Tech Brew Ride Home - Fri. 01/31 – Zuck’s Complaints About Leaks… Leaks

Episode Date: January 31, 2025

Apple’s earnings are decidedly mixed. Mark Zuckerberg’s complaints about everything he says leaking… leaked, and there are a lot of interesting details in those leaks. Looks like the SoftBank and OpenAI courtship is a serious one, and if the marriage happens, looks like Arm Holdings can be leveraged for the dowery. And, of course, the Weekend Longreads Suggestions. Links: Apple shares rise 3% as boost in services revenue overshadows iPhone miss (CNBC) Mark Zuckerberg said that Meta missed TikTok's rise because it didn't seem 'social' enough, leaked recording reveals (Business Insider) OpenAI in Talks for Huge Investment Round Valuing It at Up to $300 Billion (WSJ) OpenAI’s Sam Altman and SoftBank’s Masayoshi Son Are AI’s New Power Couple (WSJ) Weekend Longreads Suggestions: DeepSeek Chief’s Journey From Math Geek to Global Disruptor (WSJ) Chinese Quant Whiz Built DeepSeek In The Shadow Of a Hedge Fund Rout (Bloomberg) DeepSeek Debates: Chinese Leadership On Cost, True Training Cost, Closed Model Margin Impacts (SemiAnalysis) 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 Tech meme right home for Friday, January 31st, 2025. I'm Brian McCullough today. Apple's earnings are decidedly mixed. Mark Zuckerberg's complaints about everything he says leaking, leaked, and there are a lot of interesting details in those leaks. Looks like the SoftBank and Open AI courtship is a

serious one, and if the marriage happens, looks like arm holdings can be leveraged for the dowry. And of course, the weekend long-read suggestions. Here's what you missed today in the world of tech. Apple reported earnings last evening, and this is basically the tech. textbook definition of mixed. Yes, they had record revenue in their Q1 up 4% year-on-year, but iPhone revenue was down 1% and came in below estimates. Mac and iPad sales were both up 15%, but the wearables and accessories category was down 2%. Are you sensing a trend here? Tim Cook took time to point out that there are now more than 2.35 billion active Apple devices worldwide, which is a new all-time record. Amazing, right? Yeah, but,

But sales of Apple devices in China were down 11% to $18.5 billion, the largest drop in China since Q1 of 2024, quoting CNBC. Apple said it expected growth in the March quarter of low-to-mid single digits on an annual basis. The company also said it expected low double-digits growth for its services division. Apple's iPhone miss versus estimates was the biggest for the company in two years since its first quarter earnings report in fiscal 2023. At the time, Apple said its' miss. was because it was unable to make enough iPhone 14 models because of production issues in China, end quote. Yes, but Apple's services revenue reached an all-time high of $26.34 billion in Q1, up 14% year-on-year, beating estimates and helping lift the company's gross margin to a record 46.9%.

Quoting CNBC again, CEO Tim Cook's emphasis on services has transformed Wall Street's view of a company that's been defined over the decades by its iconic devices. For many years in the iPhone era, Apple's gross margin would predictably come in at between 38 and 39 percent, reflecting the company's tight grip over its supply chain and its pricing power in the market. But with iPhone growth slowing in recent years, Apple's move into services has changed the equation. The company hit a 40 percent gross margin in 2021 and has continued to expand it. Because of Wall Street's love of profit, Apple's been able to keep delivering for investors. The stock rose 31 percent last year, outperforming the NASDAQ,

and the company's market cap has climbed to $3.6 trillion, end quote. So to sum up, slightly disappointing, no iPhone super cycle of upgrades, at least not yet, and whether the glass is half empty or half full, basically depends on which part of the Apple you are focusing on. Yesterday in a meta-all-hands meeting, Mark Zuckerberg said, and I quote, Everything I say leaks, and it sucks, right? How do I know he said that? Well, because details of the meeting leaked, and there were lots of it. interesting bits in this meeting. For example, quoting Business Insider. When I look back on TikTok,

