GUY: Good morning, good morning... it is Tuesday, May 13th, 2026, and this is Morning Signal. I’m Guy, caffeinated and a little too excited about commodity curves before breakfast.

AVA: And I’m Ava... also caffeinated, but for very different reasons, because today is one of those mornings where every story somehow comes back to AI, oil, China, shipping lanes, and who actually controls the physical world.

GUY: Exactly. And that, to me, is the big takeaway. We’re not in the cute phase of AI anymore where it’s just demos and valuations and everyone arguing over who has the best chatbot. We are now in the phase where AI is showing up in payrolls, in power demand, in oil sensitivity, in supply chains, in geopolitics.

AVA: Right... and the market still kind of talks about these as separate stories. Like, there’s the AI story, and then there’s the oil story, and then there’s the U.S.-China story. What really jumped out from the podcast set today is that they’re not separate at all. They’re becoming one system.

GUY: Let’s start there with markets and macro, because I think the macro framing is the cleanest way into everything else. So over on RenMac Off-Script, and also over on Forward Guidance, Neil Dutta had what I think is the most important macro description right now. He basically said the U.S. economy is functioning like a one-engine plane... and that one engine is AI-related capex.

AVA: Which sounds dramatic, but when you actually listen to what he means, it’s not hype. He’s saying the labor market is not broadly reaccelerating. The pockets that are holding up are very specific. Non-residential construction, heavy and civil engineering, specialty trades, bits of durable-goods manufacturing... all the stuff that gets pulled along when you’re building data centers, power systems, industrial capacity.

GUY: Exactly. And the flip side matters just as much. Over on RenMac Off-Script and Forward Guidance, Dutta contrasted those strong pockets with softness in residential construction, information services, financial activities, and accommodation. So if you just look at the headline economy, you can say, hey, things are resilient. But if you look underneath, it’s narrow as hell.

AVA: And that changes how you interpret good data. Because if payrolls hold up, that doesn’t necessarily mean the economy is broadening in a healthy cyclical way. It may just mean one capex complex is doing a lot of heavy lifting.

GUY: Right. This is not 2017 broad reflation. This is not everybody hiring, everybody spending, every sector running. This is one giant industrial project propping up a lot of adjacent activity. It’s more like... if you built a whole regional economy around one giant shipyard. The shipyard’s booming, the welders are busy, the truckers are moving, the suppliers are humming... and everybody says, wow, what a healthy economy. But if the shipyard sneezes, the whole town catches pneumonia.

AVA: OK but here’s what’s interesting... on Forward Guidance, Dutta didn’t sound recessionary either. He’s not saying the economy is fake. He’s saying it’s real but narrow. And the wage data is part of why. He pointed to average hourly earnings and the employment cost index both around 3.5%, plus another month of 0.2% month-over-month wage growth. That’s not a wage spiral.

GUY: Yep. That’s the key nuance. If wages were reaccelerating to 4.5%, 5%, then you’d say maybe the economy is hot again and the Fed has a real inflation problem. But 3.5% wage growth is more like... stable, not overheating. Good enough to avoid imminent recession, not strong enough to justify a full-on growth party.

AVA: So then what does that mean for the Fed? Because over on RenMac Off-Script and Forward Guidance, Dutta’s conclusion sounded kind of awkward. Not dovish, not hawkish... just awkward.

GUY: That’s the right word. The economy isn’t weak enough to beg for cuts. Inflation is still above target. Equities are near highs. So the Fed’s natural posture is... sit tight and stay inflation-vigilant. Higher for longer, but not because growth is amazing. More because the Fed has no reason to rescue anyone.

AVA: Which is almost worse for markets in a weird way. Because if the economy were clearly rolling over, people would price cuts. If the economy were clearly booming, people would price earnings acceleration. Instead we’ve got a narrow growth engine, sticky-ish inflation, and a Fed that can’t really help.

GUY: Exactly. It’s like being in a car with one working headlight and the driver saying, “Good news, the engine’s still on. Bad news, I’m not changing course.” That’s basically the setup.

