Friday Chatter (20250404)

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

Tariffs

M: It’s like the boot has finally hit the ground on tariffs this Wednesday. According to the tariff formula, the purpose is to balance trade. If you ask ChatGPT to crunch the numbers—assuming everything else stays the same—the collected tariffs would exactly wipe out the U.S. trade deficit, or even those of all countries. Of course, there are some very interesting twists in between. I noticed that some small enclaves—because last year they bought a few U.S. aircraft engines with a trade deficit of over three million dollars—ended up with a 99% reciprocal tariff. There’s even an extra tariff on U.S. military overseas bases, and the funniest case: Heard and McDonald Islands adding a 10% tariff. Setting those aside, I think Bitcoin’s bounce back last night might have given everyone a bit more confidence—that it can carve out its own trend instead of being dragged down along with everything else amid global risk-off selling.

N: No, I think it’s some kind of weird short covering. Today—being the 5th day of this sell-off—we’re seeing strange behavior. For example, gold is down 2–3%, and even the yen (traditionally a safe haven) is falling. They should be rallying along with treasuries, right? Instead, both are down and seem to conflict with one another. The yen is down, yet not because of a stronger dollar (since gold is also down), and the USD remains strong. I suspect this is just a liquidity event or forced selling, with self-inflicted pain from forced liquidations. After all, it’s a risky asset—a direct measure of people’s risk appetite. We can all agree we’re in a risk-off environment, right?

M: Yes.

N: So, I don’t think there’s going to be a single day with one specific, large flow that can drive an independent trend.

M: I think that if Bitcoin—still a risky asset—suddenly graduates into its own independent trend this weekend, it could trigger some very interesting consequences. Being a risky asset that’s locked in, it can’t be easily sold off, which might lead to defaults. Since 2022, U.S. banks (including Goldman Sachs) have even started taking Bitcoin as collateral, and we’re now seeing credit-derived loans tied to it. I need to research more on what these loans really mean for the market. But if Bitcoin breaks out on its own this weekend, it won’t be good news for the overall market.

N: Of course—especially for the stock market, that’s obvious. To elaborate, you believe that Bitcoin and crypto are assets even Trump and his family keep a close eye on, right?

M: Yes.

N: If they were to have an independent rally over just a single weekend, it would give them the feeling that they’re invincible—because the asset they’re focused on would keep rising on its own, right?

M: And then, from a political perspective, if people find that they can’t sell this risky Bitcoin asset, they’ll double down on selling off other assets as a cover. When assets can’t be sold, more liquidations follow. We might not know which one will be the next Lehman Brothers, but I worry that a Lehman moment is just waiting to happen.

N: Okay, we don’t even have to go that far. I’m just saying this gives them an invincible feeling. You can clearly see how this negative feedback loop works: they act like they don’t care about anything and can just do whatever they want—as if they don’t give a damn. If this attitude carries into next week, I think some people are already calling it Black Monday—much like October 1987, when the stock market fell around 20% in a day. I mean, it’s not impossible.

M: I think it depends on what happens this weekend. Right now, the “Crypto Bros” are all shouting “Buy the Dip,” while the “Financial Bros” remain a bit reserved.

N: The mood is extremely negative. Since yesterday, they’ve been discussing negotiation tactics—but they’re still holding on to hope, like Vietnam saying its tariffs will drop to zero and the UK starting negotiations with the U.S. If these sentiments hold, Monday could even be a bounce-back day.

M: But I don’t even have to warn you—you’ve seen the TikTok news, right?

N: So?

M: It’s going to extend for another 75 days, which means there won’t be any deal or negotiations with China this weekend—or even in the coming weeks. This uncertainty is just going to drag on for the next 75 days.

N: I think China is a lost cause—we should brace for a price hike of around 30%. Yesterday, many checked with Chinese suppliers, and they’re only willing to budge by about 2%, which isn’t nearly enough to cover a 34% jump. The Chinese government says there will be a stimulus package, but it won’t be huge. Companies will negotiate for the U.S. to shoulder the price increase instead of relying on Chinese stimulus. So, while last week we were unclear about how inflationary these tariffs would be for the U.S. market, this week it’s clearer that most of the impact will likely be absorbed by U.S. consumers, since it’s a universal tariff.

