43: My Thoughts on iPhone 12 Cameras & Silicon, Databricks IPO, Gavin Baker on Retailers & Amazon, Heterogeneous CPU Design, Headlines vs Actual Stories, and Solar Now Cheapest Electricity in History
"Amazon’s retail business may actually be disadvantaged long term by Covid"
Correlation doesn't imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing 'look over there'. —Randall Munroe
It’s weird, when I started this newsletter project, I thought I’d quickly run out of things to say. I only had a handful of ideas going in, so they’d soon run out… Now, I’m finding that the more I write, the more things I have to say.
¯\_(ツ)_/¯
If you’ve been a lurker on Twitter or forums or have wanted to start a blog or a newsletter but haven’t because you think you’ve got nothing to contribute and you wouldn’t be good at it anyway, the way to get out of that unproductive loop is to just punch through and do it (Arnold voice: dooo iiiit!).
If you don’t, you can be 100% certain nothing will happen. If you do, I can’t promise success, but there’s a non-zero chance that you’ll find out it was a good idea, and that all those other people you’ve been reading aren’t special, they just gave it a try and have been learning from the experience ever since.
Throw your hat in the ring, you can always take it out if it’s really not for you. Be like Amazon, do lots of small experiments, then put more time and energy behind those that work, and drop those that don’t (also remember Bezos’s two-way doors vs one-way doors).
Don’t sit around waiting to know in advance that something will be 100% perfect before acting, that never happens.
→ If you’re a US citizen, please vote. ←
Investing & Business
Databricks Plans IPO in 2021
Snowflake competitor and Microsoft partner (Microsoft has also invested in the company) has no doubt been motivated by Snowflake’s 200x+ EV/sales public valuation:
Databricks is aiming to go public at a price significantly higher than the valuation in its last funding round, one of the people said. It was valued in 2019 at $6.2 billion when it raised $400 million [...]
Databricks has raised about $900 million from backers including Andreessen Horowitz, Coatue Management, New Enterprise Associates, Tiger Global Management, BlackRock Inc. and T. Rowe Price Group Inc. Its backers also include Battery Ventures, Sinewave Ventures, In-Q-Tel and Data Collective.
Databricks, founded in 2013, has more than 5,000 customers (Source)
I posted some thoughts about Snowflake’s S-1 filing pre-IPO in edition #17.
The Changing Color of Renewable Energy in the U.S.
O&G OG a.k.a. @EnergyCredit1 has a good post full of charts (who doesn’t love charts?) slicing and dicing renewable energy in the U.S., how it has changed over time, which sources are on the upswing/downswing, which states are producing the most as a percentage of their total production, which are producing the most in absolute quantities, and how things are spread around when it comes to the red/blue political landscape.
Some takeaways:
In July 2020, three Republican states (TX, OK, IA) generated more wind & solar electricity generation on a trailing-12-month basis than all 20 Democratic states, plus Washington, D.C. combined
Nearly 30% of the Democratic state power stack is generated from renewable power vs 15% for the Republican state power stack, but Red states generate more renewable power than Blue states
Republican states generate just over 50% of all renewable power in the US currently, but 2/3 of all wind and solar power
Texas dominates U.S. wind and solar renewable scene at nearly 25% of U.S. generation, an amount totaling over 70% of all wind and solar generation from Blue states
If we look at coal, the demise has been rather rapid, and is no doubt terminal:
Red States: since 2001, coal has dropped from 60% to 25% of the generation mix, natural gas increased from 15% to over 40%, nuclear held steady and renewables gained nearly 10% of the power generation stack
Blue States: since 2001, coal has dropped from 30% to 9% of the generation mix, natural gas increased from 23% to over 37%, nuclear held steady and renewables gained 13% of the power generation stack
Texas is a powerhouse when it comes to fossil fuels, but also when it comes to renewable energy:
Texas Wind + Solar Generation is > 70% From All Blue States
Texas has 30,217 MW wind capacity, ranking it number one in the US
Texas has nearly 3x the wind capacity of the 2nd largest state Iowa and nearly totals the next four biggest states combined
Texas has invested $53 billion in the wind sector, outpacing any other state by at least $30 billion
Texas represents 29% of US wind and 7% of US solar on a trailing basis
Texas alone produces more than 70% of wind and solar from all Blue states combined.
