Friday Chatter (20250307)

Friday Chatter features anonymous conversations between two or three people discussing industry rumors and providing both forward- and backward-looking insights into the market.

TSM

H: TSMC, I agree with everyone’s opinion. This deal represents the best outcome for TSMC in terms of negotiation. First of all, not having to partner with Intel is a good thing—Intel’s involvement would be a problem. Let’s be honest: TSMC essentially bought Intel’s fab, but only by getting a discount on equipment prices. Otherwise, it wouldn’t make sense. They claim Intel's 18A is equivalent to 2nm, but it’s an entirely different process. There's no way TSMC could run that process efficiently. Moreover, there’s another benefit: even though it’s still a pushmark, through this deal, TSMC can secure a full exemption from all tariffs. Considering the trips he made, I believe he must have reached some kind of soft agreement with the Trump administration—including a commitment to invest three years’ worth of CAPEX in building a factory in the US (despite TSMC’s annual CAPEX being only about 38–40 billion).

G: I think that doesn’t demonstrate any agreement on Taiwan’s security. In my opinion, Taiwan’s security isn’t even on the negotiation table. And even if it were, following the US’s global retrenchment strategy, if China’s domestic economy continues to struggle by 2027, an attack on Taiwan might become possible. Conversely, if China’s economy is doing well and is well-integrated globally, there’d be no need to attack—if they did, it would mean Xi Jinping lost his mind. On the bright side for TSMC, this deal is a very bullish sign for me. I was especially worried about one thing in the administration: racism. although we know Trump isn’t a racist (he’s just an actor from Hollywood, it’s hard to be a racist there), this administration’s Project 2025 includes many racist individuals. I was concerned about whether bribery from Taiwanese would even work without them being looked down upon for having the wrong skin color. One particularly chilling thought: among the top ten US billionaires, Jensen is the only Asian. The rest are all white.

H: Speaking of which, is Jensen still among the top 10 billionaires?

G: You’d have to check again. Being at the top has its drawbacks. I think if TSMC manages to pull off this bribery successfully—and assuming Supermicro also passed—it’s a very bullish sign. It even hints at whether NVIDIA’s export controls will relax if the bribery is just right.

H: That is lobbying.

G: Okay, it's not really lobbying—lobbying means you pay your congresspeople and senators. This is direct bribery. But anyway, that’s beside the point; it’s just how the executive branch and business synergy operate in this administration. It might surprise everyone, but with the right bribery and lobbying, this administration could have a change of heart—from enforcing strict export controls on NVIDIA to relaxing them. I believe that would be a permanent bullish turning point for NVIDIA, especially since even in its bearish downward spiral, its profits remain strong.

H: I mean, you do hear a lot of chatter about that, right? People say we shouldn’t focus on China; we should concentrate on running our own race at 100 or 200 percent, which I agree with. You can’t just try to limit your competitors—you have to compete at your best. That’s it.

G: All that chatter comes from some VCs. Their vested interest is that if China re-enters the game, they’ll have more attractive investment opportunities—especially the ones we discussed last time. Investing in developing models isn’t very appetizing for VCs because it’s like throwing tens or hundreds of millions into a void without knowing if your model will succeed, or if the application layer will benefit. Yet, you can’t simply stop investing in new models because they unlock new capabilities in the application layer; without them, there’d be no new applications at all. It’s a very tricky position—investing huge sums in models while knowing the money might never yield a return. But if China re-enters the picture, it’s a different story. They’ll have plenty more investment opportunities; they can invest in China or leverage Chinese capital to invest in models. And nobody talks about it. Even now, there’s a lot of Chinese capital in U.S. VCs as limited partners, but no one discusses it because you’re not supposed to. VCs remain very receptive and respectful toward China. That’s why you hear them advocating for lifting the sanctions on NVIDIA. Ultimately, from my perspective—and we have huge holdings in NVIDIA, so take that with a grain of salt—I think we should lift the sanctions on NVIDIA chips, as these sanctions are becoming more absurd over time (especially with the 5090D and all the 50-series debacles). Continuing these export controls and tariffs makes little sense; lifting them would benefit competition and ultimately undercut Chinese competition. I also believe that Western investors fail to realize that China has a degree of market economy. Although even after lifting sanctions, China can continue investing domestically, the support it receives with sanctions in place is completely different from that with sanctions lifted.

