Investing in the Magnificent 7: Value Creation, Value Distribution and Expectation Gaps

Section summaries

1. 12:35Freddy's Path in Tech Stock Research

   Freddy introduces himself as a former US equity analyst at KGI Securities. He starts with internet company internships from university through graduate school, when he did a lot of research to understand digital advertising and streaming media. His later interview question was to choose between Airbnb and Uber and estimate financials and key metrics for a company that was not yet public, which foreshadowed his work over the next nearly 10 years. He then looks back at the evolution from FAANG to MAG-7: back then everyone was watching consumer internet companies, and even Microsoft was seen as a dead elephant.

   ▶ Watch from here

2. 16:20Industry Trend Shifts from Consumer Internet to Enterprise Software

   At first everyone watched consumer internet, which was easy to understand because you were the user yourself. Later clients asked him to cover Microsoft, so Freddy had to move from consumer internet into enterprise software, covering SaaS names such as Oracle and ServiceNow as well. Enterprise software was hard to use directly and he did not know whom to ask, so he read annual reports and earnings-call presentations, and often attended industry tech forums and exchanged business cards; once he had built up contacts, industry insiders taught him how to read the space. Along the way, the 2019–2020 boom driven by crypto and the pandemic also exposed him to payments and fintech.

   ▶ Watch from here

3. 22:35Insights on Value Creation and Blurring Industry Boundaries

   Looking at any industry comes down to 3 things: value creation, value distribution and valuation. Value creation means being able to predict where technology and products go next, rather than staying at a surface understanding from news and reports; value distribution involves competition and cooperation, and in software, rivals from adjacent fields often burst in suddenly, such as ChatGPT threatening Google Search, which blurs industry boundaries. Many analysts who were bearish on AI back then were driven by differences in domain know-how: people with hardware or semiconductor backgrounds thought it was a bubble however they ran the numbers, while pure-software people were overly optimistic. What analysts are worst at is judging growth that has exceeded its historical trajectory, and because they fear being wrong and being unable to explain themselves to clients, they tend to rely on recent experience.

   ▶ Watch from here

4. 31:20The Big Myths of Tech Investing: "The Big Stay Big" and Single Theses

   Freddy believes the biggest myth is "the big stay big": tech's value creation and value distribution shift faster than in other industries and follow a power law, which leads people to wrongly think a leader's position can last forever. IBM and GE were once seen as unassailable, and in 2007–2008 everyone thought MySpace was the social leader, yet Facebook won when mobile internet rose, so not all of the M7 will necessarily still be around 20 years from now. The host adds that a common mistake is treating a single thesis as truth, for example assuming Apple's lag in AI doesn't matter because of its large installed base of iPhones; in fact that is just one of many variables, and one should weigh the various favorable and unfavorable factors and their sustainability.

   ▶ Watch from here

5. 37:35Factor Investing and Finding Expectation-Gap Stocks

   Freddy's main investment approach is factor investing, and factors are mostly market-cap weighted with some tilt, so large companies are already held in substantial amounts; for the individual-stock portion he deliberately does not follow market-cap rankings. He looks at large companies to understand industry narratives, and then looks for the stocks with the biggest "expectation gap": companies that are on the same track as the mainstream, whose products or customer demand can move faster than the trend, but whose revenue and profit expectations in the market are low, which can be inferred by backing out the market's growth expectations from current trading multiples. Netflix is an example, when paid subscriber growth was slowing, and it was also a research case from his time at KGI.

   ▶ Watch from here

6. 41:20Netflix's Advertising Pivot and Investment Judgment

   After Netflix's paid subscriber growth slowed, the market split into 2 camps on advertising: one did not believe ads could work at all and gave it only a media-style valuation of PE 15–20; the other expected ads to contribute about 10% of revenue the following year. Freddy judged that ad revenue might not be noticeable until about 5 years later, but at that stage 4 signs confirmed Netflix was serious: cracking down on account sharing, launching an ad plan, content such as HBO's returning, and the writers' strike raising content costs. In the short term the ad plan adds paid subscribers, beating market expectations; in the medium to long term, streaming peers exit, competition turns into coopetition, and industry ROE and ROIC could both improve. He believes the key is not value creation but changes in value distribution. The host also shared that he invested in Netflix twice: the first time around 2018–2019, after which he exited; the second time he re-entered after seeing Netflix decide to build its own ad platform, confirming it was serious about ads, and he has held it since.

