233: Nvidia Giving Up on ARM?, Netflix FCF, Microsoft & Tiktok & Twitter, Shopify Logistics, Spotify Drama, Facebook Supercomputer, Gavin Baker, and OpenSea
"A sure recipe for success has been to take a 90’s protocol"
Forever is composed of nows.
🚨🚨🚨 In case you missed it, yesterday I published 𝕊𝕡𝕖𝕔𝕚𝕒𝕝 𝔼𝕕𝕚𝕥𝕚𝕠𝕟 #𝟟, an interview with David Kim aka Scuttleblurb. His answers are great and he drops a lot of gems 💎.
🎬 🎥 👾 I had an interesting discussion with friend-of-the-show and supporter (💚 🥃) Brad Slingerlend on the topic of movie special effects. With his permission, here’s what he had to say about it:
As I was watching that fx video you shared on T2 (side note, 1991 remains the greatest year for movies ever), I’m struck by how a lot of modern CGI takes me out of the movie compared to what I remember from the 1990s era fx. The digital fx that came before in the late 1970s and 1980s were novel, and then in the 90s they became really integral to taking you *into* the world of the movie.
It feels like it was more of a mix of digital, analog, and camerawork in the 90s (special artistry voodoo), and now it’s a lot of 100% digital. Maybe that’s it or maybe it’s something else. Surely our own memories, experiences, and biases are going to play into this based on our ages.
The teaser for Amazon’s new LotR series felt fake to a degree that pulled me out of the experience, in contrast to the original trilogy, which I recall pulling me in. But, both I believe were done by Weta, so it’s probably just me.
I generally agree, but the ways I'm explaining it is partly survivorship bias (we remember most the movies that had really great special FX for their time, and the terrible ones are often forgotten as they age really badly), and partly some kind of uncanny valley curve where the state-of-the-art isn’t quite there yet, but close enough that it becomes used for things that wouldn't even have been attempted not that long ago
I tend to think that it all comes down to the taste of the artists, not the tools themselves.
ie. Denis Villeneuve's special effects look and feel solid, like they have weight and are part of the world. Nolan tends to be good at that too, and uses a mix of miniature models and CGI to add some of that tangibility that can still be hard to do fully in CGI
It’s also probably a bit like cosmetic surgery: we notice it a lot when it's bad, but when it's really good, it may not even register.
I remember seeing a behind the scenes on the excellent ‘John Adams’ mini-series (HBO, 2008), and seeing how much was CGI that I wasn’t even aware of while watching and re-watching it
I also wonder if the VFX talent pool is being stretched too thin by all the the money pouring into the space, with all the big streamers and big tech investing more each year… 🤔 There's only so many great artists in the world, and at some point you go down the curve ¯\_(ツ)_/¯
🛀 I wonder how much faster some of the biggest discoveries of the past would have been made if the people at a time had access to the modern internet.
I don’t mean the actual data, the answers to their questions without having to do the work — just the communication capabilities to reach others, broadcast your ideas to the world, for communities to self-organize, etc.
For example, what is Newton or Darwin or Einstein had been able to discuss some of their theories, ideas, and observations with other scientists from around the globe at their very early stages, before they had answers or even the right question to ask.
Would it have given them breakthroughs years earlier? Would some discoveries that were only fully refined many years later have been more solid from the start? ¯\_(ツ)_/¯
🧩 Played Kingdomino for the first time with my family. Fun game! (Thanks to my friend Jeff for letting me borrow it)
💚 🥃 You know by now:
Liberty’s Highlights is reader-supported. To support my work, consider becoming a free or paid subscriber. 👩🏼🔧
A Word From Our Sponsor: ⭐️ Tegus ⭐️
What makes a great investor isn’t pedigree or diplomas 🎓, it’s quality thinking and information 🧠
Tegus can’t provide the thinking for you, but on the information front, they’ve got you covered! 📚
With 25,000+ primary source expert calls covering almost any industry or company you may want to learn about, and the ability to conduct your own calls with experts, if you so choose, at a much lower cost than you’ll find elsewhere (70% lower than the average cost of ˜$400), this is the service to turbo-charge the depth and breadth of your knowledge 📈
Investing & Business
Nvidia to give up on ARM deal..?