H: Yeah, sure. I think they understand that. Everyone is saying that if you rely on them, they’ll use government funding and subsidies to build the entire domestic semiconductor ecosystem—from fabs upward, I believe that's right.

G: What I'm saying is that even without sanctions, the government will still pour money into domestic semiconductor production and leading-edge chips. But if you lift the sanctions, the market will boom because foreign companies competing for talent in the domestic market will muddy the effectiveness of government investment—and that’s good for foreign enterprises. At this point, I think the sanctions on application chips aren’t very useful. We might still keep the sanctions on 7nm fab equipment, because those days are numbered. China is expected to have its EUV equipment ready between the end of this year and early next year, but actual production is likely projected for around 2027. So whether you lift export controls or not, it probably won’t affect that timeline much.

AVGO

G: I don't quite understand why everyone is fixated on the 2027 number from AVGO while NVIDIA is dealing with 2025 and 2026 figures.

H: For the most part, I think people don’t buy NVIDIA’s sustained CAPEX story. They see NVIDIA as a hardware company and wonder how a hardware company can consistently demand a 75% gross margin—it just doesn’t add up. The moment they see a margin drop, even if it’s due to efforts to ramp up Blackwell, they immediately think “those days are numbered.” Competition is fierce—other companies can design their own ASICs, and the compute landscape is shifting toward inference, which doesn’t require a Ferrari-level product; you can run inference on something more like a Honda. (To be clear: “Ferrari” here refers to NVIDIA, while “Honda” refers to an ASIC like Grok or similar.) That’s why people question whether NVIDIA’s peak earnings and margins can be sustained. The market is essentially saying no, even though NVIDIA’s VMware-like recurring revenue and connectivity business give it some leeway. Plus, its AI ASIC segment is growing rapidly from a low base, with the idea that it will eventually become a $70 billion market capturing 50% of it—that’s the rhetoric for 2027, even if I don’t fully buy into it.

G: I think the fundamental issue is that we’re dealing with two types of companies. One type takes on custom contracts, while the other designs and builds products based on a multi-year roadmap. For public investors, this creates a duration mismatch. For companies that secure customer contracts—especially in ASIC design—the time horizon is about two to three years, from contract initiation to full execution. For example, Broadcom already has visibility on its expected revenue in 2027 from contracts signed in 2025. But for product companies, whose competition is other product companies, public investors have no clear idea what the final product will be or how the market will react. It could be a huge success or a massive failure. That duration mismatch is just too hard to swallow.

H: I talk to investors regularly, and the mismatch I see is that people still view NVIDIA as a hardware company—much like the debates about Apple in 2016. I’m not saying NVIDIA’s model is identical to Apple’s (not yet, anyway), but back then critics argued that Apple, as a hardware company, should have been valued similarly to Samsung—trading at mid-teens PE instead of mid-20s—and many predicted Apple would eventually drop to those levels. The same dynamic is playing out now. People view AVGO’s revenue as recurring and nearly immune, almost like a SaaS model in hardware, while others insist, “Show me the product, and I’ll evaluate you year over year.” Then I have to hedge against the risk that the next product could see a drastic drop in demand—a huge left-tail risk where the product fails and inventory piles up. That’s the biggest mismatch between these approaches.

G: I agree—running a product company is incredibly challenging. Look at Tesla: they had a few good products, then spent five years without a new release before dropping the Cybertruck. We have no insight into their product development process—why the Model Y didn’t get an incremental facelift, then suddenly did—making global sales drops hard to explain. It’s likely due to retooling for the Model Y combined with the general dislike for Elon affecting sales. We won’t have a definitive answer until March or April. By then, we should know if it’s Elon’s influence or just temporary Model Y issues. In any case, running a product company is as tough as it gets—just consider Apple over the past 15 years. They must ship a new iPhone every year with significant changes, and despite staring into the abyss, they manage to sell through clever marketing, revised payment plans, and new initiatives. Apple uses many tricks to keep the upgrade cycle going, and Tim Cook doesn’t get enough credit for that. On the Android side, the leading company is Samsung thanks to a broad product lineup ensuring smooth sales cycles. Yet, each cycle features a different leading product. In contrast, Tesla has dropped the ball—announcing no releases is easy, but maintaining a consistent product cadence is extremely hard. That cadence, along with recurring revenue, ultimately determines whether a product company succeeds or fails. Look at NVIDIA’s 50 series chips—they have many issues, yet they had to release them after two and a half years because delays aren’t an option.

