1/ Lessons From The Tech Bubble:
Last year, I spent my winter holiday reading hundreds of pages of equity research from the 1999/2000 era, to try to understand what it was like investing during the bubble
A few people recently asked me for my takeaways. Here they are -
1/ I just finished a 2.5 week trip through China today, my first visit in about a decade. I was there for family reasons, but it also happened to be my first time in the country as a tech industry observer
My amateur travel journal on the China tech market -
I guess GPT-3 is old news, but playing with OpenAIโs new chatbot is mindblowing.
Weโre witnessing the death of the college essay in realtime. Hereโs the response to a prompt from one of my 200-level history classes at Amherst
Solid A- work in 10 seconds
The problem with โUber for helicoptersโ is safety doesnโt scale
When I was at Bernstein, our aerospace analyst ran the numbers - unless you massively improve upon helicopter accident rates, a fleet of 100 air taxis operating 5 hours / day should incur 5-10 crashes per year ๐ฌ
๐ SF BAY AREA NETWORK UNVEIL | Weโre excited to announce Archerโs planned air taxi network in the San Francisco Bay Area. This goal is to connect 5 key locations across the region: South San Francisco, Napa, San Jose, Oakland, and Livermore, replacing long drives with quick
9/ What's the takeaway here?
Be humble.
For bears, it's easy to call a bubble. Anybody can do that. Timing is the hard part
For bulls, it's easy to point to the fundamentals. Historical investors weren't dumb. The hard part is matching fundamentals with price...
1/ Just caught up with a few investor friends in the consumer space last week about Ozempic and GLP1s
As far as I can tell, everything basically hinges on: how much does it matter that every consumer product in the world depends on a tiny cohort of super consumers?
8/ LESSON
#6
: Fundamentals follow price, not vice versa
The bubble popped in Q1 2000. Fundamentals didn't decelerate until Q4 2000.
It was reflexivity at work. Lower stock prices = less capex spend = less revenue growth = lower stock prices. A vicious cycle
6/ LESSON
#4
: "Tech" bubble was a misnomer... it was really a large cap growth bubble
See the valuation table below, 1 year before the top
Yes, Microsoft traded at 70x earnings. But Coca Cola was 43x. Pfizer was 92x. Every stock here was a disaster over the next 10 years...
From 1840 to 1850, private Britons cumulatively invested 40% of British GDP into the countryโs first rail network
Does anyone else find it weird how nobody ever writes about this? This would be the equivalent of US VCs spending like, $10 trillion dollars on a single thing today
17/ IMO, the fact information on the Chinese internet is scattered across myriad walled gardens without a reliable means of search seems to beโฆ pretty clearly bad for consumers?
This feels like a key area where China tech has actually regressed rather than leapfrogged the US
3/ LESSON
#1
: Everybody knew it was a bubble
Unfortunately, the quip "it's not a bubble if everyone says it is" just isn't true
Investors were comparing the internet sector to tulip mania as early as mid-98. Bernstein held an entire conference on it in June 99!
7/ LESSON
#5
: Most large cap tech stocks in the bubble had real businesses with strong fundamentals
The internet stocks were a sideshow. In 2000, the software sector had a $1 trillion market cap, 20% net margins, 20% annual growth
The problem? It was trading at 16x sales
27/ In summary:
- 1-2 superapps tying together the digital and physical worlds
- Hundreds of EV brands
- Walled gardens crippling the open webโฆ largely to the detriment of the consumer
Itโs not often you get to see a counterfactual history of technology play out in realtime
15/ Why Douyin instead of Baidu ("the Google of China")?
One of the big things I learned on my trip is how god-awful Baidu is compared to Google - way more ads, way worse relevance, and all of the websites are out of date
The decline of the Chinese web has crippled them
4/ LESSON
#2
: Calling bubbles is easy, making money is hard
In truth, the hard part about the tech bubble wasn't noticing it. The hard part was timing it
Our equity strategist tried in January 99... he was off by 14 months (and another 30 point gap in value vs growth)
5/ LESSON
#3
: Nobody knew the bubble popped until months after it did
Nobody noticed in March 2000 when it finally popped. Our equity strategist (who bet his career on it!) didn't catch on until June
14/ OBSERVATION
#4
- An internet without the webโฆ kind of sucks?