I think part of the reason why we were slow to it is because we didn't think TikTok was social. Zuckerberg said in a recording of an all-hands meeting obtained by Business Insider, we looked at it and we thought, oh, this is like a little more like YouTube. The admission came in response to an employee's question about whether Meta's current focus on artificial intelligence might cause the company to miss the next major social media trend as it did with TikTok. Zuckerberg explained at the meeting that Meta's traditional view of social interaction centered around friends posting content and commenting, caused the company to initially misread TikTok's appeal. The company failed to recognize how users were sharing TikTok content through private messages,

which has become a crucial form of social interaction across meta's platforms. Because we were too dismissive up front, it wasn't just about people commenting in the feed. It was about people seeing stuff in their feed and then sharing it into message threads, Zuckerberg said, referring to the company's instant messaging platforms, WhatsApp Messenger, and direct messaging in Instagram, where, quote, the majority of social interaction is happening, end quote. Also quoting a different business insider piece. Met a CEO Mark Zuckerberg told employees Thursday in a company all-hands meeting to buckle up for an intense year ahead and address several recent policy changes.

Zuckerberg opened the all-hands by emphasizing a sense of urgency for the year. He told staff that he expected to have a clearer sense of the company's trajectory by the end of 2025, and that AI would be top of mind. He also addressed recent policy changes related to fact-checking and programs for a diverse, equity and inclusion. This is a marathon, not a sprint, he said, and a recording reviewed by Business Insider, but honestly, this year feels a little more like a sprint to me. In a wide-ranging opening monologue, Zuckerberg predicted that 2025 would be the year a highly intelligent and personalized digital assistant reached one billion users. I think whoever gets there first is going

to have a long-term durable advantage towards building one of the most important products in history, Zuckerberg said, according to the recording. Zuckerberg also reiterated his belief that this would be the year META started seeing AI agents take on work, including writing software. Asked whether this would lead to job cuts. Zuckerberg said it was hard to know, and that while it may lead to some roles becoming redundant, it could lead to hiring more engineers who can harness artificial intelligence to be more productive. The nature of what engineering is in the future will be different than it is today, he said. Zuckerberg touched on several flashpoints in recent weeks from inside the company,

including the announcement that it would move away from third-party fact-checkers to a community notes system like that used by Elon Musk's X. He told staff to wait and see how the new system would be implemented. I'm actually quite optimistic that this is going to end up being a better system, he said, end quote. One more interesting data point that came out of this. Zuck said meta sold more than one million Rayban smart glasses in 2024, revealing those sales figures for the first time. More details on that potential marriage between OpenAI and SoftBank.

Sources say OpenAI is in early talks to raise up to four. billion at a $340 billion valuation. Open AI was last valued at $157 billion in October when it raised $6.6 billion, quoting the journal. SoftBank would lead the round and is in discussions to invest between $15 and $25 billion. The remaining amount would come from other investors. The two companies were recently in talks to value OpenAI as high as $340 billion, one of the people familiar with the matter said. After the Wall Street Journal published that figure in an earlier version of this story, the person said newer negotiations lowered the proposed valuation to as much as $300 billion. The Japanese company is helping assemble investors for the rest of the round, one of the people said.

The discussions are still in flux and could fall apart, the person said. The $300 billion valuation would include the cash OpenAI raises in the round. The funding would be used in part to help OpenAI fulfill its roughly $18 billion commitment to Stargate, a joint venture with SoftBank and others to finance the construction of new data centers in the U.S. powering OpenAI's technology. The startup also expects to use the cash to fund its money-losing business operations. At $300 billion, OpenAI would be the second most valuable startup in the world behind only Elon Musk's SpaceX, according to the data provider CB Insights, end quote. Meanwhile, where is SoftBank going to get the money to do this?