Nvidia is quietly preparing to abandon its purchase of Arm from SoftBank after making little to no progress in winning approval for the $40 billion chip deal, according to people familiar with the matter.
Nvidia has told partners that it doesn’t expect the transaction to close, according to one person, who asked not to be identified because the discussions are private. SoftBank, meanwhile, is stepping up preparations for an Arm initial public offering as an alternative to the Nvidia takeover, another person said. (Source)
Sounds like this is good for Intel and AMD, bad for ARM, and probably slightly bad for Nvidia, but not terrible — they have an ARM architectural license, they can keep developing datacenter products based on ARM like their Grace chip, they just wouldn’t have a new channel to sell Nvidia IP through, and may capture less of the upside if they help make ARM chips more competitive in the data-center through their R&D.
Intel and AMD must be rejoicing, as well as Qualcomm and other ARM chip designers, because even more Nvidia presence in the datacenter CPU market would’ve increased competitive pressure on them quite a bit, while ARM on its own won’t be able to match that R&D budget/talent pool, especially on the software stack front.
I wrote my thoughts in a lot more details about this in edition #227.
‘Nvidia and AMD GPU street prices are beginning to drop’
Speaking of GPU companies (you’d think this would be an easier segue to make..):
The slow but steady drop in crypto prices over the past month has in turn started a chain reaction in GPU pricing, as crypto mining becomes less and less feasible. Compared to last month, we're now seeing up to 11% lower card prices on eBay — and that's before the latest crypto losses. [...]
On average, nearly every graphics card dropped in price by 5% or more, with a few GPUs even breaking into the double digits — the RTX 3090, RTX 3080 (10GB) and RX 6600 all dropped by 10% or more. However, GPU volume (at least on eBay) also dropped substantially for nearly all GPU models, with only the RTX 3090 and RTX 3070 Ti selling more cards in the past week than in late December. (Source)
Netflix FCF/share since 2002
Not trying to make a point about anything specific, I just thought this was a really striking chart. They really went all-in when they decided they had to own their own content.
What an investment spike… You don’t see many companies with a decade history of positive FCF doing such a U-turn.
Alternate Timeline: Microsoft & Tiktok (& Twitter?)
It’s fun to imagine an alternate history where Microsoft bought Tiktok when it had the chance (for something like $50bn). What would this deal have looked like in more detail, and how would it have turned out?
It also makes me wonder if maybe Microsoft shouldn’t try to buy Twitter, since they seem to be the only Big Tech who can buy anything at this point.
Ideally they’d fix the broken parts and keep what works about it intact, because those parts are kind of magical and shouldn’t be messed with too much (which is what scares me every time there’s a rumor about some other company being interested… Like Disney a few years ago).
Shopify Logistics (Part 3)
Either Shopify didn’t like the reaction from merchants to its recent apparent change of mind on creating a capextastic logistics platform and reversed course, or their position was misreported/misunderstood, but either way, the company released this statement on Monday:
We will be making changes to (Shopify fulfillment network) to help merchants compete with big-box retailers, such as prioritizing two-day shipping at affordable prices and access to easy returns for U.S. shoppers.
It remains to be seen what that means in practice.
Spotify Drama (Part 7 of ???)
Not to be confused with Shopify (dyslecixs of the world unite!), here’s a new potential vector for drama with Spotify:
“I want you to let Spotify know immediately TODAY that I want all my music off their platform,” he continued. “They can have [Joe] Rogan or Young. Not both.”
This is Neil Young threatening to remove his catalogue from the service because he doesn’t like some of the content that Spotify is hosting on the podcast side.