You can read the whole thing here.
IEA: Solar is Now ‘Cheapest electricity in history’
Speaking of renewables, a new International Energy Agency report once again increases estimates on how much solar power will be built (past forecasts have all undershot reality so far too) and reports that “the cost per megawatt to build solar plants is below fossil fuels worldwide for the first time.”
For projects with low-cost financing that tap high-quality resources, solar PV is now the cheapest source of electricity in history. [...] The IEA’s main scenario has 43 [percent] more solar output by 2040 than it expected in 2018, partly due to detailed new analysis showing that solar power is 20 [to] 50 [percent] cheaper than thought.
A big part of it is financial, not technological. The cost of capital is lower than previously thought:
Previously the IEA assumed a range of 7 [to] 8 [percent] for all technologies, varying according to each country’s stage of development. Now, the IEA has reviewed the evidence internationally and finds that for solar, the cost of capital is much lower, at 2.6 [to] 5.0 [percent] in Europe and the US, 4.4 [to] 5.5 [percent] in China and 8.8 [to] 10.0 [percent] in India.
Coal is in structural decline, but both oil and gas haven’t seen their peaks yet, though if action on climate gathers speed, as it appears to be doing, current projections could be off again:
Despite a more rapid rise for renewables and a “structural” decline for coal, the IEA says it is too soon to declare a peak in global oil use, unless there is stronger climate action. Similarly, it says demand for gas could rise 30% by 2040, unless the policy response to global warming steps up.
If you don’t believe me about past solar forecasts always undershooting and being revised up, check this out:
Gavin Baker on COVID19’s Impact on Retailers, Amazon
I’ll admit I don’t spend a lot of time thinking about brick & mortar retailers (is there a better term than this, btw? I’m tired of the bricks and the mortar…), but Gavin Baker makes an interesting argument (bolded parts mine):
I believe the biggest long term beneficiaries of Covid will prove to be category leading brick and mortar retailers. By this I simply mean brick and mortar retailers who have dominant share in a category — whether it be home improvement, general merchandise, electronics or any other retail category. Their destiny has likely changed forever. Many of the perceived Covid winners such as e-commerce, videogame and streaming media companies have simply been pulled a few years forward into a future that was inevitable. Their destiny did not change. The future for those businesses simply accelerated whereas the future for category leading “brick and mortar” retailers has changed dramatically as a result of Covid, more so than for any other business of which I can think. [...]
long term steady state FCF will likely be significantly higher for category leading brick and mortar retailers who had reasonably strong e-commerce businesses coming into Covid [...]
The future was always going to be omnichannel. Pundits have been prematurely predicting this for many years, but it is finally happening. [...] Brick and mortar stores significantly lower online CAC by improving marketing efficiency (higher click through rates, higher quality scores for ads). Consumers are more likely to trust a brand they have seen in the real world. Ironic in a world where “CAC is the new rent” that one of the best ways to lower your online rent, i.e. CAC, is to pay rent offline for physical stores. [...]
Perhaps the simplest way to express what has happened during Covid is to note that Amazon has actually lost share in e-commerce during Covid. The largest e-commerce share gainers in most categories have been category leading physical retailers as well as the DTC businesses of most brands. [...]
Amazon’s retail business may actually be disadvantaged long term by Covid due to more intense long term competition as thousands of SMBs have taken advantage of services from Wix, Shopify, etc. to go online, category leading retailers are investing online like never before and brands are investing in their DTC strategies — all at the same time that Amazon may have begun to erode consumer trust with advertising.
Read the rest here (great anecdote at the end about Gavin’s “largest day of portfolio turnover of my investing career”, as a postscript, don’t miss it).