   ▶ Watch from here

7. 52:35Mean Reversion: From Valuation Expansion to Earnings Growth

   Asked whether tech stocks, which have been strong for the past 15 years, will revert to the mean over the next 10, Freddy thinks mean reversion in valuations could happen because there is indeed a premium now. If tech stocks still beat the market, the reason will not be a valuation re-rating but earnings and growth rates coming in better than currently forecast, rather than the past pattern where just telling an AI story, regardless of earnings, sent stocks up; it is still possible for earnings growth to offset the negative effect of mean reversion, but not every tech stock can do it. Choosing between the S&P 500 and a 10% allocation to each of the top 10 tech stocks, he and the host both bet on the latter.

   ▶ Watch from here

8. 53:50Apple's AI Lag: The Gap Between Fundamentals and Investment View

   Freddy agrees that Apple is indeed behind in AI, but the degree of impact depends on the angle. From a fundamentals view, Apple has a huge revenue base, including income such as search rent, so unless everyone really stops using phones in 5–10 years, the short-term fundamental impact is limited; from an investment view, lacking AI is a mark against it, and the market will question whether it can become the next AI hegemon.

   ▶ Watch from here

9. 56:20Apple Valuation Concerns and Assessing the AI Gap

   Lacking AI may change Apple's bargaining power and pricing premium over developers and suppliers, so it can no longer casually levy the "Apple tax." The host points out that Apple's revenue has grown about 5% a year over the past 2–3 years, yet the market gives it a valuation implying at least 10% growth, and this 5% gap comes from confidence that Apple can find its next growth driver; he describes the disappointment as being "an 80-point company valued like a 90-point one," says that buying Perplexity would be a plus and giving it up because it was too expensive would be a minus, and also thinks abandoning the Apple Car was a very wrong decision, as perhaps was not buying Tesla back then, which raises the question of whether Tim Cook should step down.

   ▶ Watch from here

10. 58:50Tim Cook Is a Peacetime CEO

    The host credits Tim Cook with building the iPhone into a hegemon over the past 10-plus years, with results in chips and the Mac as well. Freddy thinks he is not an expert at going from 0 to 1 but an expert at going from 1 to 1 ten-thousand, skilled at monetizing existing advantages: with the same advantage, other managers could earn for only 10 years, while he can make Apple earn for 15 to 20 years. But crossing into the next S-curve, the AI generation, he may not do well; that doesn't make him a bad CEO, it's just that wartime may call for a leader like Elon Musk.

    ▶ Watch from here

11. 1:01:20Apple's Conservative Investing vs. Microsoft's Stability

    Inside Apple, the view seems to be that over-investing before AI can prove itself is too risky, with worries that returns won't come back, and it has always been cautious about acquisitions and capital expenditure. Holding a strong hand makes it less aggressive, unlike Google, which sprang into action as a whole once threatened; Apple was the biggest beneficiary of the mobile internet era and much of its value comes from it, and it is still watching whether mobile internet can keep going and how AI will develop, so it appears hesitant. The discussion then turns to Microsoft: it has been a very stable holding over the past 10 years, arguably the most stable of the tech giants, and because of its OpenAI investment it was initially seen as keeping up with the AI wave, but the primary risk is in the productivity and business-process segment, namely Office.

    ▶ Watch from here

12. 1:03:50Office Dominance and Younger Generations' Usage Trends

    On the claim that "young people don't use Office," Freddy thinks the point is not young people's habits but the formal workplace: when joining large companies such as those in finance, both the company and its clients use the Windows/Office ecosystem, so people are naturally assimilated. He would assess it from 2 angles: first, whether flexible companies such as small firms are growing as a share of the workplace, and second, whether large companies, if AI tools can replace productivity software and reduce headcount, will still need so many software suites. On that basis, the next 5 years are fine, but after 5 years there are variables.

    ▶ Watch from here

13. 1:06:20The AI Agent Era's Impact on Microsoft and SaaS

    The discussion asks whether AI will replace software and SaaS: SaaS moats come from learning costs and switching costs, and once Agents mature those costs may disappear, and Office may someday be reduced to a gateway to Agents, much as IE was used only to download Chrome. File formats are a more structural advantage, and if Agents can convert formats automatically and seamlessly, that layer of advantage would be weakened; but that is the ideal case, and in the short term Microsoft should still have an edge thanks to its data foundations such as Microsoft Graph and the fact that enterprises' 20 years of data are built on Microsoft formats, while the longer term is hard to say. Finally the host asks: if the partnership with OpenAI ends after AGI is achieved, does Microsoft still have AI competitiveness?

    ▶ Watch from here

14. 1:11:20AWS's Anxiety and Diverging Cloud AI Strategies

    The discussion covers how cloud providers' AI strategies are diverging: apart from Google, most cloud providers lean toward diversifying the models available. AWS originally had a mindset of being a platform and waiting to collect rent, and was the slowest of the few to invest in large models, but after seeing Google Cloud and Azure accelerate, pressure rose sharply, and it rushed to invest in Anthropic and bring in multiple models such as Meta's Llama; Azure is doing something similar; Google focuses on its own Gemini, perfecting both applications and infrastructure. As for whether Microsoft needs to worry about one day being unable to work with OpenAI, Freddy thinks not, because everyone does this, and Microsoft should focus on strengthening what it already does well and keeps winning cloud market share in, with a growth rate that still beats AWS today. Finally the discussion brings up Google's innovator's dilemma.

