Tech Stock Thoughts #3: Apple, Micron, Intel, Twitter, Amazon, STMicro
Apple’s Car Plans: Better Late Than Never…and Perhaps Strategically Vital
This might be less about competing against the Model 3/Model Y than about what cars could look like 10–15 years from now.
Sources: Bloomberg, Reuters, Reuters (again), CNBC
- If you’re looking for long-term context for Apple’s plans (according to multiple reports) to build an electric car that has some measure of autonomous driving capabilities, taking a look at Mercedes-Benz’s just-unveiled MBUX Hyperscreen might not be a bad place to start. The Hyperscreen is 56 inches wide, takes up a car’s entire dashboard, contains 12 actuators to provide haptic feedback, and relies on an 8-core CPU, Nvidia GPUs and 24GB of RAM. Needless to say, it can do more than just play music and provide turn-by-turn directions.
2. Next, take a look at the tricked-out BMW X5 — replete with a large-screen TV in the back, projectors beaming content onto windows and cinema-like audio and lighting — that Intel and Warner Bros. demoed at CES two years ago, as a way of showing the kinds of entertainment experiences a car could potentially deliver if it never needs to be driven by a human (and thus has no need for a steering wheel/pedals, a traditional front-seat/back-seat arrangement or constant side visibility). Not the kind of thing that’ll be on sale at dealerships anytime soon, but perhaps a glimpse of what could be possible a decade or so from now.
3. Finally, consider what kinds of answers Elon Musk and his crew might be providing themselves to this question.
4. In terms of computing power and communications features, the typical $35,000+ car sold today is a lot different than the typical $35,000+ car sold 5-to-7 years ago. And it’s a safe bet that $35,000+ cars sold 5-to-7 years from now will generally pack a lot more computing power (and in the process, support more advanced entertainment and productivity experiences) than ones sold today. Apple is undoubtedly well-aware of this.
5. Apple is also probably well-aware of how (even though it’s set to be commercialized in a very gradual fashion) autonomous driving stands to both accelerate this trend in multiple ways. First, by creating another workload that requires a lot of computing power and advanced sensors/software; second, by freeing up time for vehicle owners to do things besides drive; and third (as shown in the Intel/Warner Bros. demo, as well as some others), by opening the door for cars to be transformed over the long run into workstations/home theater systems on wheels. Given all of this, participating in the EV market is arguably strategically vital for a company that at heart is a builder of end-to-end consumer computing platforms.
6. As others have noted, Apple has quite a few competencies (hardware design, chip and hardware engineering, UIs, software, services, etc.) that could help it differentiate an electric car. The company has also had an autonomous driving project running for several years, and (per Reuters) has developed range/density-boosting battery tech. And its AR efforts could help it create a heads-up display system for drivers (see Nvidia’s Drive IX system for some possibilities).
7. With Bloomberg and Reuters indicating Apple’s car will launch around the middle of the decade at the earliest, it’s hard not to think that it would’ve been better if Apple either kept its original Project Titan car effort going or just bought Tesla when that was still an option. I get why the former didn’t happen — a car project requires a ton of manufacturing/supply chain/R&D work, building out the necessary sales and support infrastructure would also require much effort, and a company as image-conscious as Apple has to be concerned about what the fallout would be if it launched an unreliable and/or accident-prone car. But given how mature the EV market will likely be by 2025, an earlier start would’ve been preferable.
8. On the other hand, autonomous driving will most likely still be in its infancy around 2025. While ADAS capabilities for mass-market cars will probably have improved considerably, cars sold to consumers will still require humans to often drive them, and their interiors aren’t going to look anything like the tricked-out self-driving concept cars we’ve seen. In that particular context, Apple might not be especially late to the show.
Micron’s Earnings: Mostly Solid
2021 is looking good for the DRAM industry.
Source: Micron (PR), Micron (earnings slides)
- Micron didn’t deliver a blowout earnings report, but it was a fairly strong one. They topped the hiked guidance ranges they shared in early December, and also issued February quarter guidance ranges that were fully above consensus, in spite of December disruptions at Taiwanese facilities (caused by a power outage and an earthquake) impacting its DRAM supply/costs and foundry supply constraints for logic ICs limiting how much hardware its customers are able to ship.