AVA: And then over on Macro Voices, Louis-Vincent Gave and Patrick Ceresna put a second layer on top of that, which is the energy layer. Because if your one healthy macro engine is data-center and AI capex... that engine runs on electricity, diesel logistics, steel, concrete, copper, transformers, semis. It runs on the physical economy.

GUY: Yes. And this is where I think the market is being weirdly complacent. Over on Macro Voices, Gave argued the oil curve is still pricing a fairly quick normalization of the Gulf shipping mess. He pointed to near-term crude in the high-$90s to low-$100s, while six-month oil was more like $80 to $85. That curve shape basically says: don’t worry, Hormuz reopens, logistics normalize, we move on.

AVA: And his whole point was... hold on though, why exactly should we assume that? Because if you look at the incentives, they may actually favor prolonged disruption. He floated the idea that Iran can extract shipping rents... he cited roughly $2 million per ship, and if you’re talking around 100 ships, the annualized economics get huge relative to Iran’s GDP.

GUY: That number is wild. And to be clear, as Gave framed it on Macro Voices, that’s an incentive-based scenario, not proof of formal coordination. But it matters because markets keep assuming everybody wants normalcy. Not necessarily. If Iran can monetize the chokepoint, and if Saudi can sell fewer barrels at much higher prices, the incentives get messy fast.

AVA: Right, because over on Macro Voices, Gave basically said Saudi might rationally prefer exporting 4.5 to 5.0 million barrels per day via the Red Sea at very high prices rather than pushing 8 to 9 million barrels in a lower-price environment while effectively paying rents into a hostile chokepoint. That’s not the default market assumption.

GUY: No, the default market assumption is that all producers always want maximum volume. That’s lazy thinking. Sometimes producers want maximum rent, not maximum flow.

AVA: And then the timing matters even more than the headline price. Over on Macro Voices, Gave’s real warning was that inventories, storage, and already-departed tankers have delayed the signal. So the market can look calm longer than it should. But by late May into early June, those buffers start running down.

GUY: Which lines up almost perfectly with what Marco was saying over on Geopolitical Cousins. He made the same lag argument from a different angle. Existing inventories make the shock look manageable... until suddenly they don’t. It’s like living off what’s already in the pantry and convincing yourself the grocery store isn’t closed.

AVA: Exactly. And both podcasts converged on the same escalation ladder. Over on Macro Voices, Gave said around $100 WTI is uncomfortable and inflationary but not punitive. Around $120 to $130 starts to bite materially. And $200 is economic devastation. Over on Geopolitical Cousins, they framed something similar: roughly $120 in May, $130 in June, and then the $150 to $200 range is where you stop talking about inconvenience and start talking about systemic breakage.

GUY: I agree with that ladder. People throw around oil numbers too casually. There’s a huge difference between “higher gasoline prices annoy voters” and “energy starts breaking the broader economy.” The latter happens when businesses can’t pass through costs, freight gets distorted, consumer confidence cracks, and central banks are trapped. That’s not at $95. That starts to get real north of $120.

AVA: Wait really, you think $120 is enough to do real damage in 2026? I think there’s more buffer than people realize.

GUY: I think $120 by itself is not doom. But $120 when your equity market is concentrated, when your macro growth is narrow, when your AI buildout depends on power and industrial inputs, and when the Fed can’t cut... yeah, then it matters a lot more. Oil doesn’t have to kill the economy on its own. It just has to hit the weak points.

AVA: That’s fair. And over on Macro Voices, Gave broadened that into what he called a “stockpile world,” which I thought was one of the more important macro ideas in the whole brief. His argument is that countries and companies can’t rely on financial reserves the way they used to. Treasuries are nice, but if shipping security is less reliable and drone warfare makes maritime control harder, then you need physical inventories... oil, gas, fertilizer, food inputs.