My aunt (same one) was trying to update her city government ID. In the US you'd just Google it. In Chinaโฆ I saw her search Douyin (the Chinese TikTok) to look for a local influencer to explain it via video
11/ OBSERVATION
#3
- What does the internet look like without the web? Apparently, China
I don't think I saw a single person open a browser at any point on my trip. For my extended family, the "internet" is WeChat + maybe 5-10 smartphone apps (nobody uses PCs outside of work)
2/ OBSERVATION
#1
- Yes, everything really does run on WeChat
If you're a foreign traveler visiting China, you really must set up WeChat Pay and Alipay beforehand. For me, this was the Chinese equivalent of Whatsapp + Chrome + Venmo + my credit card + my subway card + Doordash
7/ How many American industries have been propped up for the last 50 years by the poor impulse control of a small cohort of people with eating problems, drinking problems, and gambling problems?
Guess weโll find out in a year
1/ IMO, the most interesting thing about the ongoing blowup of the quick commerce sector - Gopuff, Getir, Gorillas, JOKR, etc. - is that pretty much all of this already happened 20 years ago with Webvan, the online grocery darling of the first tech bubble
A few quick thoughts...
1/ In September 1993, then-Microsoft exec Nathan Myhrvold wrote his landmark memo "Road Kill on the Information Highway", laying out a dozen-ish predictions on the rise of the internet
27 years later, I think it's a super interesting case study. Let's evaluate the predictions -
6/ OBSERVATION
#2
- The future of automobiles is happening in China today
I frankly don't spend a lot of time thinking about cars, but ten minutes on the street in China made it obvious this was the world's most dynamic car market
12/ In the US, we default to finding info on Google, which crawls the open web
In China, the open web basically doesn't exist - instead, every major app has its own walled garden of content. Douyin has its own garden, WeChat has its own garden, Xiaohongshu has its own gardenโฆ
4/ Something especially surprising to me: you order via WeChat even when you're dining in a restaurant. This was true both at fast foods (where the cash register was usually closed) and dine-in (where waiters now only bring you food instead of taking orders)
3/ I didn't use cash a single time on my 19 day trip. Everybody took WeChat Pay, from Michelin-starred restaurants, to McDonalds, to butchers at the farmers markets in tier 3 cities, to performing musicians in national parks
9/ To me, it feels very obvious the Chinese EV market is in a bubble- no different from the hundreds of American automakers founded after the invention of the internal combustion engine
I wouldn't be surprised if 80% of these brands cease to exist by the next time I'm in China
16/ I witnessed this firsthand trying to book tickets for a local history museum. Baidu was no use; the museum website hadn't been updated in years. There was an online reservation system only accessible through WeChat. How did I learn this? Watching a video on Xiaohongshu
1/ The more time I spend in the corporate world, the more I understand why everybody just hires ex-consultants and investment bankers
Itโs not because McKinsey or Goldman Sachs actually teach you how to do the job, per se
Itโs because hiring undergrads is a free rider problem
8/ What was most impressive was the sheer number of domestic EV brands. The blue plates I saw were still 50/50 Western vs. Chinese cars. But the green plates were basically 100% either A) Teslas or B) a dozen crazy Chinese brands I'd never heard of
1/ There's a well known anecdote from The Everything Store about Jeff Bezos visiting Harvard Business School in 1997. The students told Bezos "you really need to sell to Barnes and Noble and get out now"
Less well known: you can read the actual 1997 HBS case study from the class
1/ I was rereading Softwar (the 2004 book on Larry Ellison and Oracle) this morning, and the main thing that stood out to me is that pretty much every idea in software today was already basically around 20 years ago
Wait, Amazon just revealed that they finished training a 200B parameter LLMโฆ and nobody noticed
This is from SVP James Hamiltonโs Jan 15 talk at CIDR 2024. Trained on 5x more compute than Facebookโs Llama 2. I literally havenโt seen anyone write about this?
