490: TSMC & Hungry Datacenters, Marketing to AIs, Patrick Collison & Stripe, Ringtone Empire, Bee Apocalypse, Magnesium, and Rome

"they were bored and over it"

Your assumptions are your windows on the world. Scrub them off every once in a while, or the light won't come in.

―Isaac Asimov

🎭 🙌 📉 The eclipse was a great object lesson for my kids about expectations vs reality, and how our expectations going into something have a *REALLY BIG* impact on how we perceive that thing.

They had been hearing about the eclipse for weeks at school, doing art and science projects about it. They wanted to eat popcorn while we watched, treating it like a big party…

And 10 minutes into it, they were bored and over it.

It’s not that they didn’t like it, but there was no way it would live up to their expectations.

Afterwards, we discussed how to avoid this. Like if there’s a film you love and you want to recommend it to someone, don’t tell them it’s the best, most amazing thing ever and that you know they will LOVE it. Instead, give them just enough to convince them to watch it, but nothing more than that. Anything above that line *decreases* the odds that they’ll enjoy it.

Related to this, Tim Urban wrote a good piece back in 2013 (time flies!) about expectations & happiness:

• 📄 Why Generation Y Yuppies Are Unhappy

☀️🌝🌎🤔 Are we gonna see a bunch of companies with "Eclipse-adjusted EBITDA" in Q2? 'Solar-normalized revenue' and 'Penumbra-corrected cash flow’?

🍳 Back in the summer of 2023, I got a Matfer Bourgeat carbon steel fry pan (I got a good deal on it — surprise!).

I’ve been meaning to tell you about it, but it keeps getting bumped down the list… So here’s a primer on why carbon steel may be for you:

I’ve been happy with mine. I did three coats of seasoning when I first got it using a wax called “Lancaster Cast Iron Seasoning” from Amazon. It did a good job and it has been decently non-stick and well-protected since.

I also got a chainmail scrubber. It works great on cast iron and carbon steel.

---

A Word from our Sponsor: 💰 Brinker Advisor 💰

Are you ready to take control of your investments and secure your financial future? Our premium subscription is designed to empower you with the tools and insights you need to make informed decisions and maximize your returns.

🔑🔓 When you become a paid subscriber, you'll gain exclusive access to:

• Marketimer Model Portfolios I, II, and III, Income Portfolio, Active/Passive Portfolio 📊📈🏦

• BFIA Model Portfolios Aggressive, Moderate, Conservative, Tax-Exempt, Exchange-Traded Fund Portfolio 🧭

• Recommended List of No-Load Mutual Funds (updated daily) 📑🕵️‍♂️

• Full Access to All Archived Posts 🗄️📜

All this along with our favorite investment ideas and our views on the economy, monetary policy, and related topics.

⭐️ Sign up here and get 20% off your first year with the coupon code “Liberty20” (offer good through June 30th, 2024). ⭐️

---

🏦 💰 Liberty Capital 💳 💴

🇹🇼 TSMC’s Energy Footprint & Hungry Datacenters 🐜🔌⚡️🏭

Friend-of-the-show Dylan Patel (🐜🔍) posted this slide showing TSMC’s annual energy consumption along with this context:

Datacenter vs semiconductor fab: a typical hyperscale datacenter, such as Meta’s “H” building, has a power capacity of 50-100MW. This is akin to a small to mid-sized fab; however, we estimate TSMC’s “Gigafabs” require north of 200MW!

TSMC indeed has an energy footprint roughly equal to Google's and growing 10% CAGR 2016-22... but slower than hyperscalers (Google +17% CAGR)

Just for fun, I went to the site of Taiwan’s TaiPower and dug this up:

“Renewables” (a mix of hydro, solar, and wind — “PS hydro” is pumped storage, so more like a battery) are over 30% of the installed generation capacity but just under 10% of energy produced.

Coal and gas account for almost 80%. while nuclear is unfortunately losing share (hopefully that changes thanks to the efforts of locals like Angelica Oung).

Zooming out, the energy required for all these chips — not just making them but powering their operation — is compounding at a rate high enough (much higher than GDP/population growth) and over long enough that it went from ‘not significant’ on the global macro level to rapidly becoming something potentially disruptive for electric grids. It can take years to build new generation and transmission capacity, the ship needs to be steered long in advance. There’s no time to waste.

Dylan Patel writes:

From a supply perspective, sell side consensus estimates of 3M+ GPUs shipped by Nvidia in calendar year 2024 would correspond to over 4,200 MW of datacenter needs, nearly 10% of current global datacenter capacity, just for one year’s GPU shipments. [...] Nvidia’s GPUs are slated to get even more power hungry, with 1,000W, 1,200W, and 1,500W GPUs on the roadmap.