GUY: His line was great. You can’t sprinkle Treasuries on your field to grow wheat, and you can’t shove Treasuries into your car. That’s exactly right. We spent decades in a just-in-time global system where financial claims felt good enough. Now we’re moving toward just-in-case physical control.

AVA: And China, according to Gave on Macro Voices, may already be the model here. He said China officially reports about 1.3 billion barrels of oil reserves, but his firm estimates the true figure closer to 1.8 billion. More than the rest of the world combined, in his telling. That means China can buy on weakness, step back during spikes, and manage shocks better than countries that treat inventories as dead capital.

GUY: That’s a huge strategic advantage. And he also noted Chinese March oil imports were still up 8% year over year, which suggests preparation rather than panic. Again, that’s the kind of detail people ignore because it’s less exciting than a missile headline.

AVA: Over on Macro Voices, Patrick Ceresna also gave a specific trade expression for this restocking regime. Instead of trying to pick one commodity, he liked the Invesco DB Commodity Index Tracking Fund, ticker DBC, around 29.80, and specifically mentioned the January 15, 2027 $30 call around $3.

GUY: I actually think that’s elegant. You get convexity if the structural commodity thesis is right, and your downside is defined if the geopolitical premium fades in the near term. Because here’s the problem with chasing front-month oil headlines... you can be right on the thesis and still get wrecked on timing. DBC gives you a broader restocking basket and a longer horizon.

AVA: Although over on Excess Returns, Edward Chancellor would probably say if you’re making one big cross-asset statement, the cleanest expression is still gold.

GUY: Yep. And he put it well on Excess Returns... gold is “an asset without a liability.” In a world where U.S. equities are expensive, debt dynamics are deteriorating, and bonds may not be the old shock absorber anymore, gold becomes a very rational portfolio anchor.

AVA: Especially because over on Excess Returns, Chancellor’s bigger claim is that the secular bond bull market likely ended in 2021 or 2022. He reminded listeners that long-rate cycles can run 30 to 40 years. So the idea that we’re just going back to the pre-2022 world of ever-lower rates... he basically thinks that’s wishful thinking.

GUY: I think he’s right. People still act like zero rates were natural law and 5% nominal yields are some weird temporary punishment. No. The weird thing was the prior era. And if we’re in a structurally higher-rate world, a lot of valuation frameworks break... especially for long-duration assets, private markets, housing, and any business model that depended on free money.

AVA: Chancellor also made a really interesting point on Excess Returns that even after rates rose, there’s still a lot of liquidity sloshing around because central banks printed roughly $8 trillion during the COVID era. So you can have speculative booms and capex bubbles coexisting with higher rates. That sounds contradictory until you remember that the system is still digesting this giant liquidity aftershock.

GUY: Right. Higher rates don’t instantly erase excess capital. They just change where it hurts first. Public markets adjust fast. Private stuff pretends longer. Which is why Chancellor also warned on Excess Returns that private credit problems are still “percolating,” and that U.S. housing is basically stuck because prices haven’t adjusted enough to the new rate regime.

AVA: That stuck-housing point is underappreciated. It’s not a crash story. It’s a paralysis story. Owners don’t want to sell because their financing is too good. Buyers can’t afford the new payment math. So housing becomes an illiquid museum of old mortgage rates.

GUY: That’s perfect... an illiquid museum. And it matters for macro because housing used to be a broad transmission channel. If housing is jammed, then again, you’re back to the one-engine economy. AI capex is doing the running while big legacy cyclical channels are clogged.

AVA: One more market point from Macro Voices before we move to tech... over there, Patrick Ceresna said equities are ripping to 52-week highs, semiconductors are leading, but breadth is only around 50 to 55%. That’s not a healthy broad rally.

GUY: No. That’s a head fake unless you know what’s under the hood. If only half the market is participating, then the indexes are basically being levitated by the narrative sector of the moment. And over on Macro Voices, Gave sharpened that by noting semis have gone from roughly 10% of the S&P 500 two years ago to 17% now. He compared that to energy rising from about 10% in 2006 to 16% by 2008 during peak-oil mania.