5/ Given low labor costs, I doubt this pattern of "dine-in on WeChat" is driven by economics? Seems like consumer preference
I'm already seeing more restaurants in Chicago adopt "pay on your phone" a la Toast - seems like the next logical step given the higher cost of labor
@SethLiebert
So thatโs the thing - theyโre significantly worse than cars!
Helicopters have historically been more dangerous than private planes, and private planes already have a 10x higher fatality rate per mile than cars in the US
1/ A Google colleague recently observed to me that computer science tends to reinvent the wheel every 20 years: โa new generation just reimplements old ideas, but with more computeโ
Whatโs fascinating is this 20 year cycle seems to coincide with the timing of tech stock bubbles
7/ You can measure electric vehicle adoption in any Chinese city by counting the license plates - green is electric, blue is gas
Beijing and Shanghai looked 30-50% electric (no surprise given the plate restrictions). But even tier 3 cities like Tangshan or Zhangjiajie were 10%+
10/ The heads up displays + infotainment systems on these local EVs are *much* more sophisticated than what you'd get from Ford or Volkswagen. Basically, you staple a massive Android tablets onto your dashboard. Feels inevitable that the Western market will eventually trend here
2/ Every document hereon comes from my former employer Bernstein Research's internal research archive, which extend back to 1994
Unfortunately, they're not available to the public (even Bernstein's client website cuts off at 2003), but happy to give more details if necessary
18/ Just logged onto inflight WiFi - alright you folks asked for it
OBSERVATION
#5
- Livestreaming is a massive industry in China with no obvious Western equivalents
No matter where you go in China, youโll inevitably walk past somebody livestreaming on Douyin from their phone
22/ Nobody in China trusts web content anymore; a surprising % of the livestreams I saw were just basic product Q&A that weโd typically ask Google
If you can't figure out how to buy tickets up Tianmen Mountain on Baidu, why not just ask the guy livestreaming there instead?
This is why I struggle with paying 30x sales for SaaS stocks: Yeah, you might be growing 50%/year with 130%+ net retention rates *today*, but on the other hand, this 2004 Infoworld article lists 15 leading APM products and literally *zero* of them still exist
In 1993 the CTO of Microsoft predicted the internet would hollow out Americaโs cities as all information-based activity ceased to be geographically bound
If you have strong beliefs on the future, bet on the most obvious thing! Donโt bet on the 3rd order implication of the thing!
Met an OpenAI engineer that so strongly believes weโll have ASI soon that he sold his SF house and is looking to short the housing market of any cities that rely on human intellectual capital
This is pretty eye popping - the average US retail investor now expects a 17.5% forward return?
Reminds me of this December 1999 Bernstein survey where the average retail investor expected a 19% annual return on equities
Actual forward 10-year returns (1999-2009): -2.3% CAGR
25/ I also saw multiple teenagers texting their friends on QQ, which hilariously, is basically China's version of AOL Instant Messenger (launched 25-ish years ago as a forefather of WeChat)
I guess this is the China tech equivalent of bringing back baggy jeans
26/ Think I'll end it here for now
I've visited China before, but after a 10 year absence, this was by far my most fascinating trip
There's not a lot of other places in the world today where you get off the planeโฆ and basically none of your smartphone apps work anymore
24/ Everybody knows about Gen Z in the US fleeing Facebook for TikTok over the last 5 years
The Chinese equivalent: both my kid cousins have a WeChat but only use it to text their parents - otherwise "WeChat is only for old peopleโ
They chat their friends on Douyin instead!