H: I agree. Think of it like fast fashion—most hardware companies are judged product by product. If you have a huge hit, it’s almost destined to fade because the odds of back-to-back hits are slim. Typically, investors wait during a bust, then when the next hit comes, the profits are enormous. I believe what you’re saying—and I implicitly agree—is that founder-led product companies (like Apple and NVIDIA, where the founder is deeply involved in product design) are an exception. They might have minor missteps, but they rarely recover because they’re too bogged down in managing every detail, unlike leaders who’re detached from day-to-day operations.

G: Tim Cook isn’t the founder, after all. There’s a spectrum among hardware companies. For example, Samsung is a hardware company but isn’t seen as a product company in the same way. The same goes for Micron and even AMD—which is more hardware-oriented—while NVIDIA and Apple (and even Tesla) are firmly in the product company camp. AVGO, on the other hand, leans more toward hardware. So, the observations you made apply more to product companies than to hardware companies in general. Many hardware companies don’t release consumer-facing products; they manufacture components for others to integrate and sell, which makes them harder to evaluate as standalone entities.

H: But I mean, we’ll see how this cycle turns out for NVIDIA. However, one thing I’m still trying to wrap my head around is that it’s not so much about NVIDIA—it’s more about the broader macro backdrop.

Scott Bessent

G: I think that's what we need to discuss right now. We've talked enough about semiconductors—let's move on to macroeconomics and talk about Scott Bessent. He’s been appearing on TV frequently these days.

H: I mean, I don’t even know what his motivation is at this point. During Trump 1.0, whenever Trump announced some crazy policy, Steve Mnuchin would go on TV to calm the market. But Scott Bessent? Oh my gosh, he does the exact opposite—trying to spook the market even more with all his talk about a detox. And then he says things like, “the stock market doesn’t really matter; you can be up 20% for two years in a row and still get voted out.” What is he aiming for? Besides that, he looks absolutely terrible on TV—he comes off as an idiot, even a retard, and he speaks so slowly. If you’re going to spout all that contradictory nonsense, at least do it fast, for crying out loud.

G: That's what I was thinking. He probably hired a public speaking coach recently, which explains his odd TV presence and slow speech—he’s trying to apply the lessons from his coach.

H: I listened to his podcast right before he was nominated for Secretary of the Treasury, and he sounded completely different—smart, seasoned, educated, a true veteran of the financial markets. But on TV, he comes off as if he’s saying, “I’m so good, you have no idea what you’re in for,” and, “I don’t care about you.”

G: I think Bessent is seriously underrated. He provides a rare, unfiltered glimpse into what this administration is really thinking—no sugarcoating—and that raw honesty unsettles the market because it's just the truth. In contrast, figures like Marco Rubio or J.D. Vince blend market-friendly sound bites with their own spin and the administration’s messaging, making their speeches seem palatable until you spot the glaring gap between their words and what actually happens—leaving you to wonder, "What's really going on?" But I suspect Bessent’s time in the administration is numbered; he’s downright terrible on TV. Meanwhile, Rubio might stick around longer—especially after his clash with Elon—because reality TV loves a good spectacle, and the president seems to favor that kind of show. And when it comes to the recession, you're missing a key point: those who hit their prime during the Clinton years—mostly white, male, and supremely confident—treat a downturn like a minor hiccup. Elon, Bessent, and Trump all come from that era (though Elon, as a younger immigrant, operates on an entirely different level), and they genuinely believe they can bounce back from any downturn. That said, I have this superstitious belief—there’s a sort of mystical force at work, a collective consciousness, or even a God-like push—that favors certain narratives while disliking others. You know the kind: a tale of a hero fighting a villain or slaying a greedy, evil dragon, where the hero always wins, no matter what. Yet when it comes to the other kind—an overconfident man predicting the future—that never pans out. Look at Alan Greenspan’s endless boom prophecy, which ended with the Great Recession; we’ve seen it time and again. There are forces—strange, inexplicable forces—that ensure narratives of overconfidence in predicting the future, as appealing as they are, simply don’t come true. I totally understand their perspective: if we somehow navigate this detox phase, get the budget deficit under control, slash Social Security and Medicaid, and let some of the...