AVA: That comparison made me pause. Because he’s not saying semis are fake the way people used to say dot-coms were fake. He’s saying transformative themes can still become dangerously concentrated trades.

GUY: Exactly. The bubble question isn’t “is the technology real?” Railways were real. Cars were real. Aviation was real. Telecom fiber was real. The question is whether investors overfund the infrastructure, misidentify the winners, and pay insane multiples at the point of peak narrative confidence.

AVA: Which is a perfect bridge into tech and AI, because that debate was basically the whole day.

GUY: Let’s go there.

AVA: So over on Invest Like the Best, Anthropic CFO Krishna Rao gave what I think is the most honest explanation of frontier AI economics you’ll hear from someone actually inside the machine. He said compute is “the lifeblood of our business” and “the canvas on which everything else gets built.” That’s a very revealing phrase. It means compute is not an input to the business. It is the business.

GUY: Yes. And I liked that because it kills the lazy SaaS analogy. People still talk about these labs like they’re software companies with unusually large cloud bills. No. They’re something closer to utilities crossed with research labs crossed with hyperscale infrastructure operators.

AVA: Exactly. And over on Invest Like the Best, Rao explained why compute procurement is so brutal. Buy too much and you can bankrupt yourself. Buy too little and you can’t serve users, you can’t train frontier models, you fall behind, and maybe you die anyway. There’s no comfortable middle when demand and capability are both moving exponentially.

GUY: His “cone of uncertainty” framework was the key thing for me. On Invest Like the Best, he basically said small forecasting errors compound enormously over one to two years, while supply is constrained by long procurement cycles. So if you’re wrong by a little today, you’re wrong by a lot later... and you can’t just order a gigawatt of compute next Thursday to fix it.

AVA: Right. This is why I think AI infrastructure planning now looks more like utility planning than cloud budgeting. It’s a long-lead, physically constrained, capital-intensive capacity problem with nonlinear demand and very high strategic cost of underbuilding.

GUY: And over on Invest Like the Best, Rao also said Anthropic has tried to reduce some of that risk through platform diversification... Amazon Trainium, Google TPUs, and Nvidia GPUs. Plus they’ve apparently spent years building orchestration software so they can use these chip families more fungibly.

AVA: Which is strategically huge. Because if you’re dependent on one vendor, one cloud, one chip architecture, you don’t have a business... you have a hostage situation. Multi-chip flexibility gives you pricing leverage, supply flexibility, and maybe some resilience if one ecosystem tightens.

GUY: Though let me push back a bit. I like the strategy, but I think Nvidia still matters far more than people want to admit. The whole market keeps looking for a “see, they don’t need Nvidia” narrative. But over on Big Technology Podcast, the Anthropic-SpaceX deal showed where the rubber meets the road. They’re taking all of SpaceX’s Colossus 1 capacity... described there as 300 megawatts and 220,000 Nvidia GPUs. That is not a “vendor diversification solved it” story. That is a “give me every high-end GPU you can find” story.

AVA: I don’t buy that as a contradiction, though. Over on Big Technology Podcast, Alex Kantrowitz’s point was that this capacity deal had immediate product consequences. Anthropic doubled Claude Code 5-hour rate limits, removed peak-hour reductions, and lifted API ceilings. That suggests the bottleneck was real and operational. Not theoretical. Not just “we’d like more hardware someday.” Real users were running into real walls.

GUY: Fair. And that’s important because when people say AI demand might be overstated, bulls can now point to visible usage constraints and say, no, look, once the infrastructure showed up, the product gates moved right away.

AVA: But then over on Big Technology Podcast, Ranjan Roy gave the bearish version, which I think is also worth taking seriously. He warned that some of the last 6 to 8 months of enterprise AI usage may have been “token maxing”... basically loose budgets, experimentation, prestige projects, fear of missing out. Companies spending because they can, not because the workload is durable.