M: So you can’t really shift production to lower-tariff regions to escape this; the maximum relief might be around 10%. Moreover, previous tariffs have already driven supplier prices so low that they can’t afford further cuts—meaning the U.S. has to absorb the cost. This isn’t a new scenario; just yesterday, more Chinese suppliers were discussing it. In fact, there were rumors as early as last week that Walmart was pressuring its Chinese suppliers to lower prices.

N: In a typical scenario, costs might be 20% plus a currency offset of 5–7%, and if suppliers lower their prices by that amount, it might not seem too bad. But this time is different. The dollar index (DXY) year-to-date is probably down 7%, meaning the dollar’s purchasing power is lower while costs remain high. I agree with you—and if everyone faces the same tariff, there’s no reason to negotiate because everyone knows it applies universally. Besides, preparations for this have been going on for a very, very long time.

M: China has been bracing for a cumulative 60% tariff for a long time—and now it’s effectively at 64%, right in the range they anticipated.

N: Yeah, they’re thinking, “Worst case: if we stop trading with you, we’re fine.”

M: Looking at it another way, it’s understandable. U.S. agricultural products have plenty of substitutes—mainly because they’re cheap. Take Utah, for example: it produces lots of farm goods that end up being sold to China rather than consumed locally. For farmers, that’s profitable, but for the local population, it isn’t necessarily a good thing. So aside from farmers taking a hit, it might not be all bad for those regions.

N: I don’t think it’s surprising if China retaliates. If the U.S. only adds, say, a 30% tariff on China, China might not retaliate because over the past few years they’ve shifted production bases to India, Vietnam, and elsewhere. But now—even those regions face high tariffs—making it very hard for Chinese companies to operate. The Chinese government might think, “Since you’re at war with everyone, I’ll just team up with everyone else to take you on.”

M: It’s unlikely they’ll align with Europe or Canada, but they might find more common ground with Asian or some developing countries.

N: I don’t think it’s a long-term alignment, but when negotiating with Trump, he might try to cozy up to Europe for a bit—I think that’s possible.

M: For Europe, let’s be honest: an alignment with the UK, France, or especially Germany is unlikely—Germany’s automotive industry is in direct competition with the U.S. However, countries like Italy, Spain, and Portugal—previously part of China’s Belt and Road—might be more open.

N: Hungary.

M: Hungary was part of the Belt and Road, but due to Biden’s policies, these countries have pulled out.

N: Now they’re rejoining the Belt and Road.

M: That’s not out of the question either, but another boot hasn’t dropped yet—how exactly will tariffs on chips be levied? There’s still little clarity on that. So that’s another issue looming. Even before that happens, I’m already sensing some momentum toward buying Nvidia.

N: This is straightforward. First, companies like Nvidia have solid margins—enough to absorb a shortfall in the worst-case scenario. Second, they enjoy significant pricing power; their products are in high demand and sold mainly to large enterprise customers. In contrast, memory companies like Micron or Microchip deal in commoditized, consumer-facing products with thinner margins and weaker pricing power, and their consumer demand isn’t as robust. So if you have to choose among semiconductor companies based on pricing power and quality, I’d lean toward companies like AVIDIA, AMD, or Avago. But let’s be honest: if a recession hits, the entire semiconductor sector could tank—possibly by as much as 20% in two days. There’s a reason for that.

M: But if these tariffs are here to stay—which seems highly probable—there will be a slow electronics upgrade cycle. All electronics might see a 5–10% price increase (for example, a $799 iPhone could jump to $899). And if the upgrade cycle is slow, more AI computing will shift from edge devices to the cloud. At the end of the day, I believe…

N: I have a question: why, in an upgrade cycle, would computing shift from the edge to the cloud?

M: It’s a slow upgrade cycle—that’s exactly why. The idea of an “AI PC” is becoming obsolete, so people will move more computing to the cloud. And we all know you can’t levy tariffs on bytes. This means some companies—like Oracle and OpenAI, which plan to build data centers in the U.S.—will be at a disadvantage. Meanwhile, data centers in places like Osaka or Malaysia won’t face high equipment tariffs and will be more competitive.