H: Trump said he's not going to cut all of that.

G: It's going to happen—you'll see a wave of boomers pass away, and the economy will improve as the age imbalance is corrected, with more young people ready to work (even if they aren’t college educated and are primed for manufacturing jobs). It might seem like everything will magically work out, but I believe that if a narrative rests on one man confidently predicting the future, it just won’t pan out. It’s not rational; it’s superstitious.

H: I just don't trust any of those fuckers—I think they're going to screw up big time. Do I believe the situation is as dire as they claim? Probably not, because I’m not that bearish. Everyone I talk to seems overly pessimistic, expecting the market to tank, and non-US investors are even more bearish. They compare the current US fear to China’s 2022 lockdowns and irrational policies. But honestly, if you go out to dine, travel by plane, or visit places like Hawaii (which we just did), you don’t see any signs of a consumer slowdown—and that’s what really matters for the economy.

G: Hawaii isn’t a fair comparison—tourist numbers there haven’t recovered since the Maui fire. You can have a booming economy while the stock market remains depressed. Look at China after the 2015 crash: its GDP grew at 6–7% per year, even though the stock market lagged. The US is more complex, but what if the stock market stays down without the economy falling into recession?

H: It's unlikely.

G: Unlikely, but what if it does happen?

H: No, you have to consider the composition of wealth. In China, 70–80% of wealth is in real estate, which is why after 2021—when domestic developers went bust and property prices tanked—China entered a massive deflationary period where the wealth effect disappeared and spending stalled. In the US, over 70% of family wealth is in equities, making the situations quite similar. My point is, if the stock market crashes, consumer confidence will suffer, triggering a deflationary cycle that stifles economic growth. You can’t engineer a prolonged stock market drawdown in the US without it impacting consumer confidence and the real economy.

G: I think that by simply scaring public investors, they might trigger the recession they’re aiming for.

H: Dude, that’s why I’m not feeling that bearish. First off, we’re at a 4.25% Fed funds rate—so by slightly hurting consumer confidence and introducing recession talk, you can push down inflation (since inflation is largely about expectations and confidence), giving the Fed room to cut rates. Secondly, oil prices have dropped significantly, saving consumers money. And third, without rehashing all the data (I’m not a macro investor), the overall signs suggest I’m not overly bearish. I do believe those old white motherfuckers will eventually screw up big time—but not right now.

G: I think you're painting a rosier picture than even they hope for. You’re saying that we are going to…

H: I mean, at least in the short term.

G: We’re going to have a pretty painless soft landing, with inflation, interest rates, and even the 10-year treasury yield coming down. That’s why they want to engineer a recession—if they can trigger one without an actual recession, why even aim for it? Then the economy would start growing again. I think that’s a rosier picture than what they were even hoping for. Let’s end on that positive note.

H: Let me add one thing before we wrap up on that positive note. I don’t think their actions will cause much short-term damage, but they are seriously undermining the US dollar’s status as the world’s reserve currency—which will have long-term implications.

G: I don’t even think they want the US dollar to remain a long-term reserve currency, considering they’re setting up swap agreements between USD, USDT, and USDC.

H: That's the fucking dumbest thing, right?

G: That swap agreement essentially gives them the right to print as much USD as they want. There’s no regulatory oversight for USDT and USDC—USDC is self-regulated domestically, and USDT is self-regulated internationally, meaning there’s basically no regulation at all. So, if they establish a swap agreement between USD, USDT, and USDC, I see it as even more bullish than spending tax dollars to buy Bitcoin for a reserve. It basically means they can print unlimited USDT and USDC, and if that agreement comes to fruition as rumored, it shows they don’t really care about maintaining the USD’s reserve status. It’s simply not their concern. Maintaining a global currency isn’t a central priority for this administration.

H: Okay, let's end on that note.