    ▶ Watch from here

15. 1:15:05AI Search Costs and Google's Dilemma

    The host thinks introducing ads into chatbots is not incongruous: once a free-tier user's problem is solved, they still have the attention to look at ads, so results could be good. Google has publicly said AI Overview's revenue conversion is no worse, but the key to the innovator's dilemma is cost; the host speculates that how much higher AI search costs is uncertain, but it should be around 10 times. Freddy thinks whether Google falls into the dilemma is up to Google, but once AI is introduced, costs will certainly rise: according to the discussion, Google's overall gross margin is about 60–70%, and search alone may be as high as over 90%, so profit will definitely fall after introducing AI. But with Google forced to do it, earning less is better than being crushed.

    ▶ Watch from here

16. 1:18:50Google's Revenue Dilemma in Shifting to AI

    Google Cloud and the advertising business have very different cultures, and AI-related products are mostly built on Cloud; but advertising accounts for over 70% of revenue, so any shift toward AI will lower near-term EPS. If Google chooses to sacrifice profit first to protect revenue and market share, its valuation may first need to be cut 30–40% to be reasonable. Transformations usually don't happen all at once; if the profit margin has to fall 30%, it might be spread over 10 years at 3% a year. Meanwhile Cloud's revenue share must reach at least 20–30% to earn a higher cloud valuation like Amazon's, or a new growth curve such as Waymo must fill the gap.

    ▶ Watch from here

17. 1:21:20Coatue's 2030 List Omits Google and Apple

    The host saw Coatue's Fantastic 40 list (its forecast of the top 40 tech companies in 2030), and neither Google nor Apple is on it. Freddy thinks that although Google needs a reset, Gemini is clearly the strongest model right now, and if a startup owned it, it would surely make the list; he guesses the list assumes the value of digital advertising shrinks sharply and that Google Cloud and Waymo can't carry the whole company, which is too pessimistic for Google, and the logic for Apple is similar, but in 2030 the iPhone should still be number one in phones. They also guess the list has a marketing purpose, deliberately generating discussion, because Coatue's venture arm has recruited quite a lot of industry talent and did not explain it well.

    ▶ Watch from here

18. 1:25:05Andy Jassy Understands Cloud Best: Buy AWS, Get E-commerce Free

    Freddy thinks Andy Jassy is the most underrated CEO among the M7, and no one understands the cloud better. His current valuation view on Amazon is "buy AWS, get e-commerce free." The cloud direction is right: all the models that should be connected have been connected, DeepSeek has cut AI costs sharply, and the more powerful open-source models there are, the better for AWS, because more applications will be hosted on AWS; AWS is like infrastructure, and as long as demand continues, it can control its own gross margin. E-commerce being hit by tariffs is the main reason the market undervalues it, but he thinks that is cyclical, and even with permanent tariffs everyone would adjust on their own, so there's no need to be overly pessimistic.

    ▶ Watch from here

19. 1:27:35Amazon Building America's Everyday Super App

    The host observes that Amazon is gradually becoming Americans' everyday "super app": expanding from e-commerce into groceries, prescription delivery, health insurance, streaming video, and even competing with Starlink in low-earth-orbit satellites, taking over one field after another, wanting to handle every part of daily life itself, and making good progress. Freddy adds that in recent years the market has focused too much on AI, so everyone only looks at AWS, but when he first entered the industry, people spent nearly half their time discussing how to expand e-commerce, international business and improve their profit margins; Amazon has in fact kept growing step by step.

    ▶ Watch from here

20. 1:31:20Amazon Valuation and Meta's Search Ad Opportunity

    The market thinks Amazon's non-AWS business has a long-term growth rate of only around 10% or a high single digit, which seems boring, so its valuation has kept sliding; Freddy thinks that as long as the growth outlook improves (for example from 10% to 15%), it wouldn't be valued as it is now, and he also hopes the company breaks out its reports in more detail and more transparently, because the market values it on a sum-of-the-parts basis, AWS is easy to value, and the non-AWS business ends up looking shortchanged. The discussion then turns to Meta: Zuckerberg remains CEO as the founder and has high voting power, which makes driving change easier; the advertising business makes money, and the other new investments are growth opportunities. If AI lets Meta carve off a small piece of search advertising (the largest piece of digital advertising, most of which Google takes), EPS growth of 25–30% a year could last another 5–10 years, without needing a new S-curve to convince the market.

