Tech Brew Ride Home - Mon. 04/15 – Onshoring And Halving

Episode Date: April 15, 2024

Really looks like high-end re-shoring of chip tech is happening, and happening soon. What if satellite telephony becomes a table-stakes smartphone feature? Get ready for the bitcoin halving. Does anyone have invites to Airchat? And get ready for 4TB, yes, TB, SD cards. Sponsors: Shopify.com/ride ArcticWolf.com/techmeme Links: Samsung boosts Joe Biden’s chipmaking ambitions with Texas plant upgrade (Financial Times) Exclusive: Google Pixel 9 series to get emergency satellite connectivity, new modem (Android Authority) Bitcoin ‘Halving’ Will Deal a $10 Billion Blow to Crypto Miners (Bloomberg) ChatGPT essay cheats are a menace to us all (Financial Times) Naval Ravikant’s Airchat is a social app built around talk, not text (TechCrunch) 4TB SD cards are arriving in 2025 for your cameras and laptops (The Verge) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the TechMean right home for Monday, April 15th, 2024. I'm Brian McCullough today. Really looks like high-end reshoring of chip tech is happening and happening soon. What if satellite telephony becomes a table stakes smartphone feature? Get ready for the Bitcoin halving.

Does anyone have invites to air chat and get ready for 4 terabyte? Yes, 4 terabyte SD cards. Here's what you miss today in the world of tech. The U.S. has awarded Samsung $6.4 billion in Chips Act funding, saying the company plans to make two nanometer chips at its Taylor, Texas plant by 2026, two years ahead of when TSM is going to roll up a similar plant, two nanometer chips. So the cutting edge top of the line manufactured here in the U.S. in just a couple of years. This is on-shoring we can get behind, I think, quoting the Financial Times. U.S. Commerce Secretary Gina Raimondo announced on Monday that the South Korean chipmaker was set to produce chips at the two nanometer level of miniaturization at a new fabrication plant or fab it was building in the city of Taylor in Texas. It will be part of a $40 billion investment made by the company in capabilities ranging from the manufacturing of microprocessors to advanced chip packaging and research and development work.

Samsung's investment will be topped up with as much as $6.4 billion in direct funding received under the U.S. chips and science. Act, a week after the U.S. government announced that TSM would receive up to $6.6 billion from the flagship subsidy program to support its chip expansion in Arizona. The first of Samsung's new Taylor fabs will start making two nanometer chips in 2026, according to a senior U.S. official. TSM is set to produce two nanometer chips at an Arizona facility from 2008. An initial $17 billion investment, part of a planned $40 billion in total was announced by Samsung in 2021 to build its first fab and Taylor, which will now make two and four nanometer chips. The new capital spending adds a second fab that will also make two nanometer and four nanometer chips as well as the

construction of an advanced chip packaging for 2.5D packaging of processor and memory chips. Advanced packaging is a crucial stage in the production of artificial intelligence chips such as NVIDIA's H-100, which is used to train generative AI systems such as OpenAIs chat GPT. TSMC, the world's biggest contract chipmaker, carries out 2.5D packaging for Nvidia's most powerful chips, but it currently has no plans for such an advanced packaging facility in the U.S., end quote. It would be interesting if this became standard table stakes for flagship smartphones. Sources tell Android authority that Google's new Pixel 9 series, Next Generation Pixel Fold,

and an underdevelopment 5G tablet will use a new Samsung modem that has satellite connectivity. Quote, the TensorFlow G4 coming first to the Pixel 9 series, is bringing a brand new modem, the Samsung modem 5400. While it's not a Qualcomm modem like some would prefer, it's still an upgrade over the previous generations. Our source couldn't give any exact specs, but as always, we can expect it to be faster and more power efficient than the previous generations. Additionally, the software stack has been upgraded again, which should improve stability. Finally, there's support for 3GPP-REL-17, which brings support for 5G non-terrestrial networks, or NTN, satellite-based communication in other words.

The modem is coming to all cellular-enabled tensor for devices, including the Pixel-9 series and the next-generation Pixel Fold. Our source has also informed us that Google is in the early stages of developing a 5G tablet codenamed Clementine with the same modem. However, it is currently unknown when and if the device will make it to market. As we previously reported, Google has made significant progress in integrating support for satellite-based messaging in Android 15. Thanks to NTN support in the Exinos Modem 5400, the Pixel 9 series will be the first to support