GUY: And that’s the crux. Short-run scarcity does not automatically prove long-run economics. Railroads were scarce until they were overbuilt. Fiber was scarce until it was dark fiber everywhere. The fact that something is scarce in year one tells you almost nothing about whether investors have overfunded it by year four.

AVA: Which is why Edward Chancellor’s argument on Excess Returns landed so hard. He wasn’t doing the dumb “AI is fake” thing. He explicitly said real technologies can still produce terrible shareholder outcomes. He cited British railways, early autos, aviation, and late-1990s telecom. The pattern is familiar: real technology, real use cases, too much capital, poor winner selection, overbuilding, shakeout.

GUY: That’s right. And he added another layer that I thought was very sharp. On Excess Returns, Chancellor said capex waves mechanically inflate reported profits because one company’s spending becomes another company’s revenue. So when the whole ecosystem is in build mode, everyone looks richer. But some of that richness is self-referential.

AVA: And then the accounting critique. Over on Excess Returns, he pointed to reports that GPU depreciation schedules have stretched from about 3 to 3.5 years out to around 6.5 years. That’s a huge deal. Because if the hardware is economically obsolete in three years but the accounting says six and a half, current earnings can look a lot prettier than true economic returns.

GUY: This is one of those boring details that can matter a lot in 2026 and 2027. If useful life assumptions are too generous, then profits are being flattered right at the point of peak investment enthusiasm. That’s classic cycle behavior.

AVA: I’ll push back slightly there. Extending depreciation isn’t automatically fraud or fantasy. If the secondary workloads improve and hardware utilization broadens, some chips may indeed have longer usable lives than people first assumed.

GUY: Sure. But the burden of proof is now on managements. Because frontier hardware obsolescence is brutal. If a next-generation chip gives you massively better performance per watt, your old hardware doesn’t need to be physically broken to be economically dead.

AVA: That’s fair. And Chancellor also went even deeper on Excess Returns into capability skepticism. He said he’s “very, very skeptical” of strong AGI or singularity claims built on large language models, basically because they’re next-token systems. He cited model-testing results with the best systems still around 2% hallucination rates, plus examples of expensive enterprise failures.

GUY: Which matters because a lot of these valuations assume not just growth, but mission-critical adoption. And mission-critical adoption requires reliability. If your error rate is 2% on legal drafting or code generation or medical workflows, that can still be catastrophic depending on the context.

AVA: Although over on 20VC, Patrick Forquer made the application-layer bull case in a very concrete way through Legora. And I think his argument is worth hearing because it’s more sophisticated than “AI is cool.” He said people are using the wrong market denominator. Legora isn’t targeting just the roughly $40 billion legal-tech market. It’s going after the roughly $1 trillion legal services market, because AI can displace billable work, not just software budgets.

GUY: That’s the best bull case for AI applications in one sentence. The TAM is not the software line item. The TAM is the labor pool behind it.

AVA: Exactly. And the numbers he gave on 20VC were kind of nuts. More than $50 million of qualified pipeline in one month after the Jude Law campaign. Pilot-to-closed-won conversion of 78%. Typical ACVs in the $250,000 to $500,000 range, with some high seven-figure deals. Growing from about $3.5 million ARR to $70 million ARR in roughly a year, then crossing $100 million ARR early the following year.

GUY: Those are venture-capital catnip numbers.

AVA: They are. But they also imply something deeper... if they hold up. If AI apps can genuinely capture service spend with that kind of velocity, then the revenue pool is much bigger than legacy software comps suggest.

GUY: I agree, but I’d add a giant caveat. Over on 20VC, Forquer also said comp plans were running around 8x to 12x quota-to-OTE, and average attainment last year was 280%. He framed that as basically the business outrunning the plan. Maybe. Or maybe it means the whole market is in a heat-distortion phase where normal SaaS efficiency metrics don’t mean what they used to.

AVA: Right. In other words, these numbers may be real and still not be normalizable. Hypergrowth inside a gold rush can tell you something important about demand, but not necessarily about what the steady-state business will look like.