    ▶ Watch from here

21. 1:37:35Meta's Metaverse, AI Ads and Smart Glasses

    Freddy was always bearish on the metaverse, and only turned positive on Meta once it decided to stop investing in the metaverse (by then it had burned US$20 billion). The pressure at that time came from 2 factors: the metaverse's continued losses, and Apple's IDFA limiting revenue growth, so revenue and profit were squeezed at the same time; now the money-making side looks very good, and as for generative AI's improvement of ad ROAS and back-end products, he originally did not believe it at all and is still half-convinced, but based on the numbers and channel checks it has indeed delivered results. On smart glasses, Freddy thinks their impact is at most 1/4 that of phones, and Meta should not bet the fate of its whole empire on a single device, but should add use cases and interconnect with other devices; he also thinks Google's Android XR is not 100% serious. The host, however, thinks smart glasses will be very big, close to phones, with the two coexisting.

    ▶ Watch from here

22. 1:40:05NVIDIA's Growth Slowdown, and Assessing Tesla's EVs and Vision

    Freddy believes NVIDIA's growth will certainly decelerate from its current high rate; the only question is how fast. Slowdown usually comes from competition, depending on whether AMD can catch up or the ASIC camp can absorb new AI demand. From the history of tech, specialized products eventually become general-purpose products, and once in a general-purpose environment, segmented needs emerge, where ASICs are more efficient in specific scenarios; only when NVIDIA's price-performance in those segments falls behind rivals will the ASIC camp as a whole pose a larger challenge, and only then will growth slow markedly. The host adds that recent growth is about 60–70% and last year exceeded 100%, and asks whether revenue can reach a 30% compound annual growth rate over the next 10 years; Freddy thinks it's possible. AMD will find it hard to catch up, because NVIDIA has a complete ecosystem of CUDA, NVLink and GPUs, and AMD must compete across the board, requiring far greater scale than ASICs that target only specific customers and applications. As for Tesla, EVs face fierce competition from Chinese makers, sales in China have collapsed, and the next 12–18 months won't be great; an individual car owner is unlikely to buy a Tesla for FSD, but it would help if the buyer becomes a fleet platform like Uber. Visions such as FSD, RoboTaxi and robots won't show significant numerical contributions until about 5 years later, but its technology is ahead, and if Physical AI really is the most important thing of the next 5–10 years, Tesla is in the best position.

    ▶ Watch from here

23. 1:55:05M7 Wrap-up: Three Picks for a Decade

    Tech-leaning institutional investors, especially those whose capital doesn't need to worry about short-term cash needs, are happy to hold Tesla; pension funds or endowments wouldn't be so keen to buy it. Asked to pick only 3 companies to hold for 10 years, Freddy's answer is Meta, Tesla and NVIDIA, with Amazon cut and Apple and Google not considered. The discussion then moves on to the next topic.

    ▶ Watch from here

24. 1:56:20Crypto's Long-Term Outlook as an Established Asset Class

    Freddy believes Bitcoin has been established as an asset class and no one will call it a scam anymore; even if it isn't a currency, it is at least a kind of commodity, and asset allocators see it that way too. Large ETF issuers such as BlackRock have begun launching ETFs backed by the physical asset rather than futures, letting ordinary investors participate systematically, much as venture capital went from a group of people pooling money into a fund to becoming a mainstream asset class. With Circle and stablecoins joining, the direction has become steadier; even after exchange blowups and price declines, mainstream asset managers still treat it as an asset class, and once it came under regulation everyone became better behaved. He is optimistic on crypto in the long run, believing it will be allocated at some percentage and will become correlated with inflation and other assets.

    ▶ Watch from here

25. 2:00:05Investing in US Tech: ETFs vs. Individual Stocks

    If you are bullish on US tech as a whole, Freddy thinks buying an ETF such as QQQ or VGT is quite enough; tech ETFs often miss names, because Google and Meta are classified under communication services and Amazon under consumer discretionary, so to hold all these names QQQ is still the most complete. Actively picking stocks to generate alpha is very difficult: you must understand how the company makes money, know how you differ from market expectations, and the position must be big enough to be worth the single-stock risk, and it can still be washed out by macro or monetary-policy issues, unless you have a validated insight and believe it is right. He stresses that the most fundamental things are value creation and value distribution, and as long as the time horizon is long enough and the direction is right, time will correct valuations, so there's no need to fixate on P/E; if a company will rise 200% in the future, paying 20% too much now doesn't matter. The M7 have all been through multiple tech cycles, so don't casually call them bad companies. The host sums up: only pick individual stocks if you can form an insight on a company that differs from the market's; otherwise just buy QQQ.

    ▶ Watch from here