The specifics don’t interest me much, and Neil Young alone won’t move the needle (🎶 the needle and the damage done… 🎶) , but the idea is intriguing: What happens if Taylor Swift and Beyoncé and Kanye start making demands, or a bunch of big stars band together to put pressure on as a group 🤔
I’m not saying it’d be the end of the world for Spotify, but could be quite annoying and hurt the brand if really popular people keep saying bad things about you with really huge megaphones. Drama drama drama…
And it also would suck for music fans who suddenly can’t listen to artist X because they get caught in the crossfire of a bunch of political stuff… ugh
Netflix Films: Valued lower by viewers because not in theaters?
Compound with an interesting theory:'
I wonder if not having nflx films globally released in theatres first lowers the value of nflx films in consumer minds relative to movies that release in traditional theatre windowing.
For example, if I can get the new spiderman on streaming service when it's showing in theatres that streaming service immediately feels more valuable to me.
In comparison, if red notice is not that big in theatres or not showing in your local cinema at all, having red notice show up on nflx, the perceived value prob lower? Red notice might be a blockbuster, but I don't feel like it's a blockbuster since it has no presence in my local cinema.
Consumers still have mental anchoring to blockbuster = big cinema release
Consequently, I don't feel like red notice added a ton of value to my $nflx sub even though it was a nice film to have
What do you think?
You can leave a comment with this convenient button:
Interview: Gavin Baker, Atreides ⚔️
I enjoyed this interview of Gavin Baker by Patrick O’Shaughnessy (I almost went with the dad joke of “Gavin 👨🍳🥖🥐🧁” ):
Here’s a highlight:
I think there's two parts to inflation. The first one is supply chain driven, the shortage of goods everybody's read about. Ships stacked up the port, we can't get enough semiconductors to make cars. I am so relaxed about that. It is very rare for me to have a view on something like that. I just think we now have hundreds of years of history and capitalism is amazing at solving problems. It is so good. And we have seen a massive supply response.
This is a statistic, I actually just ran this this morning. Amazon has spent more money on capex […] in the last two years than they did in the preceding 20 years.
Think about that. From 1999 to today spent 62 billion on capex. They're going to spend 87 billion in 2020 and 2021. [...] [TSMC], their 2022 capex is going to be many multiples of 2018. 2021 and 2022, they'll spend more than they did in the preceding five years. So there is a massive supply response coming.
Liberty’s Highlights is reader-supported. To support my work, consider becoming a free or paid subscriber. 🦦
Science & Technology
Facebook’s New Supercomputer (16,000 Nvidia GPUs)
Face… Meta announced a new AI supercomputer, which should be the “fastest AI supercomputer in the world when it’s fully built out in mid-2022”. The previous one was:
The first generation of this infrastructure, designed in 2017, has 22,000 NVIDIA V100 Tensor Core GPUs in a single cluster that performs 35,000 training jobs a day.
The new daddy is called RSC:
AI supercomputers are built by combining multiple GPUs into compute nodes, which are then connected by a high-performance network fabric to allow fast communication between those GPUs. RSC today comprises a total of 760 NVIDIA DGX A100 systems as its compute nodes, for a total of 6,080 GPUs — with each A100 GPU being more powerful than the V100 used in our previous system. The GPUs communicate via an NVIDIA Quantum 200 Gb/s InfiniBand two-level Clos fabric that has no oversubscription. RSC’s storage tier has 175 petabytes of Pure Storage FlashArray, 46 petabytes of cache storage in Penguin Computing Altus systems, and 10 petabytes of Pure Storage FlashBlade.
So how does it compare to the previous gen?
Early benchmarks on RSC, compared with Meta’s legacy production and research infrastructure, have shown that it runs computer vision workflows up to 20 times faster, runs the NVIDIA Collective Communication Library (NCCL) more than nine times faster, and trains large-scale NLP models three times faster.