BTC

N: I have a question: what does “American-made Bitcoin” even mean?

M: It basically means that Bitcoin miners in Texas use Texas-grade electricity to mine Bitcoin right here in the U.S., and those bitcoins…

N: Does it have to be Texas?

M: Well, Texas is where most Bitcoin mining operations are located because many believe they’ll get long-term tax benefits. Also, while Texas electricity isn’t exactly cheap, the spot pricing lets miners operate during the cheapest hours—making Texas ideal for Bitcoin mining.

N: I have another question: how do you know if Bitcoin is American-made? Does it come with some kind of stamp?

M: First, everything on the Bitcoin blockchain is public so we can trace which bitcoins were mined by whom. That said, there isn’t a tagging system yet. So if U.S. miners claim their Bitcoin as tax-free, it is tax-free on the chain. However, once Bitcoin crosses to off-chain systems (like mixers such as Tornado) and then comes back, it’s hard to tell if it remains tax-free. For miners, though, it’s quite easy to claim tax benefits in the long term.

N: I have a question: for Bitcoin already in circulation, how can I tell if it’s American-made or qualifies for an exemption?

M: For older bitcoins, it’s harder to tell since you can’t track which mining pool they came from—it might even be mined on someone’s personal computer in 2010. Fortunately, most of those coins are locked up anyway. Newer bitcoins, mined through pools, are much easier to trace, so we can determine a significant portion as American-made.

N: Interesting. But if I have a coin in my Bitcoin wallet right now, can I tell whether it’s American-made?

M: Yes—if a coin was mined recently and can be traced back to a specific miner, you can tell. But as bitcoins go through numerous transactions, even if they started out as American-made, there will come a point when that status is lost and taxation kicks in. Essentially, every Bitcoin can be traced back to its mining block, and if that block was mined using U.S. electricity, then that Bitcoin is considered U.S.-made—but only until it goes through enough transactions to obscure its origin. Then it might be “tainted” and lose its tax-free status. Ultimately, there has to be a cutoff for tax purposes.

N: Yeah. Now, here’s another question: If I mine Bitcoin in Iran where electricity costs almost nothing (which many miners do for the margin), that Bitcoin obviously wouldn’t be American-made by rule. But how does that work in circulation? Does it mean Bitcoin effectively forks from that point?

M: No, it’s not a fork. Bitcoins mined in Iran will simply flow back into circulation. Unless U.S. authorities decide to ban Iranian mining—which might force a fork and lead U.S. (and likely Chinese) miners to exclude them—that’s not happening. So they just become part of the circulating supply and are recognized as non–American-made, subject to export controls. Essentially, even if it’s technically illegal, once it’s in circulation, authorities won’t pursue criminal actions over it.

N: But then you’re effectively creating two different types of circulating Bitcoin, right?

M: No, no, no.

N: If I have the option to buy Bitcoin, I’d naturally choose American-made Bitcoin for the tax benefits.

M: And then?

N: So in circulation, Iranian Bitcoin would likely have a permanent discount relative to American-made Bitcoin.

M: It could be.

N: And that discount effectively creates two separate circulating assets.

M: It might appear as two systems off-chain, but on the Bitcoin blockchain itself, one Bitcoin is still one Bitcoin—you can’t actually exchange one for the other as separate assets.

N: What does that even mean?

M: Off-chain, however, it’s conceivable that one Bitcoin might be valued at $80K while another is at $70K. In that case, you’d need some kind of smart contract to obscure its origin—but I don’t think that’s practical.

N: No, it still confuses me. If you can swap U.S.-mined Bitcoin for Iranian-made Bitcoin on-chain, but off-chain they trade at different discount rates, wouldn’t that create a risk-free arbitrage opportunity? I could just convert my holdings and exploit the difference.

M: That’s why I think it’s not possible. On-chain, the origin is always identifiable—whether it’s Iranian or U.S. Bitcoin.

N: So, as I was saying, you’ve effectively forked the two types—meaning that even on-chain, no one is willing to swap them directly.