Well, a different journal article says they might borrow against there more than $140 billion arm stake to fund the investments. Quote, Masa Sond's commitment to Open AI and Stargate ensures that Sam Altman's company will have ample cloud computing firepower in the coming years, just as it has ended the exclusivity portion of its Microsoft deal. And by leading a funding round that would be the biggest in Silicon Valley history, Sond is enthusiastically endorsing Altman's plan to keep spending gobs of cash on leading edge AI systems. For all of us, the AI era represents a once-in-a-generation opportunity to help build a better, safer, healthier, and more prosperous

future, Sahn wrote in a widely distributed email message Thursday. It was the $32 billion purchase in 2016 of Arm Holdings, a chip design company that became SoftBank's Golden Goose. The company surged in the halo of the AI rush after SoftBank relisted it on the NASDAQ in 2023. Arm stock has tripled since then, and SoftBank's stake is worth more than $140 billion. an asset that gives son plenty of wiggle room financially. Borrowing against its arm's stake would be one way for SoftBank, which had around $30 billion of cash at the end of September, and has vowed to keep a large buffer on hand to fund OpenAI. It could also sell some assets, such as its stakes in T-Mobile and Deutsche Telecom, worth a combined $27 billion, according to fact set.

SoftBank has held early talks with potential lenders to help fund its investments in OpenAI and Stargate people familiar with the discussion said, end quote. Time for the weekend long read suggestions, and look, this was the week of Deepseek, right? So, two different profiles of Deep Seek founder Liang Wen Fang, a math geek who in 2015 founded hedge fund High Flyer, which somehow Deep Seek was incubated inside of. This first one is from the journal, quote, some call him China's Sam Altman. Others compare him to Jim Simmons, the pioneer of quantitative investing. Liang Wen Feng shares a lot with both innovators, but his impact might end up being as great. Born in 1985, Liang grew up in Zhang Jiang, one of the port cities on the southern Chinese coast that have traded with the world for centuries. He was a straight-A student in school and began studying calculus on his own in junior high school before going on to China's prestigious Jey Jiang University. As a student, Liang started writing AI algorithms to pick stocks. In 2013, a few years after graduation,

he started Jacoby, an investment firm named after German mathematician Carl Jacobi. In 2015, he founded High Flyer with two college friends. It was a pioneer in China in applying newer AI techniques so computers could absorb many types of data, not just market prices, and tease out significant patterns. Unlike high-frequency traders who try to get a jump of a few milliseconds on others, Liang and his colleagues have focused on medium-frequency training where positions can be held longer. Liang drew inspiration from Simmons, a mathematician and pioneer in quantitative finance who founded the Long Island Hedge Fund Renaissance Technologies and was using machine learning techniques in the 1980s. Liang wrote the introduction to the Chinese version of

The Man Who Solved the Market, a book about Simmons and his team. Whenever I encounter difficulties at work, I recall Simmons's words. There must be a way to model prices, Lange. wrote. Over the past five years, at least five funds High Flyer managed produced average excess returns of more than 20 percent compared with market benchmarks, according to financial data provider Simu Pi Pi Wang. But there have been bumpy patches too. In 2021, High Flyer apologized to its investors for poor performance, saying it misread which sectors of the market would perform well. Another rough spot occurred last year when China's financial regulator clamped down on quants, accusing them of herd behavior that exacerbated volatility. Richard Dewey, a technology

executive and former hedge fund manager, said Deep Seeks AI development strategy of doing more with less, a smaller team, less money, and fewer chips reminded him of quant traders. They seem to have focused on ringing a lot of signal out of a relatively small amount of data, which is similar in spirit to what's required in quantitative training, Dewey said. Like some other quants, Leang prefers to be thought of as an engineer rather than a trader, according to people close to him. In 2019, Liang's team started building computing systems with Nvidia graphics processing units by late 2022 when OpenAI released ChatGBTBT. Only a few Chinese companies had more than 10,000 Nvidia high-end chips in hand,

and High Flyer was among them. It's like buying a piano, Liang told Chinese tech publication 36KR in 2023, talking about the chip purchases. Firstly, it's because you can afford it. And secondly, it's because you have a group of people who are eager to play music on it. One of Liang's biggest decisions was to make his code open source, meaning anyone can access it. He said he wanted DeepSeek to break the monopoly of big tech companies, end quote. And this is Business Week's profile, quote.