Impressive, but it’s not done yet:
Through 2022, we’ll work to increase the number of GPUs from 6,080 to 16,000, which will increase AI training performance by more than 2.5x. The InfiniBand fabric will expand to support 16,000 ports in a two-layer topology with no oversubscription. The storage system will have a target delivery bandwidth of 16 TB/s and exabyte-scale capacity to meet increased demand. (Source)
Peter Attia Podcast: ‘COVID Part 2: Masks, long COVID, boosters, mandates, treatments, and more’
This is a follow-up to the podcast that I wrote about in edition #224.
Peter wrote a separate piece:
There’s a lot of food for thought here. The section in the podcast on ‘long covid’ was particularly elucidating (and reassuring), IMO.
I think the important thing — and you may start to recognize this as one of my favorite words — is to try to be as well-calibrated as possible based on the known facts at the time, and not get stuck on what was the best thing to do with what we knew and the different circumstances of 2 years ago or whatever.
Centralized vs Decentralized, Life-is-Trade-Offs Edition
Moxie Marlinspike (real name: Matthew Rosenfeld, best known at the CEO of Signal, though he recently announced he was leaving that job) recently wrote an interesting piece about Web3.
I don’t want to discuss the whole thing — many others already have — just one of his points:
A protocol moves much more slowly than a platform. After 30+ years, email is still unencrypted; meanwhile WhatsApp went from unencrypted to full e2ee in a year. People are still trying to standardize sharing a video reliably over IRC; meanwhile, Slack lets you create custom reaction emoji based on your face.
This isn’t a funding issue. If something is truly decentralized, it becomes very difficult to change, and often remains stuck in time. That is a problem for technology, because the rest of the ecosystem is moving very quickly, and if you don’t keep up you will fail. There are entire parallel industries focused on defining and improving methodologies like Agile to try to figure out how to organize enormous groups of people so that they can move as quickly as possible because it is so critical.
When the technology itself is more conducive to stasis than movement, that’s a problem. A sure recipe for success has been to take a 90’s protocol that was stuck in time, centralize it, and iterate quickly.
I think this is a deep insight.
People spend too much time focusing on the technology itself — its architecture, lines of codes, what it can do — and not enough on what’s going up one level up, at the level of the group of people developing it, and what may cause them to move fast or slowly, to succeed or fail, to be able to direct their energies in productive directions in a coordinated manner, or to splinter/shatter into a thousand sub-projects that don’t collectively achieve much movement in any specific productive direction…
OpenSea Bug Used to ‘Steal’ NFTs
While there are many benefits to decentralized platforms where “code is law”, one downside is that when bugs are exploited to steal your stuff, it’s a lot harder to do anything about it.
The exploit appears to originate from the ability to re-list an NFT at a new price, without cancelling the previous listing. Those previous listings are now being used to purchase NFTs at prices specified at some point in the past - which is often well below current market prices.
The bug was first reported some weeks ago but doesn't appear to have been exploited to any significant extent until today.
Here’s what happened:
For example at around 7am today, a Bored Ape Yacht Club NFT #9991 was purchased for 0.77 ETH ($1,800). This family of NFTs currently sell for at least $198,000. Twenty minutes later the hacker sold the NFT for 84.2 ETH ($196,000) – realizing a profit of $194,000.
One attacker, going by the pseudonym "jpegdegenlove" today paid a total of $133,000 for seven NFTs – before quickly selling them on for $934,000 in ether. Five hours later this ether was sent through Tornado Cash, a "mixing" service that is used to prevent blockchain tracing of funds. (Source)
I’m curious if OpenSea will ultimately take the hit and make its customers whole — the reputational damage may be costlier than just paying…
But if there’s ever a very large-scale exploit of this kind where hundreds and hundreds of listings are affected, the math may be pretty ugly.
Liberty’s Highlights is reader-supported. To support my work, consider becoming a free or paid subscriber. 🌵