than choosing individual names to sell. I will say it again: most people who trade tech professionally have no fucking clue what these companies actually build. They look at numbers, watch earnings reports, and listen to buzzword-laden commentary from research analysts.

Look at Apple. Their supply chain is threatened by factory shutdowns and a huge market of theirs has basically been put on ice, but their stock is near all-time highs. They haven't made a game-changing product since the days of Steve Jobs, and people usually say their value now comes from their execution. But how can you execute if your suppliers are shut down and one of your major markets is closed for business?

INTC is facing tons of pressure in enterprise and PC from AMD, their 10nm has been a disaster for them, and yet their stock popped 9% on earnings due to cloud demand. It's trading near dot-com bubble levels. They are getting hurt in their competition vs TSMC on fab, and losing market share to AMD, yet their stock is like a rocket ship. Doesn't this seem weird?

QCOM has been hit hard by several regulators and in a number of lawsuits for its IP bullying, it has come out and said that coronavirus is going to hurt smartphone manufacturing and sales, and yet its stock is trading near all-time highs.

Don't ever confuse stock performance with company performance. Boeing stock is trading at more than double where it was when the 737 Max 8 took its first flight four years ago.

I'm honestly extremely pleased with my iPad Pro and Airpods Pro, both are honestly impressive pieces of tech. I'm coming from an EE background, so I might not be able to really distinguish nuances in software, but hardware-wise they are impressive.

I can't agree with you on the Apple perspective, and it's my subjective opinion, I might be also biased since I have a large position on APPL. However, I can agree on the idea that the stock price can be completely disconnected from the fundamentals of the company.

Apple’s most recent earnings call seems to suggest they’re doing just fine expanding into wearable accessories and monthly subscriptions as well.

Seems like half of NYC has airpods when I’m walking around and in most of the stores here they won’t be in stock until March (a jump from mid February a couple of weeks ago).

That could be supply chain disruption or they’re selling like hot cakes.

Very satisfied with my pro pod pair post purchase as well.

Everything you say is true, but traders are only reacting to the money supply. The FED has been cranking dollars at a level that in unheard of since the depths of the Great Recession. It seems that they're committed to reelect the person in the HW at any cost. This money has to go somewhere, and for traders so much better if this all comes crashing later, there's more money to be made on the downside.

"Doesn't this seem weird?"

I have no idea if INTC is a good stock to buy, but it sounds like you have an idea of who they are that is based on a lot of ordinary people and their stereotypes about the company from the media and from using computers for years.

A company can be reshaped much faster internally than their public image, but I don't think you have to be an insider, a genius, or a professional to figure it out. You just have to read their reports to the SEC. Not for the decimal places, but for one or two significant figures and the text.

What I'm saying is, I don't know if x86 CPUs have the significance you think or not. If people did buy the stock based on "cloud demand", then I would say "huh, I wonder if that's justified, what proportion of their business is it and how is it growing?" Rather than assuming they are who I think they are. IBM, GE, many other companies have strong images of the companies they used to be with the public, so I imagine many people own stock without any idea of what they really have.

I can’t agree with you there. I would say most professionals trading these stocks in volume know more about tech than what’s being discussed here. I speak from my own experience but Wall Street has changed quite a bit in the past two decades and domain knowledge is what gets you paid. These guys also have access to c-suites to continue to build on domain knowledge.

Haven't made a game changing device since Steve Jobs? The Apple Watch completely dominates that product category. There is no alternative that can even compare.

The entirety of "wearable, home, and accessories" amounted to 10.9% of revenue for the past quarter, which included Christmas, and 9.4% for all of fiscal 2019. In addition to the Apple Watch, that category includes Airpods, HomePod, Apple TV, Beats, iPod touch...

Just because you like the Apple Watch doesn't mean that it has changed the way the company does business. And the fact that Apple doesn't break out gross margin for this category tells us that they aren't particularly proud of that number. Gross margin on products as a whole is about 34%.

Apple wants to paint the picture that they are diversifying away from phones, which accounted for over 3/5 of their revenue this past quarter, and shifting toward services and other products.

But it's one thing to repeat a rosy narrative, and another entirely to back it up with financial statements. Apple is still a phone company, and if the Apple Watch really drove profits for them, they'd break "wearables" out to its own line item, or at least show us the gross margin for "wearable, home, and accessories."

https://s2.q4cdn.com/470004039/files/doc_financials/2020/q1/... https://s2.q4cdn.com/470004039/files/doc_financials/2019/ar/...

The Apple Watch by itself is estimated to be larger than the iPod was at its peak. (http://www.asymco.com/2019/12/12/ipods-pro/) and the Airpods if they are not already larger than the iPod at its peak soon will be.

As far as the iPod Touch, the entire iPod line was less than 1.5 million a quarter when they stopped breaking out the numbers (years before they stopped reporting volumes of their other lines).

The iPhone when it was introduced was already introduced into a market with a 1 billion device/year run rate. Now close to 5 billion people own cell phones (https://www.bankmycell.com/blog/how-many-phones-are-in-the-w...).

Also, 61% of their revenue comes from phones. (https://sixcolors.com/images/content/2020/financials-2020-1-...)

And of course the iPhone is going to sell well during the Christmas quarter and after they introduce new phones compared to most quarters.

https://sixcolors.com/post/2019/07/apple-third-quarter-2019-...

But the last time I checked, even a category that is only 10% of Apple's revenue still puts that category's revenue above all but the top 100 companies in the US.

And that wouldn't be the first time semi stocks do well right before a crash. In 2000 they did fabulously, but at the expenses of the largest number of unprofitable tech companies ever. According to some, we're following the same path.

No one can innovate when they can’t understand. No one can understand when they can’t read the data sheets for chips they don’t use.

Want people to build things with your crap? Don’t fight them when they try.