That’s just GPUs from Nvidia! There are tons of other chips from Intel and AMD and ARM and the custom chips of the hyperscalers.

Additional context:

Estimated global data centre electricity consumption in 2022 was 240-340 TWh, or around 1-1.3% of global final electricity demand. This excludes energy used for cryptocurrency mining, which was estimated to be around 120-240 TWh in 2022, accounting for 0.4-0.8% of annual global electricity demand.

2022 was before the generative AI boom really got going… At least ETH going proof-of-stake helped, but there are still plenty of crypto ASICS churning out hashes out there.

All this is in parallel with the electrification of transportation and building heating via heat pumps…

🤖📢 Marketing to… AIs 💡(the birth of the AIO industry 🫣)

I admit I hadn’t thought of this before, but as soon as I heard the concept, it made all the sense in the world.

Now that more and more people are asking questions directly to an AI and getting an answer instead of a bunch of links with short blurbs from which they themselves pick, there will be a huge incentive to be good at “marketing” to AIs so that they recommend your products and services to their users. Or even just describe your stuff in a positive way.

Not only that, but because AIs are becoming increasingly able to use tools, APIs, and eventually other AIs, there will be marketing to AIs directly to convince them to use your tools and AIs rather than someone else’s.

It’s one thing to market to a machine hoping it’ll pass on your message to a human (a kind of AI SEO) but it becomes a bit more mind-bending when the goal is to market *only* to a machine, hoping to convince it to use another machine.

Thanks to strong incentives for ranking highly, the SEO industry has played a big part in ruining the web by pumping out an endless stream of garbage. I hope we can find better counter-measures when it comes to AIO… 😬

🗣️ Interview: Patrick Collison, Co-Founder of Stripe 💳

That was a good one, wide-ranging, as expected with PC and Dwarkesh.

I like the part where they discuss the different kinds of innovation at startups vs mature companies, how they think about infrastructure, and the biotech research that falls between the cracks:

• 🎧 Patrick Collison (Stripe CEO) - Craft, Beauty, & The Future of Payments

Interesting riff on the importance of craft and beauty even at large scale:

The best people consider themselves craftspeople in their domain. And they really, above almost all else, want to work with the best other people.

And so I think it may almost be true that even if from a customer facing standpoint, craft was not valued by the market.

You actually might still want to build an organization that index is very heavily on this because you just want the best people for other reasons. And now as it happens, I think customers do in fact value it and I think the evidence is broadly consistent with that.

But yeah, I think it's very hard to assemble groups of the best people if you don't take the practice of the work super seriously.

This makes a lot of sense to me.

From outside a field, observers often think of creators as “artists”.

But inside, the people making things mostly think of themselves as craftspeople, just plugging away at it, trying to get a little bit better, learning a few more tricks, gaining a few more skills, and improving iteratively while always looking for a new breakthrough.

They largely measure results by their own internal yardstick (they’ve usually spent years honing their taste and are their own harshest critics) or based on feedback from a few respected peers. 📏

📳 That brief moment when the telcos and the music industry minted money with Ringtones 💰💰💰💰💰💰

I dug up some old articles about ringtones (which, at the time, were expected to keep growing for a long time along with the mobile revolution!):

• “In 2006, ring tones accounted for over 10% of the $ 32.3 billion worldwide music market.”

• “In 2006, global ringtone sales were $4.4 billion, up from $3.7 billion in 2004”

• “Jamba was behind the popular "Crazy Frog" ringtone and was bought for $273 million in 2004 by VeriSign” (!!!)

• “In 2009, Lil Wayne's "Lollipop" became the biggest-selling ringtone ever, going five times platinum.”

Having never bought ringtones, I didn’t realize that they were sometimes (often?) a subscription:

• “Ringtones could be quite costly, with fans paying up to $3 for a 15-second clip of a song that they could only keep for 90 days before having to repurchase it. This subscription-like model made ringtones very lucrative for mobile carriers and record labels.”

It was such a craze that the media even had ringtone anecdotes:

• “During a somber after-death ceremony at a hospice facility, someone's phone went off playing "Another One Bites the Dust" by Queen. The awkward moment was diffused when the deceased patient's family member started laughing, saying it was perfect for their loved one”

Graph via Chartr

---

🧪🔬 Liberty Labs 🧬 🔭

This is not a real photo. I was curious how well Midjourney V6 did macro shots of bees 💚 🥃

🐝🐝🐝🐝 What happened to the “Bee Apocalypse” from a few years ago? 🍯 🐝🐝🐝🐝

Starting in 2006, there was widespread concern about bees. Beekeepers reported losses of 30-90% of their hives, with the bees mysteriously disappearing.