- Just as importantly, Micron noted it’s seeing DRAM pricing finally improve within several end-markets (fits with recent reports) and forecast “further tightening” for the market throughout 2021, with full-year supply growth trailing demand growth.
3. With DRAM accounting for 70% of Micron’s revenue and likely over 80% of its gross profit, point #2 matters far more than the relatively mixed outlook Micron gave for NAND, where prices are still declining but the company is hoping price elasticity and its not-so-subtle calls for rivals to keep their supply growth in check will help yield better second-half conditions.
4. One minor blemish: Micron slightly took down its 2021 DRAM bit demand guidance to high-teens growth from ~20% growth. However, the company attributed this to 2020 bit demand growth coming in better than previously expected, and (with corporate desktop and server/storage spend still pretty weak) Micron might be erring on the side of caution when it comes to expected enterprise hardware demand.
5. On the flip side, Micron was pretty enthusiastic about expected 2021 demand from smartphone OEMs and hyperscalers, while highlighting growth drivers such as DRAM/NAND content boosts within 5G phones, the memory needs of AI/ML workloads and Intel and AMD’s 2021 server CPU refreshes (notably, Intel’s Ice Lake CPUs will support 8 more memory channels per socket, 2 more than current-gen Cascade Lake CPUs and on par with AMD’s Rome and Milan CPUs).
6. After opening higher on Friday, Micron’s stock sold off and closed down 2.1%. But with Micron up 50%+ in the two months leading up to its report, this feels like a case of moderate profit-taking in response to a report that was good but didn’t meet the highest hopes of some bulls. And as I noted last Thursday, it’s still not hard to make a case that Micron’s stock isn’t all that expensive.
Intel is reportedly in talks with TSMC and Samsung about foundry deals — Given what Intel said on its July and October earnings calls (in the wake of its 7nm manufacturing setback) about outsourcing more, I would’ve been surprised if Intel wasn’t holding such talks. But there are still a few interesting details in Bloomberg’s report:
- Intel wants TSMC and/or Samsung to make “some of its best chips.” This aligns with one my 2021 tech predictions (#3).
- Any chips Intel obtains from TSMC through this effort “wouldn’t come to market until 2023 at the earliest and would be based on established manufacturing processes already in use by other TSMC customers.”
- TSMC is offering the use of its N4 (4nm) process, which is (just as TSMC’s N6 process is relative to its popular N7 process) meant to be an evolution of its recently-launched N5 process.
A 2023 ETA would likely put the arrival of any N4-based Intel CPUs well behind that of AMD’s first Zen 4-based PC and server CPUs, which are expected to be made using N5 and launch by early 2022. Chances are that a number of N5-based ARM PC and server CPUs will also have launched by then (indeed, Apple has already rolled one out).
It might also be telling that TSMC is offering the use of N4 rather than its N3 process, which is expected to reach volume production in 2H22 and will probably see big initial orders from Apple. With Intel still quite committed for now to its own leading-edge process development, and with TSMC having strategic relationships with the likes of Apple, AMD and Nvidia, TSMC’s mindset towards Intel might be “We’re willing to help you, but we won’t make you our top priority.”
Could Samsung, which is hungry to gain ground against TSMC, be more willing to roll out the red carpet? Intel has said it’ll provide more details about its foundry plans during its Jan. 21 earnings call.
Twitter bans Trump — I’ll leave the political commentary here to others. Instead, I’ll note that the ban comes less than 3 months after Twitter posted a pretty disappointing Q3 mDAU print in spite of the election and the return of sports. It’ll be very interesting to see what they have to say on their Q4 call (set for Feb. 9) about how mDAUs have trended following the Trump ban.