Android's native satellite implementation. Initially, the feature will be provided by T-Mobile, built-in collaboration with SpaceX, with other providers possibly added in the future. The satellite link will allow for texting but not calling. Additionally, a special satellite gateway app will allow for easy communications with emergency services via a feature called emergency SOS. We've already seen earlier indications about how this could work, but we can now preview it in more detail. First, you will be asked a few basic questions to narrow down the

situation. You will also be given a choice to notify your emergency contacts, after which you will be able to message with emergency services to get you the help you need. Many of the questions will also have predefined answers to choose from to make the process quicker. Due to the nature of satellite connectivity, mobile solutions using it typically need very specific alignment to connect. Google has designed animations to help this process, end quote. Something to put on your radar. The next Bitcoin halving, halving, is right around the corner. This will cut miners validating transactions to 450 from 900 now and costing the industry around $10 billion in annual revenue. And it's expected to happen April 20th, because of course it is, 420. Quoting Bloomberg,

for enthusiasts of Bitcoin, a once every four years software update called the halving has long been held as one of the keys to propping up its value. Around April 20th, the solving will cut the amount of Bitcoin that miners can earn each day for validating transactions to 450 from 900 now. Based on Bitcoin's current price, it could spell revenue losses of around $10 billion a year for the industry as a whole. Marathon Digital Holdings, Clean Spark, and other miners, which compete for a fixed Bitcoin reward by solving mathematical puzzles using super-fast computers, have invested in new equipment and sought to buy smaller rivals in an attempt to cushion the drop-off in revenue. This is the final push for miners to squeeze out as much

revenue as they can before their production takes a big hit, said Matthew Kimmel, a digital asset analyst at coin shares. With revenues across the board decreasing overnight, the strategic response of each miner and how they adapt could well determine who comes out ahead and who gets left behind, end quote. Granted, Bitcoin has reached new highs after previous halvings, helping to mitigate the periodic drop in mining rewards and the increase in the cost of doing business. The event this month is coming after the digital currency has more than quadrupled since November 2020. Yet, the margin of success for the industry keeps getting finer. Miners will need to continually spend more money in a never-ending technological arms race for smaller rewards. And while the energy-intensive

validation process has always made mining expensive, companies now face even more competition for power from the burgeoning and deep-pocketed artificial intelligence industry. The soaring price of Bitcoin has helped offset those power costs and fueled growth in crypto mining. Since the first specialized machines came into play in 2013, the aggregate market cap of 14 U.S. listed miners has grown to about $20 billion, according to an April 1st report by J.P. Morgan Chase. While U.S. listed miners are the face of the industry, they account for just about 20% of the sector's computing power, according to crypto researcher, the miner mag. Private miners make up the rest and could be more vulnerable after the having, as they typically must tap debt financing or venture capital to cover

their needs, whereas public companies can raise funds via share sales. As the hoopla has revved up around the event, some traders are betting that mining stocks will fall total short interest. the dollar value of the shares borrowed and sold by bearish traders stood at about $2 billion as of April 11th, according to an estimate from S3 partners. That short interest accounted for almost 15% of the group's outstanding shares three times more than the U.S. average of 4.75%. The update, the fourth since 2012, was programmed by the anonymous Bitcoin creator Satoshi Nakamoto to maintain the hard cap of 21 million tokens to keep it from becoming inflationary as a currency. The situation differs from four years ago when Bitcoin was trading under $9,000,

and most of the mining activity took place in China. Since then, much of that activity has shifted to the U.S. driving competition for electricity, end quote. I'm not saying this to jinx anything, but the lowest hanging fruit of the feared AI apocalypse still hasn't quite been picked yet. Yes, AI fakery is making its way around the various nooks and crannies of the internet, but it's not the overwhelming deluge that we might have feared. Again, at least, not yet. But also, remember that first AI panic around cheating in academic settings? Again, it's not been the step change that some were fearing. An analysis of more than 200 million student papers submitted over the last year shows only around 11% of them had signs of AI help in them,

and only around 3% of papers contained at least 80% AI writing, quoting the Financial Times. In the U.S., Stanford University researchers said last year that cheating rates did not appear to have been affected by AI. Up to 70% of high school students have long confessed to some form of cheating, and nearly a year after Chat Chept's arrival, that proportion has not changed. At university's research shows half of students are regular generative AI users, not necessarily to cheat, but only about 12% use it daily. When it comes to the number of student essays written with the help of AI, rates appear relatively steady, says, turn it in, a plagiarism detection software group that has a tool for checking generative AI use. It's said that students have submitted more

than 22 million papers in the past 12 months that do show signs of AI help, which was 11% of the total it reviewed. More than 6 million papers or 3% of the total contained at least 80% of AI writing. That is a lot of papers, but the percentage of AI writing is virtually the same as what Turnitin found last year when it carried out a similar assessment. AI usage rates have been stable, says Chris Karen, Turnitin's chief executive, and as he told me last week, just because you're using ChatGPD does not necessarily mean you are cheating. Some teachers and faculty allows some level of AI assistance in writing an essay, but they also want that properly cited, he says, AI can be incredibly useful for doing research and brainstorming ideas, end quote.