Headlines about Colony Collapse Disorder with photos of dead bees were frequent, and the whole thing was dubbed the Bee Apocalypse, with graphs showing how many hundreds of billions of dollars were created by our pollinator friends and the massive impact their disappearance would have on ecosystems.

I’m not trying to downplay it using hindsight bias. It was truly concerning at the time, and nobody knew if it would keep going or stop, or what was to blame.

Various groups blamed their favorite targets: Cell towers, high voltage power lines, toxins and chemicals, global warming, parasites, pesticides, fungi, habitat loss, etc.

Honeybee populations bounced back. The phenomenon lasted about 3-5 years and is no longer considered a major threat. Globally, honeybee populations increased 45% over the past 50 years and are at a record high.

In the U.S., honeybee colonies hit a 20-year high in 2014, increasing from 2.4 million hives in 2006 to 2.7 million in 2014. Beekeepers have been able to rebuild colonies by splitting hives.

Source: USDA annual report on honey-producing colonies in the U.S

A meta-lesson is that while the problem was going on, we all knew about it. But when it ended, that didn’t get nearly equivalent attention. It’s like this with many things, so it’s easy to get the impression that there are only problems out there and they never get solved, they just accumulate.

Another meta-lesson: in the moment, it’s very easy to extrapolate, but if we zoom out, we often find out that what seemed like an intractable problem was just a blip (it doesn’t mean we shouldn’t care about problems or assume everything just fixes itself — we should work super-hard to fix problems without panicking). 🛠️⚙️

🥬 💊 What do you know about Magnesium? 🩻🦴🔍

I enjoyed this primer on magnesium’s role in our bodies:

• 🎧 The Science of Magnesium and Its Role in Aging and Disease | Dr. Rhonda Patrick

I’m making a few changes after listening to it:

1. I’m lowering the strength of my priors about Magnesium L-Threonate. I’ve been taking it before bed for a while, both as a general magnesium supplement and because it’s apparently best at crossing the blood-brain barrier. As Rhonda explains, the studies on it are not super convincing AND it’s not the best at providing elemental magnesium. It’s not bad, maybe just not as good as I thought it was.

2. I’ll be adding magnesium glycinate to my cocktail of supplements (bis-glycinate is the same thing, just the full name of C4H8MgN2O4). I’ll also try to eat more leafy greens, but I know I don’t eat a lot of them every day and I’m not likely to start all of a sudden, so supplementation is my safety net.

Given that approx half of the population lacks sufficient magnesium, which plays a crucial role in vitamin D metabolism and is linked to conditions such as osteoporosis, heart problems, and cancer, it's worthwhile to educate yourself about this mineral and, if needed, take steps to address deficiency.

---

🎨 🎭 Liberty Studio 👩‍🎨 🎥

⚔️🏛️👑 Rome (2005, HBO) 📺 💰💰💰

It has been ˜15 years since I saw the series ‘Rome’ (HBO, 2005).

I remember *really* liking it. I wonder how well it has held up. I’ve heard from a lot of people telling me it has — I’ve bumped it up on my list of things to rewatch soon. If you’ve never seen it, it’s definitely worth watching and likely much better than whatever else you were going to watch next!

It’s kinda tragic they could never execute their full vision and had to rush through the story during S2.

If it had come out in 2015 instead of 2005, the budget — which was the big problem with it at the time, with the first season costing $100m making it one of the most expensive shows ever — would likely have been no problem. Amazon is reported to be spending $1bn on that Tolkien series, with over $460m on just S1 (while I haven’t seen it, it looks pretty mediocre).

It would also likely have found an audience thanks to the on-demand nature of today’s platforms. We forget how difficult it was for complex shows with long story arcs to build momentum in the pre-streaming world.

Anyone who misses the first few episodes is probably hopelessly confused and can’t get into it. Even if you’re there from the start, if you miss an episode or two in the middle of a season, you could also be left behind…

By the time DVDs came out for S1, the show was canceled (it got axed in the middle of S2 production).

We’re overdue for a new truly great TV show set in that world. Something epic, with ‘Games of Thrones’ production values but set in antiquity.

These thoughts were triggered by listening to the excellent podcast Cost of Glory (🎧), specifically part 1 of Alex’s new series on Pompey (Kid Butcher!). Check it out!