DeepSeek's research was funded by high flyers' R&D budget, Liang said previously. It drew computing resources from the Quant Fund, which had amassed 10,000 Nvidia GPs in 2021, prior to U.S. bans on exports of sophisticated Nvidia chips and other graphics processing units. Liang recruited engineering talent almost exclusively from China. Many were fresh out of top universities, interns in their final leg of doctoral studies, and Olympiad medal holders. Wang said Liang ran many experiments on his own and Deep Seek operated much like a research lab. It started small, but as they got real progress, they started to get excited, he said. The startup began periodically releasing models, seemingly impervious to even stirred up by the U.S. ban on exports of cutting edge

AI accelerator chips. Deep Seek released its R-1 advanced AI reasoning model on January 20th, the same day Donald Trump was sworn in as America's 47th president. Earlier that Monday, Liang attended a closed-door business symposium in Beijing that was hosted by Chinese premier Li Kieng. Their experts in technology, science, education, and other fields offered their opinions and suggestions for a draft government work report, according to the official news agency. Video footage on YouTube shows Liang sitting across the table from Li and speaking with the Chinese leader nodding attentively. Liang is assumed to own 51% of High Flyer. That would give him a stake worth $71 million based on a comparative analysis, according to the Bloomberg Billionaire's Index. If DeepSeek reaches the same potential as Open AI valued at roughly $150 billion, the founder could potentially be in line for a massive windfall.

Some have questioned whether Liang's Deep Seek is as promising as it appears. Shortcomings include the startup's infrastructure's ability to handle global traffic waiting to try its service, or the app's handling of sensitive subjects such as the 1989 protests in Tiananmen Square and queries on Chinese leader Xi Jinping. Experts have also questioned the assumption that DeepSeek was building with 10,000 A-100 Nvidia chips, with analysts like Dylan Patel speculating that Deepseek needs at least 50,000 of Nvidia's far more powerful chips, the H-100s. Meta platforms, for instance, operates the equivalent of 600,000 Nvidia H-100s, end quote. Yeah, on that, and finally this week, the great site semi-analysis has a deeply detailed look at Deepseek and the contours of their operation, which, again, is estimated to have access to around 500,000 Hopper GPUs, a total server capex of $1.3 billion, and a GPU spend of more than $500 million over its history. quote, Deepseek took the world by storm. For the last week, Deepseek has been the only topic that anyone in the world wants to talk about. As it currently stands, Deep Seek daily traffic is now much higher than Claude, perplexity, and even Gemini. But to close watchers of the space, this is not exactly new news. We have been talking about Deep Seek for months. Each link is an example. The company is not new, but the obsessive hype is. Semmy analysis has long maintained that Deepseek is extremely talented and the broader public in the United States has not cared.

When the world finally paid attention, it did so in an obsessive hype that doesn't reflect reality. We want to highlight that the narrative has flipped from last month when scaling laws were broken. We dispelled this myth. Now algorithmic improvement is too fast, and this is too somehow bad for Nvidia and GPUs. The narrative now is that DeepSeek is so efficient that we don't need more compute, and everything has now massive overcapacity because of the model changes. While Jevins Paradox 2 is overhyped, Jevins is closer to reality. The models have already induced demand with tangible effects to H100 and H200 pricing. We believe the most interesting implications is specifically on margins, and what that means for the entire ecosystem.

Below, we have a view of the future pricing structure of the entire AI industry, and we detail why we think Deepseek is subsidizing price, as well as why we see early signs that Jevin's Paradox is carrying the day. We comment on the implications on export controls, how the CCP might react with added deep seek dominance and more, end quote. As I say, it's in depth, it's technical, it's worth reading. This weekend I've got a bonus episode that looks at something we don't talk about all that much. You, you know, you create a startup, you enter a market, you attack a problem, all with a set of assumptions set in a specific period of time. And then

the world changes. Maybe there's a global recession. Maybe there's a global pandemic. Maybe there's a, I don't know, meteor strike. Or, as in the case of the company we're profiling this weekend, there is a change in political administrations, and suddenly the market you were targeting might have to face headwinds you didn't need to even worry about a year or so ago. It's an interesting look at, as I say, an interesting startup problem that doesn't get discussed very often. Listen to that. Talk to you on Monday.