1/ I donโt typically do reviews, but Super Founders was one of the most intriguing tech business books Iโve ever read
Most business books are just anecdotes and opinion. This book actually compiled quantitative data on 400 startups over 13 years to measure base rates for success
A new one for the Ozempic watchlist: the top 10% of customers account for 39-47% of fast food sales, per the latest credit card data from Earnest Research
1/ Just caught up with a few investor friends in the consumer space last week about Ozempic and GLP1s
As far as I can tell, everything basically hinges on: how much does it matter that every consumer product in the world depends on a tiny cohort of super consumers?
2/ What happens if it turns out weโve actually invented an all-purpose anti-addiction drug?
I suspect itโs not properly appreciated how many consumer categories follow a power law distribution of consumption
19/ Most of the streamers I saw were selling something
- A tour guide promoting airline flights to Xian
- A street food vendor streaming himself making sandwiches
- A travel influencer narrating his hike up Tianmen Mountain
- A merchant peddling durian at the farmers market
etc
When I started on Wall Street years ago, an older colleague advised me that the point of reading sellside research was to disagree with it - if you didnโt disagree with anything you were reading, you werenโt thinking about it hard enough
1/ Just realized we recently passed the 30 year anniversary of Nathan Myhrvold's internal Microsoft memo "Road Kill on the Information Highway."
This 1993 memo is probably the most prescient set of 10-ish predictions ever written on the rise of the internet
6/ This 2017 paper looks at 22 more CPG categories: the top 20% of US adults account forโฆ
70% of ice cream consumption
75% of coffee consumption
77% of soda consumption
87% of cigarette consumption
Something Iโve been pondering - the single biggest advantage that startups have over incumbents is probably the willingness to take insane reputational risks
I usually avoid discussing Tesla on Twitter, but Tesla in the early 2010s might be the best example of this ever
1/ This episode of
@AcquiredFM
was impressive: it might actually be the most thorough history of AWS ever published (take it from someone who spent a *lot* of time studying early 2000s IT infrastructure at my last job)
But it still leaves out one thing -
1/ I totally missed this news- Getir just shut down their US operations. The 2020-24 โquick commerceโ boom is officially over
The most interesting thing: pretty much all of this already happened 20 years ago with Webvan, the online grocery darling of the first tech bubble
1/ Just caught up with a few investor friends about the public equities job market today. You definitely feel a creeping dread around the rise of pod shops - Citadel, Millennium, and the ilk are now a quarter of all buyside seats (and probably >50% of incremental seats)
McKinsey estimates generative AI will create $2.6-4.4 trillion of economic value
Honestly, I mostly just feel bad for the 22-year old McKinsey BA with an econ degree from Dartmouth who was stuck having to make this up in Excel in February after Googling what a โtransformerโ was
Just learned Wikipedia publishes audited financials. Apparently they have $112M in annual expenses... for what??
Editors and admins are unpaid. Web hosting costs $2.4M a year. The rest appears to be overheard. Not thrilled to see this is where $150M in annual donations are going
A thing I sometimes think about is how during the 1997 to 2001 tech bubble, the telecom industry collectively spent $1 trillion dollars building the infrastructure of the modern internetโฆ and now 99% of people today donโt remember this ever happening
In retrospect, this was ridiculous: in 2021, Palantir invested over $700M in startupsโฆ on the condition they buy $700M in Palantir software. This singlehandedly drove a third of the companyโs revenue growth!
Departing Bernstein analyst's thoughts on investing in tech:
"There is a misguided obsession in many tech circles around predicting the future of technology. In contrast, I would posit that predicting the future is actually pretty easy - the hard part is making any money on it."