It's that time again, time when I ask if anybody has invites to a new social network. Well, not new exactly, but Naval Robicant's air chat apps sort of blew up late last week, so if you have an invite, please let me know. Brian at Techmeme.com. Here's the deal with AirChat, according to TechCrunch. A previous version of AirChat was released last year, but the team, led by Angelus founder Naval Ravacant and former Tinder product exec, Brian Norgaard, rebuilt the app and relaunched it on iOS and Android yesterday. Currently, invite only. AirChat is already ranked number 27 in social networking on Apple's App Store. Visually, AirChat should feel pretty familiar and intuitive

with the ability to follow other users, scroll through a feed of posts, then reply to like and share those posts. The difference is that the posts and replies are audio recordings which the app then transcribes. When you open air chat messages automatically start playing and you quickly cycle through them by swiping down and up. If you're so inclined, you can actually pause the audio and just read text. Users can also share photos and video, but audio seems to be what everyone's focused on and what Ravikant describes as transforming the dynamic compared to text-based social apps. After joining AirChat this morning, most of the posts I saw were about the app itself with Ravikant and Norgaard answering questions and soliciting feedback. Humans are all meant to get along

with other humans. It just requires the natural voice, Rabicant said. Online text-only media has given us this delusion that people can't get along, but actually everybody can get along. This isn't the first time tech startups have bet on voice as the next big social media thing, but airchats, asynchronous, threaded posts make for a pretty different experience than the live chat rooms that briefly flourished on Clubhouse and Twitter spaces. Norgaard argued that this approach removes the stage fright barrier to participating because, quote, you can take as many passes at composing a message on here as he like and nobody knows. In fact, he said that in conversations with early users, the team found that, quote, most of the people using air chat today are very

introverted and shy. Personally, I haven't convinced myself to post anything yet. I was more interested in seeing how others were using the app. Plus, I have a love-hate relationship with the sound of my voice. Still, there's something to be said for hearing Ravikant and Norgard explain their vision, rather than just reading the transcriptions, which can miss nuances of enthusiasm, intonation, etc. And I'm especially curious to see how deadpan jokes and shit posting translate or don't into audio. I also struggle a bit with the speed. The app defaults to 2X audio playback, which I thought sounded unnatural, particularly if the whole idea is fostering human connection. You can reset the speed by holding down the pause button, but at 1X, I noticed that

I'd start skimming when listening to longer posts, then I'd usually skip ahead before listening to the full audio, but maybe that's fine, end quote. Again, any invites, Brian at Techmeme.com or at Brian MCC on Twitter. Finally today, long-time listeners know I have a fascination with storage. I remember when I got my first computer with a 100-gibite hard drive and how spacious and capacious that felt. But I also remember how my first MP3 player could hold but one album at a time, only one. I load Primal Scream's Exterminator album and go for a run, and at Track 11, this device was out of storage, so it was either hit repeat or stop running. I usually stopped running. So here we go. Quoting the Verge. It wasn't that long ago when we were impressed by a

400-gigabyte micro-SD card, but now Western Digital is showing off a full-size SD card with 10 times the storage capacity. The new Sandusk Extreme Pro 4-Torabyte SDx. card is designed for cameras and laptops, and it's the first time we've seen the storage size on an SD card. While Western Digital is previewing its 4-terabyte SD card at the NAB show this week, it won't actually be available until 2025. Anantec notes that the card uses the ultra-high-speed I-UH-S-I interface, which supports up to 104 MBS data transfers in the UHS 104 mode. It also has minimal sequential right speeds of 30 MBS, enough for 8K video recording. The SD Association first announced its secure digital ultra-capacity or SDUC card specifications in 2018, promising that SD cards would soon hold up to

128 terabytes of storage. It's taken quite some time to get to the 4-terabyte point, and Western Digital isn't disclosing what type of NAND is inside the upcoming card, so even higher capacities are likely a ways off. Still, if you're eager to record beyond 4,000, 4K resolution with 4 terabytes of space, then Western Digital should have something ready next year. The big question is how much a 4-terabyte UHS-1 SD card will cost? Given the 1-terabyte versions of the same card retail for around $140, you're probably looking at $600 or more for a 4-terabyte version, end quote. I think I've brought this up before, but I've fallen in love with the feature in Apple News Plus

where certain articles are turned into podcast episodes so you can listen to instead of read them. Can anyone listening to me right now put me in touch with the Apple News team? I mean, I do a daily news show that could be perfect for this section of Apple News Plus. We'd love to have at least the conversation. Brian at Techmame.com. Talk to you tomorrow.