Crossover Interview: Invest Like the Best x Acquired
I’ve enjoyed both episodes of this crossover. I already mentioned the one where Patrick is interviewed by the Acquired hosts in edition #38, but there’s also a recent one where the Acquired hosts are interviewed by Patrick.
The discussion of what they consider the best acquisitions of all time is interesting, but I found particularly thought-provoking the parallels that they made between yesterday’s media businesses and today’s tech/software businesses. It makes a lot of sense to me, but I had never quite thought of it that way.
And the point about Android having huge value to Google as a defensive asset that keeps TAC lower than it otherwise would be is one I’ve often made in the past and totally agree with. When Eric Schmidt was on Apple’s board and saw the iPhone coming early, he no doubt saw that a new middleman between Google and its customers was about to appear on the scene, and felt that they had to neutralize that threat at almost any cost.
Oh, and speaking of Acquire, I enjoyed their episode about Intel’s legendary switch from being a memory maker to a microprocessor maker.
It’s also a good reminder to read Andy Grove’s books (I’m thinking of ‘Only the Paranoid Survive’ and ‘High Output Management’), if you haven’t.
Science & Technology
‘Ahead rounds produce a high density cloud of sub-projectiles’
Very impressive tech, magnetically programming every round in-flight as it passes the muzzle.
What if more of Earth’s top engineers were working on things other than ways to blow people & things up and make people click on ads, though? It’s a cliché, but that doesn’t make it false…
My First Impressions on the iPhone 12 Pro
On Friday I received my iPhone 12 Pro (Pacific Blue, 256gb). My first impression, as a replacement for an iPhone X, is that the camera improvements and much faster compute on everything alone are worth it (I took the photo above while testing the camera).
The quality of the photos that come out of this thing is really impressive, and the low-light and HDR improvements are especially notable vs the X, which was considered to have a really good camera and fast SoC at the time (and is still plenty good — my wife will inherit it to replace her iPhone 7).
It’s still impressive to me (I work hard at not fully losing that sense of wonder I had at technology when I was 12) that in my pocket is an internet-connected super-computer with many orders of magnitude more RAM than my first computer — a 386 DX 25mhz — had in total storage (around 200 megabytes, I think), and that every time I take a photo, there’s a system on a chip with 11.8 billion transistors etched at 5 nanometers that fire up, with 16 machine learning inference cores working to make things look better and compensate for the physics of the relatively small lens and sensor (we’re in the second inning of the era of computational photography — I expect amazing things to keep happening on that front).
big.LITTLE Core CPU Design
Speaking of Apple’s A14, one of the things that I like about that CPUs design is the heterogeneous cores, like ARM’s big.LITTLE design (this isn’t new for the A14, Apple has been doing it since the A11 and ARM’s had Cortex designs that support it for years).
This is a way to get more energy efficiency out of a system-on-a-chip (SoC). The simple way to explain it is that it’s harder to have one core that is both optimized for low power and for the fastest speed.
There are no designs without trade-offs, so if you have the transistor budget, it’s easier to have one of each, and switch between them as needed. For example, if you’re just listening to a podcast and doing things that don’t require much compute, it may all happen on one or two low-power cores. But if you launch an app or go play a game, it’ll switch to one or more high-power cores for maximum performance.
So you could have a toaster oven that is a pretty good toaster and oven, but sometimes, you need both a real toaster and a real oven to get the best of both worlds. And since we’re not really cranking clock speeds that much anymore, chips have been sprawling “horizontally” and gaining more cores. May as well mix and match different types of cores so that there’s at least one available for each type of situation.
More Thoughts on Apple’s A14 Chip
Also about the A14: it's interesting that the performance jump over the A13 wasn't as high as might have been expected (it’s 17-18% faster than the A13 — which isn’t bad!)
If you’re curious, both the A13 and A14 are 6-core designs, with 2 high-performance cores and 4-low power cores. The A13 has 8.5 billion transistors total at 7nm and the A14 has 11.8bn at 5nm. The performance cores are clocked at 2.66Ghz on the A13 and 3.1Ghz on the A14.
