No. 25: Musings about Audio and Twitter, Memory Usage, Gavin Baker Semiconductor Interview, US Healthcare Chart
I took a short break from writing. Not intentionally – but it just sort of turned out that way. But I’m back in the game this week! And it looks like I’ll be releasing a few longer form pieces over the winter.
Audio/Media
I think about media quite a lot – mainly the fact that companies fight for a finite amount of consumer attention. Apparently, screen time is quite large for Americans – between 5 and 7 hours per day. The only thing that increases the total amount of possible media screen time is productivity gains giving people extra discretionary time.
What’s interesting to me is as more of my time goes towards exercise or housework, the audio format is the only thing that’s usable for me during these activities. After all, I can’t exactly watch a Netflix series and go for a run, but I can easily do it while listening to a podcast.
Not only is audio a perfect companion for things like housework, cooking, or exercise - it makes these times the most valuable learning opportunities of my day. So, though I’ve probably slowed down on reading books, I wonder whether my learning has slowed down at all!
I guess if my habits are in line with larger trends in media consumption then this bodes well for audio platforms like Spotify – if they can become independent of the large record labels.
Twitter has supercharged my investment learning. It’s safe to say that without it, I’d struggle to access some of the insights that I can otherwise get. In addition, it’s also likely that Capital Buddha would have significantly less readers without it.
The platform can be a double-edged sword. It can be an echo-chamber where I might be smelling my own farts for long periods of time. However, it can often be a very powerful curation tool – where I follow industry experts or other curators who have excellent information which I need for my investments or profession.
It’s not that books are obsolete, but I’m finding more that Twitter and podcasts, carefully chosen, can provide higher ROI on time than most books. Not to mention, I found many interesting people in my investment network through Twitter and attended many great discussions in the Spaces product.
Twitter barely captures any of the value that it creates in the platform. So, it will be interesting to see whether it can capture value and then reinvest proceeds into growing the platform.
Memory Usage
This article is a short one – and illustrates that memory intensity of devices tends to go up over time. Every single phone I’ve had has run out of space within a year. Granted, I usually buy second-hand phones at least a generation behind – but it seems that as phone specs get better, the size of media and content goes up.
More storage = food/cat pictures in higher resolution.
This trend leaves me confident that bit shipments of memory – both DRAM and NAND are going to increase for years to come. Additionally, considering that there’s only three main players left in the DRAM business and a handful in the NAND business – along with secular demand growth trends for years to come, I believe memory has become an investible space. But since the industry operates off a large fixed-cost base, it’s likely there will be industry downturns in profitability which can be used to add exposure to the sector.
Interview with Gavin Baker about Semiconductors
Gavin Baker is awesome. I’d highly recommend reading the full article if you are interested in technology generally. This excerpt was super interesting, explaining how Google created its own proprietary chips which let them save on building football fields of data centres:
Now, we have a positive demand shock in the form of artificial intelligence. Sure, PCs have leveled out, smartphones are slowly additive, the electronic content of cars is growing and so is the internet of things. But AI, because data quantity is so important for quality, is much more semiconductor intensive. For me, a light bulb went off when I read this article in «Wired» about how Google’s engineers started to deploy voice recognition based on AI to Android smartphones. They realized that if each Android phone used Google’s voice search for just three minutes a day, Google would need to build twice as many data centers. What that means is that AI is so incredibly computation intensive that it led Google to develop its own chip called the Tensor Processing Unit or TPU. It’s kind of a graphics processor competitor and it saved Google from building a dozen new data centers.
Chart of the Day
I understand that a certain level of administration is necessary to save the time of doctors – but what this chart shows is sheer bloat. Bureaucracies develop, people hire more people, then send emails to each other all day long. More meetings, more alignments, more vested interests blocking all progress. It’s sad to see that this is where 20% of US GDP goes – especially since US health outcomes are not the greatest in the OECD.
I often have conversations with friends in start-ups that say that sometimes the most counter-productive thing they can do is hire someone. It’s a hard balance to strike to know when to hire someone, versus simply trying to allocate work amongst your existing team better.