23/ OBSERVATION
#6
- It is a truth universally acknowledged, that every child in possession of a smartphone, must be in want of a social network not occupied by their parents
- a small observation I made while meeting two of my kid cousins (both <10 years old) for the first time
1/ I increasingly expect China will reach mass-market availability of self-driving cars 3-4 years before the US
Itโs not necessarily that Chinese self-driving tech is betterโฆ itโs just that the Chinese government seems way more tolerant of self-driving cars hitting pedestrians
21/ Why does anybody buy things this way?ย
Partly it's just cheaper - this is especially true for ecommerce livestreams, which tend to focus on discounting. These are some pretty cheap duriansโฆ
But partly I suspect livestreaming fulfills an informational void unique to China
1/ This article on Pinduoduo (the 3rd largest e-commerce platform in China) is ostensibly focused on their new US venture Temu, but it quickly devolves into the most outrageous profile of Chinese corporate culture that Iโve ever read
Mostly, this chart just makes me feel bad for the 22-year-old sellside associate who had to manually input 900 datapoints from an economic history textbook into Excel just so his boss could make a point about interest rates in the Neo-Babylonian Empire
20/ You can also buy products directly off livestreams in Douyin, Kuaishou, Xiaohongshu - they all have built-in shopping carts and contracts with 3rd party delivery services
One of my cousins estimated she made about a quarter of her last 12 months' ecommerce purchases this way
1/ In 2019, a fintwit topic de jour was how Doordash was torpedoing the economics of food delivery with crazy subsidies (culminating in Grubhubโs disaster quarter of Q3 19)
Fast forward today: Doordash killed Grubhub and is now annualizing $2.5B of profits (and $1.5B ex-SBC!)
Departing Bernstein analyst's thoughts on investing in tech:
"There is a misguided obsession in many tech circles around predicting the future of technology. In contrast, I would posit that predicting the future is actually pretty easy - the hard part is making any money on it."
It is moments like this when we gape in collective horror that IBMโs mainframe hardware business still generated revenues in 2022 that were *flat* with their revenues in 2000
The more time I spend at Google working on LLMs, the more Iโm struck by how young everybody is
The most knowledgeable people I know in the field are like, 5 years out of undergrad. A couple people didnโt even start working in ML until like 2022, and now theyโre world class
During Netflix's 2014 international rollout, my old firm sized their TAM in Brazil at 8 million - this was every household in the country with 2+ Mbps broadband.
Yet Netflix has over 15 million Brazilian subscribers today.
This is something I think a lot about.
1/ This 2014 paper offers a very interesting answer to a question that I hadnโt actually thought to ask:
Why were none of todayโs great global software companies born in Japan?
1/ From 1999: "It now seems reasonable to forecast within two decades, 30-50% of the nationโs electric supply will be required to meet the direct and indirect needs of the Internet."
Actual outcome: US electricity demand growth flattens to zero for the next two decodes
Very interesting Mosaic interview with a former IBM VP
Apparently IBM contemplated building a public cloud competitor to AWS as early as 2005 - but the idea was shot down by the head of IBM's IT outsourcing business, as it would cannibalize existing outsourcing contracts
Honestly, it's pretty shocking how many hotels and cruise stocks have basically recovered entirely from COVID when you add the debt and share dilution
Royal Caribbean, Norwegian, and Carnival are all trading at 94-101% of their pre-COVID enterprise values...
This is something I remind myself of a lot - long term equity returns are not consistently distributed
The last 140 years of the US stock market have seen FIVE 15-year periods of <4% annual returns
When I wrote this thread last year, growth tech valuations had just hit a 20-year high
Just 16 months later, and the average P/S multiple of the most expensive tech stocks has fully regressed to the 20 year mean
No idea if this is the end, but wow - what a crazy 16 months
1/ Lessons From The Tech Bubble:
Last year, I spent my winter holiday reading hundreds of pages of equity research from the 1999/2000 era, to try to understand what it was like investing during the bubble
A few people recently asked me for my takeaways. Here they are -
A useful rule of thumb I havenโt seen before: 1 NVIDIA H100 = roughly the same electricity consumption as an average US home (~1 kW)
So a 100k GPU training cluster has roughly the same needs as Cincinnati. Honestly Iโm not sure if this is higher or lower than I expected
1/ Thoughts on the Myth of the "First Mover"
This thread by
@danrose
stirred something I've been thinking about for a while - the myth of first mover advantage
To this day, most people assume Amazon Web Services was the first cloud computing service. This isn't quite true
I was at Amzn in 2000 when the internet bubble popped. Capital markets dried up & we were burning $1B/yr. Our biggest expense was datacenter -> expensive Sun servers. We spent a year ripping out Sun & replacing with HP/Linux, which formed the foundation for AWS. The backstory:
1/ A followup thought: the more time I spend working in tech, the more inclined Iโve become towards believing in the โgod of straight lines on chartsโ
Understanding the drivers of Mooreโs law is hard, but blindly extrapolating Mooreโs law was easyโฆ and 100% worked for 50 years
1/ Looking back on the history of computing in the late 20th century, I think we seriously underrate how much Mooreโs Law basically gave a free time machine to anybody paying attention
8/ For employers, seeing McKinsey on a resume has a similar signaling value to a top undergrad
ie not, โthis person will definitely be good at the jobโ
But rather, โthis person is guaranteed to have the bare minimum 2-3 years training that I didnโt want to do myselfโ
Rereading Mark Meekerโs original Internet Trends report from 1996 today. Just noticed the appendix listed ~120 internet firms from the period
I went and checked how many of those firms still exist today. I countโฆ 22
Easy to predict the future of tech. Hard to call the winners!