M: On-chain, imagine one wallet holds U.S.-mined Bitcoin and another holds Iranian-made Bitcoin. If one wallet sends U.S.-mined Bitcoin, the exchange of value happens off-chain—and that might reflect a different price. Essentially, there is no pure on-chain exchange between different types of Bitcoin. They might have different values on decentralized or centralized exchanges, or when converted to Ethereum, but on-chain one Bitcoin is still one Bitcoin.

N: Now consider a scenario where every transaction is in Bitcoin—for example, I buy something with 2 Bitcoins and you owe me 0.5 Bitcoins in change.

M: No, no—there isn’t an exchange mechanism like “giving change.” It’s digital cash. Using Bitcoin as a medium of exchange creates issues because you wouldn’t normally use an appreciating asset for daily transactions. For example, if you buy a pizza, the store might price it in U.S.-mined Bitcoin so that you only need one coin, but if you use non–U.S.-mined Bitcoin, you might need two coins for the same pizza. This doesn’t mean there are two separate circulation systems; it just creates two different price points for the same item—underscoring why Bitcoin is a poor medium of exchange.

N: So basically, you’re creating two different kinds of assets.

M: Yes.

N: So we start with one asset, but it ends up splitting into two distinct assets.

M: Exactly—it's not that you value one Bitcoin differently from another, but rather…

N: Okay, so if I’m a U.S. citizen, I’d definitely want to buy U.S.-mined Bitcoin because of the tax benefits.

M: Because you’re a patriot.

N: Anyway, what’s the implication? If everyone mines Bitcoin in the U.S., does that mean we’ll run out of electricity?

M: No, no, no. First, people have some very twisted logic about this. They claim that mining Bitcoin means we need to build more power plants—and that by doing so, Bitcoin can absorb the excess power, which in turn makes electricity cheaper in the U.S. It’s a very convoluted rationale, but if you assume building a power plant is free, it kind of makes sense.

More Tariffs

M: In the past couple of days, more news about tariffs has emerged—including reports of internal conflicts. For example, Elon and Bessent don’t agree with the proposal currently being announced.

N: Bessent is questionable, but Elon has already distanced himself.

M: However, Lutnick, Paul Navarro, and Trump himself seem to be the driving forces behind this framework. Even Sacks hasn’t spoken up in a long time.

N: Does staying silent now mean disapproval?

M: Uh, I think they’re still strategizing—figuring out how to get involved in the discussion without alienating their billionaire friends. At the end of the day, this is part of a “Great Reset.” Next week, some people might even start jumping off buildings, and as small and medium businesses go bankrupt, there will be…

N: Let’s take a step back. Your read is that if these tariffs remain in place for just a week or two, it could cause many industry supply chains to collapse—leading us into a massive crisis.

M: Right. We’ve already discussed that—in negotiations between suppliers and importers, suppliers refuse to absorb these tariffs. That leaves price hikes as the only option, which will undoubtedly be inflationary.

N: In the short term, it acts like a consumption tax. Initially, it’s inflationary, but over time it will depress demand—pushing us toward a recession, or worse.

M: Then you end up with winners and losers. For instance, I think the niche industry of building AI data centers in the U.S. will take a hit—especially for new companies like OpenAI and XAI, which have already invested billions in U.S. data centers.

N: Here’s the deal: whose deal did OpenAI broker—Larry Ellison? And who’s behind XAI—Elon? Even if Elon gets pushed out of Washington for disagreeing with the trade deal, a $300 million donation still carries weight, and being a lifelong friend of Trump counts for something. Those two will probably be fine. Even if it isn’t a tariff reduction or exemption, they can likely secure government support. I’m just speculating, but it looks like this new regime is effectively picking winners and losers—they’ve got a trump card, to some extent.

M: I think it’s tricky because taking government subsidies isn’t popular among conservatives. That’s why I feel if these tariffs stick—even for three weeks—they’re likely here to stay. In that case, many small and medium businesses will go bankrupt, and I doubt the government will support them.

N: The closest analogy is COVID—a massive supply chain disruption. But the key difference is that during COVID, central banks and governments responded super fast with monetary and policy measures: extended loans, grants for small businesses, and near-zero interest rates. However, according to Powell’s comments on Friday, the central bank isn’t planning to cut rates now.

M: They’ll only cut rates once a major bankruptcy forces their hand—but that’s just cause and effect. It’s not an overnight decision.