At first I thought the smaller than usual YoY improvement was because both the A12 and A13 were such speed monsters that another similar jump couldn't be expected every year (nothing goes up in a straight line) and that since these older chips were still among the fastest in the industry, all this time later, Apple didn't feel the pressure to pull another rabbit out of the hat.
But thinking about it some more and looking at the die area reserved for the ML acceleration and "neural engine" cores, it seems like they decided to spend a lot of the transistors budget on that.
This may not show up on benchmarks, but it’s definitely making the product faster and better.
For example, there's probably a bunch of computational photography stuff that makes for better photos and faster processing that may not show in regular benchmarks, but that further differentiate the performance of the A14 from the competition.
Industry benchmarks don't measure these machine learning tasks on consumer phone SoCs because they aren't standardized, but it doesn't mean there isn't a crapload of operations taking place in the A14 that are contributing in noticeable ways to the user experience, possibly in more valuable ways than if the same die area had been spend on just faster speed on non-ML processes that would’ve resulted in nicer benchmark scores.
In other words, let’s be careful not to forget why we have benchmarks. They’re proxies for performance. But performance that isn’t captured by benchmarks is still real, and what matters most in the end is the user experience, not the benchmark score. Optimize for the right target.
The Headline vs The Actual Story
I thought there were a couple good recent examples of how sometimes we have to be careful to check beyond headlines…
First, there was a study by Boeing, the Department of Defense and United Airlines about the risk of SARS-CoV-2 (people call the virus COVID-19, but that’s the disease, the virus itself is SARS-CoV-2) in planes:
Headline:
Risk of COVID-19 exposure on planes 'virtually nonexistent' when masked, study shows
Someone took a closer look at the study and found what they actually did:
If you're wondering what factors Boeing and United smoothed over in their study to get this result, here they are:
-It studied 767s and 777s, not the planes you actually fly on.
-It assumes everyone sits facing forward at all times without talking or turning their heads.
-It assumes everyone wears a mask at all times, even though you eat and drink on planes.
-It only accounts for aerosols, not large droplets from sneezing or talking.
-It does study surfaces...but again, only aerosols from masked people, not direct spit droplets.
-It assumes every air vent is fully open at all times.
-It doesn't account for people getting on and off the plane or moving through the cabin.
-It doesn't account for the risk of being at the airport.
Another one where it was important to go beyond the Reuters headline:
AstraZeneca COVID-19 vaccine trial Brazil volunteer dies, trial to continue
Important detail: “According to O Globo, the man was in the trial's control group and was given a placebo instead of the experimental vaccine.”
So while there’s certainly something that went wrong and there should be an investigation, the vaccine itself isn’t in cause, despite all the usual suspects jumping on the story to stoke their confirmation bias…
The Arts & History
A Walk in the Woods (Shorter than Bill Bryson’s)
So a few days ago, I’m walking in the forest, and some wood chips fall at my feet. I look up and there’s this guy:
So I keep going and decide to follow a trail that I’ve never walked before:
Annihilation (2018)
Speaking of bears in the woods… (now that’s a smooth segue)
If you’ve seen the film and enjoyed it, I think you may like this video that goes into some of the metaphors in the film, and tries to not only focus on literal plot points.
If you haven’t seen the film, this isn’t for you, as it’s full of spoilers.
It’s a film that was quite an experience for me. I saw it in theater at a late viewing, alone, in basically an empty room. A friend was supposed to come with me, but his wife had their second daughter that night… I went in knowing nothing about it. I don’t normally enjoy scary movies, and rarely see them, so my tolerance is very low, unlike horror fans.
So it was quite the experience, and I think I was able to see it as it was intended, more as a sensory/emotional experience than as a bunch of plot points that may not all make literal sense.
Anyway, I have fond memories of it, but I think it’s not for everyone, and you have to be in the right frame of mind to get the most out of it.
I suggest watching it at night, in the dark, with the sound turned up (or headphones), and to just accept that it’s not “realistic”, whatever that means in this context.
James Randi has Died
Sad news. In honor of the man, I recommend watching the 2014 documentary about him, ‘An Honest Liar’.