I think most of us have a decent intuition which modes of transport are riskier than others (eg motorcycles are likely riskier than, say, buses), but I did not fully grasp the magnitudes here -
Flying a private plane is *39x* more deadly per hour than driving a car!
This chart was very surprising to me
Itโs easy to say Blackberry was always doomed, ex post. But they outgrew the iPhone for 4 straight years from 2007 to 2011. How many of us wouldโve honestly looked at this chart in 2011, and said โyeah, I think Blackberry is going to zeroโ?
13/ While RIM peaked at 50m+ units sold in 2011 ($19B sales), that was the first full year iPhone outsold BlackBerry.
Blackberry share of the exploding smartphones fell from 21% in 2009 to 2% in 2013 (same year RIM renamed to Blackberry).
It stopped making phones in 2016.
I think about this a lot - the lifespan of an average tech company will likely be shorter than the career of its average employee
Of the 10 largest tech companies in the 1970s, only 1 remained on the list 20 years later (IBM). Of the 90s cohort, only 3 remained 20 years later
11/ The one catch? Timing
Nearly all these predictions took 15-20+ years to play out. WFH is still 25+ years in the making. Nothing in the memo (except shorting newspapers) would've been investable on any reasonable timeframe
Predicting the future is easy. Making money is hard!
Tech isn't synonymous with growth. Correlation not causation
When you look beyond FAAMG, US tech has grown revenues at just 3% over the last decade - no faster than the rest of the market. Over the last 25 years, tech ex-FAAMG has actually *lagged* the market (5% vs 6%)
1/ A fun piece of tech history that I came across today: an online copy of Microsoft's 1986 IPO prospectus. At just 52 pages, a light piece of weekend reading!
Some takeaways from my skim -
4/ At a mutual fund, even a junior analyst gets to pretend theyโre part of Das Kapital rather than Das Labor
But at Citadel, youโre just working on the factory line for Ken Griffin like everyone else
Does anybody else find it weird how hard NVIDIA tries to hide, like, the actual specs of their new GPUs?
The spec sheets are impossible to find, the FLOPS are inflated 2x by sparsity that nobody uses, and they say the B200 is a 30x improvement vs the H100 when itโs reallyโฆ 2.5x
It's somehow incredibly hard to get actual specs of the new Nvidia GPUs, between all the B100/B200/GB200/sparse/fp4 numbers floating around.
@tri_dao
linked this doc which thankfully has all the numbers in a table:
1/ This is the clearest articulation I've ever seen of the purpose of "superapps":
"When your app offers more services, customer acquisition cost (CAC) per service goes down, and CAC is the most important cost for Internet businesses."
1/ Every time I hear about the "metaverse" becoming the next internet (ugh), I don't think about the history of the internet - instead, I think about the history of the information superhighway
Aka the half-baked, corporate, also-ran version of the internet beaten by the web