N: I mean, unless a big bank goes bankrupt, the Fed will likely just sit it out. Small business bankruptcies aren’t really the Fed’s problem. If a major bank collapses or the repo market tightens drastically, then the central bank would need to step in. Otherwise, I don’t think the Fed will get involved.

M: For the short term, I agree—it’s unlikely that building AI data centers in the U.S. will receive subsidies or preferential treatment if tariffs rise by, say, 20%. For many small companies like Lambda Lab, CoreWeave, or Nebius, it’s much easier to build data centers outside the U.S.

N: I think most companies will simply build their data centers in Canada. Why not?

M: Everyone except for OpenAI and XAI (who have already invested heavily) and Facebook—which has the disadvantage because it hasn’t built many data centers outside the U.S.—will likely opt to build abroad. Other companies are better positioned because they already have a strong multinational presence.

N: You mean Google?

M: Yes—Google, AWS, Microsoft, and other cloud service providers are well positioned to shift orders from the U.S. to countries like Malaysia, Japan, or even parts of Europe and Canada. After all, you can’t levy import taxes on bytes and bits.

N: I think a 20% hike—or even close to 30%—is too steep. The tariff on Taiwan is around 27%, right? That’s too much for end consumers to fully absorb. Either the supply chain or the ODMs, who are already operating on razor-thin margins (around 7–10% gross margin, with most at 7%), would have to bear the brunt. It might end up being split between companies like Nvidia and the end user.

M: No, I think it’s hard to imagine a scenario where that cost is passed on to the retail side for computing.

N: No, I mean enterprise.

M: It’s tricky—CSPs can absorb a lot. For example, AWS charges about $10 per hour for an H100 per card, which is huge compared to the nominal price of $1.5 to $2 per H100. They have absurd profit margins and can absorb the cost. However, enterprise users, who are very sensitive to price increases (since it affects all their projections), might see prices rise more slowly than consumer products. This could push us toward a recession due to demand elasticity. That’s why I’m planning to stock up on coffee beans tomorrow—maybe a year’s supply. There is no supply of 5090s now, so I might have to take one for the team. And yes, an iPhone might end up costing an extra $100, which is acceptable.

N: But did you see the estimates? Some analyses suggest iPhones could be about $300 more expensive under the new tariff regime. If Apple only passes $100 on to the consumer, then they’d have to absorb $200 themselves—or split it with their suppliers.

M: It really depends. I might even cancel my annual upgrade contract with Apple and switch to an outright cash purchase—because here in the U.S., tariffs apply, whereas they don’t elsewhere. I’d rather skirt the tariff (almost like tax evasion) than pay an extra $300. Realistically, I think we’ll end up paying an extra $50 to $100 on average. However, those stuck in AT&T contracts or unable to pay full price for an iPhone will suffer. People with more financial flexibility can work around it, which might become the norm here in the U.S.—I experienced similar tactics with small items when I was in China.

N: Yeah, I think consumer electronics components are in even worse shape than, say, AI servers. But at the end of the day, if we head into a recession and demand drops, everything will be in trouble.

M: Do you want to talk about Lama 4? It’s set to release tomorrow.

N: Oh, okay. What are the expectations?

M: Lama 4 is going to be based on a MOE (Mixture of Experts) architecture. For the initial models, don’t expect them to be exceptional—that’s the consensus. They’re just the first step; reasoning model will come later. Since it’s a MOE model, don’t expect to be able to host it on your own. It’s clear that most of these advancements are headed to the cloud. Tariffs only worsen the situation—discouraging electronic upgrades and leaving insufficient compute power. In short, models are getting larger and will be hosted in the cloud, which raises issues of privacy and security. People are willing to compromise on those issues simply so they can use the technology. For edge devices, tariffs and the slow pace of efficiency improvements mean they can’t keep up. We still haven't reached AGI yet, so people are not going to reduce the size of the model or make it dumber in order to fit on a device. People have to make the model smarter and smarter and smarter. Even if that means it has to use more compute, more memory, and the model itself is going to be bigger, with more training data—everything just has to scale up. The only complication is that AGI is delayed due to the tariff, which is the biggest open question.