GUY: Exactly. One of the most dangerous assumptions in a boom is taking this quarter’s urgency and treating it like a 10-year annuity. That’s where investors get killed.

AVA: So if we synthesize the tech section... over on Invest Like the Best and Big Technology Podcast, the bullish evidence is that compute scarcity is real, product bottlenecks are real, and infrastructure procurement is now strategic. Over on Excess Returns and Big Technology Podcast, the bearish evidence is that real scarcity can coexist with eventual overbuild, accounting may be flattering returns, and some current usage may be prestige spending.

GUY: That’s the right synthesis. I’d phrase it even more simply... the bulls have operational evidence, the bears have historical precedent. And both can be right at the same time.

AVA: Which brings us to geopolitics, because a lot of this turns on whether the physical system stays functional long enough for the AI buildout to keep compounding.

GUY: Take it.

AVA: So over on Geopolitical Cousins, the key argument was that the Hormuz crisis is being misread if you’re only looking for immediate collapse. Marco’s point was that the real issue is delayed impact. Inventories and reserves mean markets can stay calm longer than they should. But once those buffers run down, the repricing can be much more violent because it becomes about physical scarcity, not just headlines.

GUY: Which again rhymes with Gave on Macro Voices almost perfectly. Late May to early June is the window. That’s when pantry economics gives way to supply-chain reality.

AVA: Yes. And over on Geopolitical Cousins, Marco interpreted the Iranian strike near Fujairah as a calibrated signal. Not all-out war, but a demonstration that bypass infrastructure is still vulnerable. That matters because a lot of comfortable Western analysis kind of assumes, “Well, if Hormuz is messy, flows will just reroute.” Maybe some do. But if alternate nodes are vulnerable too, then reopening transit becomes a negotiation problem, not just a military escort problem.

GUY: That’s a really important distinction. Logistics networks are only resilient if there are safe substitutes. If the substitutes are also targetable, then the system is much less redundant than people think.

AVA: Over on Geopolitical Cousins, the China angle got even more interesting. Jacob said Beijing instructed companies not to comply with U.S. sanctions on five Chinese refiners linked to Iranian crude, citing a 2021 blocking measure against what China sees as unjustified foreign laws. That is China openly resisting secondary sanctions.

GUY: And Marco’s interpretation there was brutally practical. China is basically calling Washington’s bluff. Because actually removing Iranian supply from the market would be inflationary and painful for both China and the U.S. So sanctions are becoming less about moral signaling and more about who can force whom to absorb the inflation.

AVA: Exactly. That’s why I think the upcoming Trump-Xi summit matters more than people realize. Over on Geopolitical Cousins, Jacob argued the summit shouldn’t be read as merely an Iran meeting. He said Trump basically sees China as “a deal to be made,” and that this could look more like a return to the unfinished logic of the 2019 Phase One approach than a pure security confrontation.

GUY: In plain English... if Washington needs industrial disinflation, supply-chain relief, and maybe a softer enforcement stance around Iran-linked flows, then it may have to choose pragmatism over maximal confrontation.

AVA: And that choice is bigger than tariffs. It affects semiconductors, energy, defense inputs, grid buildout, solar, rare earths... pretty much everything the AI economy needs.

GUY: Over on Macro Voices, Gave made that exact point from the market side. He suggested semis may be rallying not just because “AI good,” but because markets are starting to price a broader U.S.-China-Gulf recalibration. If Gulf data-center projects look vulnerable or uninsurable, more capacity comes back onshore in the U.S. And if more capacity comes back onshore, then the U.S. suddenly needs more power, probably more solar, Chinese panels, rare earths, and some form of détente around industrial inputs.

AVA: That’s the dot-connecting exercise people need to do. An oil chokepoint story can become a U.S. power-buildout story, which becomes a solar-import story, which becomes a China-policy story, which loops back into the AI capex story.

GUY: And that’s why I don’t love the clean “decouple from China no matter what” rhetoric. Over on Geopolitical Cousins, Marco’s structural point was that the U.S. cannot simultaneously absorb an Iran oil shock, rebuild defense and logistics capacity, and run a maximal anti-China policy. That combination is just too inflationary.