This absolutely. The chip people always make custom protocols for /everything/.

Custom protocols are not really the end of the world, it’s when they make a custom protocol and then use copyright to force you to do certain things in order to use it.

Are you talking about NVIDIA and their Cuda stack or some another manufacturer?

They really do slow down development though, and flrxe you to hire specialists who can make sense of the often terse datasheets.

Wouldn't a lot of the complexity be due to the fact that these are likely the most complex pieces of engineering on earth?

Complexity and nicely written datasheets /doc are tractable. Obfuscated docs and average complexity is cryptic.

Thanks for the post, tipped me off to the current rock-bottom ram prices to max out the ram of my desktop I built 5 years ago.

There is interesting news about Xilinx specifically: they are no longer allowed to sell the Huawei, so their stock declined recently. The consequence is that Huawei needs to switch to home-grown ASICs, and I'm sure the pressure is on for a high-end home-grown FPGA vendor.

So this is just one company, but I'm wondering if tariffs in general are a part of this decline.

https://epsnews.com/2019/10/28/xilinx-cuts-huawei-from-futur...

https://www.nytimes.com/2020/01/20/business/economy/trump-us...

"hurt by a trade war between the largest chip producer, the U.S., and the largest consumer, China"

This sentence confuses me. Seems like they have it backwards. Can someone clear this up?

It depends. Do you include Taiwan in China or not? If not, then yes.

In fact, U.S. is the largest manufacturer of semiconductors. In 2018, nearly half of the total semiconductor market was occupied by firms based in the U.S., according to the Factbook2019 of SIA.

Here's the historical total graph, not the year to year change graph.[1] Unclear if this is constant dollars. That's from the World Semiconductor Trade Association, which collects statistics and resells them, expensively. The Bloomberg article is what you get from the free tier, press releases.

[1] https://www.wsts.org/

Coincidentally, I've been wondering about dumping some cash into a semiconductor ETF like eg the VanEck offering (SMH). I noticed that there are leveraged ETFs like SOXL that magnify returns and I typically associate these riskier offering with precious metals and so forth. I hadn't realised that semiconductors seem to fall into that same "cyclical commodity" category.

Question: why then are there recent media reports that DDR4 and SSD prices going to be increased? Is this Samsung's unilateral oligopolic move, another price fixing event, a ploy to stimulate panic buying and/or some other structural move related to real ability to supply? (Yes, I remember the hyper-commodification DRAM price wars and price fixing of years gone by.)

Uhhhh I think it is because everyone is waiting for the next gen consoles of Xbox and PlayStation. There are even rumors for a Nintendo switch "x"

And they all need chips.. And electric vehicles are also slowly catching up. They need chips too.

Its not a bubble burst, just a generation change (and wait)

This must be mostly attributable to the massive NAND and DRAM prices falls in 2019.

If only the US had the infrastructure to do everything from mining the materials all the way up the chain to building custom PC’s, wouldn’t that be cool? If we could do everything “in house” so to speak?

Seems like it should be a priority for homeland security really...

Not sure about you guys but I would be willing to pay a premium if I knew that every part in my computer was created right here in my own country.

Pretty nasty pressure move to strangle your rival's access to chips. New foundries can't be built over night.

https://archive.md/3Ms5l

This article sounds alarmist. "Revenue fell 12% to $412 billion in 2019, the Semiconductor Industry Association said Monday in a statement. That’s the biggest drop since 2001, when industry sales slumped 32% as the dot-com bubble burst."

I don't see how 12% is a skyfall number. Especially when compared to 32%.

It was pay-walled so I did not read any further.

> It was pay-walled...

For others who hit the wall. https://outline.com/VYJEa5

We can argue about adjective choice, but a 12% revenue drop in any industry is going to get people's attention.

The environment in which that 32% drop happened involved thousands of companies failing within a short time span. I personally ended up leaving the Bay Area after being out of work for over a year. (I came back.) Just to offer some context to hang those numbers on.

I don't see that happening at all currently. The housing market is insane in the bay area, as usual, even out here in the tri-valley area. as far as livermore, houses are selling like hot-cakes, not even on the market for a week.

Didn't intend to imply we're on the edge of a tech-recession or anything close to that.

But I'm pretty sure property prices are not a leading indicator of tech crashes. In the first dot.com crash they didn't drop until after the job market carnage was well over.

https://www.bayareamarketreports.com/trend/3-recessions-2-bu...

Revenue falling at all in any tech industry is usually cause for extreme concern, as all the investor numbers assume infinite growth.

"as all the investor numbers assume infinite growth"

This is absolutely not the case in semis.

Some of these companies had near monopolies for long stretches of time. If some of that is busted up through competition it could drop revenue while still being healthy. If you're invested broadly in the whole sector you might lose, but if you're invested in the successful challengers you win.

If actual demand was lowered rather than just supply changing due to effective competition, then it would be a bigger concern.

Perhaps that’s the real problem there.

If companies like Intel and AMD are still turning a respectable profit, there’s no cause for actual panic, yet.

Investors should learn that intractable P/Es on established companies are, at best, unwise in all but the edge cases.

Let's stop discussing a hypothetical. Analysts in this space are well aware of the cyclical nature of it, and have better data than some of the companies in question. The premise is false.

Cyclical trough/peak earnings normalization is decades old from industrial and resource sectors.

P/Es aren't the best way to trade companies like AMD where past earnings are not indicative of future earnings.

Turn off the javascript on the page.

using "reader view" in my browser and then refreshing works for me.

I did not think that would work- why even bother with the darn paywall! I just guessed that if you disabled js the page would fail to load like other news sites.

Anyhow, thank you!

hmmm every time I'm at the store I see like 3 new brands of chips I've never seen before.

Reply on news.ycombinator.com