Also, it’s worth noting that Twitter was desperate enough to boost usage in Q4 that it backtracked on its decision to (mostly) stop showing the likes and follows of followed accounts within Timelines — a decision that it actually talked up in its Q3 shareholder letter (see page 7). As it is, the fact that Twitter has been so dependent on using such a crude approach to surfacing content from non-followed accounts — one that depresses activity by making users sensitive both to which tweets and whose tweets they’re seen as liking — speaks volumes about how (unlike Facebook, Instagram, YouTube and TikTok) Twitter has failed at using pure ML/algorithm-based approaches to driving content discovery on a large scale.
Jim Keller joins AI accelerator startup Tenstorrent as CTO — Keller is a chip industry legend whose resume includes stints at Apple, AMD (twice), Tesla and most recently Intel. Tenstorrent is one of a slew of prominent AI accelerator startups; others include Cerebras Systems, Groq and Graphcore, the last of which just raised $222M at a $2.77B post-money valuation.
From the looks of things, Tenstorrent is partly looking to differentiate its accelerators by trying to offer superior compute/power/bandwidth efficiency, as opposed to delivering unmatched horsepower. Also, like Intel’s Habana Labs (and Nvidia/Mellanox going forward), addressing networking bottlenecks is a priority — Tenstorrent’s next-gen Wormhole accelerator will pack a 16-port/100G Ethernet switch. Time will tell just how successful they are — Nvidia’s software stack/CUDA ecosystem remains a large moat, and there’s more to meeting the needs of AI developers than designing impressive silicon — but Keller’s involvement is arguably a fresh vote of confidence not only in Tenstorrent’s engineering work, but in the long-term market opportunity for AI accelerators.
As an aside: If you’re interested in the chip industry’s evolution, I highly recommend listening to this podcast Keller did about a year ago with Lex Fridman.
Facebook’s Jan. 2021 mobile gaming report — The report features data from a survey of 13,246 gamers in the U.S. and several other markets. Among the findings:
- Mobile gaming audiences have grown by a healthy double-digit percentage since March, and average weekly gaming time spent is also up over 10% in many markets.
- Gamers in most markets are nonetheless spending less on average than they did pre-COVID, with U.S. average monthly game spend down 28% to $17.
- New gamers are both playing and spending more on average than existing gamers, are more likely to play games from “core” rather than casual genres, and are also more partial to multiplayer/social experiences.
The survey provides additional reasons to think COVID will have a meaningful long-term effect on the gaming industry’s addressable market. Some of those who have taken up gaming since last March might stop playing as reopenings gain steam, but their average engagement and spending activity suggests a lot of them have found a new pastime that they won’t be fully giving up.
Amazon shutters Prime Pantry — File this under the “it’s not a bug, but a feature” category. When it launched in 2014, Prime Pantry shipped boxes containing up to 45 pounds of groceries (including various heavy/bulky items) for a $5.99 shipping fee. But as its warehouse/logistics infrastructure kept expanding and becoming more cost-efficient, Amazon gradually improved Prime Pantry’s terms. Most recently, it provided free delivery to Prime members on all $35+ Prime Pantry orders, while lowering the order minimum to $10 for those paying a $4.99/month fee.
Now, Amazon is simply folding Prime Pantry’s selection into the rest of its direct e-commerce listings. Browsing around a bit, it looks as if many of the items previously offered via Prime Pantry still feature $35 free-shipping order minimums, but also support free same-day delivery. Meanwhile, though scaling out the AmazonFresh grocery delivery service remains a work in progress, AmazonFresh and Whole Foods are the only local grocery delivery players to provide free deliveries.
STMicro pre-announces a strong Q4 — Expected Q4 revenue of $3.24B is soundly above a prior Q4 guidance midpoint of $2.99B and implies 18% annual growth. STMicro cited strength within its Personal Electronics segment, which counts Apple as a major customer, as well as improving demand for automotive products and microcontrollers.
Given all of the reports there are right now of strong iPhone demand and auto chip shortages, it makes plenty of sense that business is good for STMicro right now. It’s also probably good for many other suppliers of iPhone and/or auto chips (e.g., Qualcomm, Skyworks, Qorvo, NXP, TI, Microchip, ADI, Renesas).