AVA: I agree. And honestly, this is one place where market people can be more realistic than political people. Markets understand constraints. Politics often performs through absolutes. But if China remains a source of excess industrial supply and disinflation, then some degree of selective accommodation is just arithmetic.

GUY: Right. Arithmetic beats ideology eventually.

AVA: One more geopolitical note from The Diary of a CEO... Anne Applebaum framed Trump-related enrichment as a market risk, not just a moral complaint. She said Trump’s reported net worth rose from $2.3 billion to $6.5 billion in about two years, and she pointed to the $2 billion Saudi investment in Jared Kushner’s fund as an example of influence-seeking through proximity to power.

GUY: That matters because if diplomatic and commercial decisions start looking like family-office monetization opportunities, then cross-border capital allocation gets distorted. Investors have to ask: are these contracts, licenses, and investments being awarded on economic logic or on political access?

AVA: Exactly. And once that distinction blurs, your risk premium should go up. Not because corruption is new, but because the transmission from geopolitics to capital allocation becomes less predictable. A normal market can price policy. It’s much harder to price personalized patronage networks.

GUY: Which, by the way, also feeds back into AI financing. Over on Big Technology Podcast, there was speculation that Gulf-state capital may already be more hesitant after the Iran war. If that’s true, some of these giant AI firms may end up leaning toward public markets sooner than expected.

AVA: That would be a fascinating shift... because then public-market investors inherit more of the geopolitical financing risk that private capital might have absorbed before.

GUY: OK... let’s pull the threads together, because this is where the whole thing gets interesting.

AVA: Over on RenMac Off-Script and Forward Guidance, Neil Dutta gave us the one-engine economy. Over on Macro Voices, Patrick Ceresna gave us a narrow equity market with only 50 to 55% breadth. Those are the same phenomenon in two different languages.

GUY: Exactly. The economy is concentrated in one capex ecosystem, and the market is concentrated in one narrative ecosystem. That can keep working for longer than bears expect... but it also means your failure modes are highly concentrated.

AVA: Over on Invest Like the Best, Krishna Rao’s “cone of uncertainty” for compute procurement sounded eerily similar to Gave’s “stockpile world” on Macro Voices. Different domains, same logic. You can’t rely on just-in-time access anymore. You need physical control over scarce capacity before you actually need it.

GUY: Yes. That’s one of the deepest connections in the whole briefing. Sovereigns stockpiling oil and fertilizer... AI labs locking down megawatts and GPU fleets... it’s the same strategic move. Everyone is shifting from financial optionality to physical optionality.

AVA: And the China angle sits right in the middle. Over on Macro Voices, Gave even floated a pretty bullish RMB scenario... something like 5 to 8% annual appreciation over a three-year window. That’s not just an FX call. It implies a broader world where China is less isolated, capital rotates toward Asian yield assets, and some U.S.-China thaw emerges.

GUY: He also mentioned names like PetroChina and China Mobile in that context, plus life insurers. Which is interesting because it’s not the usual flashy China-tech pitch. It’s more like: if the world revalues physical assets, energy security, and income streams, then boring Asian yield names may benefit.

AVA: Over on Excess Returns, Chancellor’s anti-bubble framework fits here too. His point was that you don’t necessarily short the mania. You buy what the mania starves. He used 2020 to 2021 energy as the classic example, when energy fell to around 2% of the S&P 500 and Tesla’s market cap exceeded the entire listed North American energy sector.

GUY: That was one of the great absurdities of the last decade. And he gave Garrett Motion as an example... around 40% market share, 60% incremental share, trading near 7x earnings, later up roughly 150% in his telling. The lesson isn’t “buy turbochargers forever.” It’s that narrative disruption often leaves perfectly viable old-economy assets dramatically mispriced.

AVA: He extended that on Excess Returns into consumer names too... Pernod Ricard, Campari, Diageo, Rémy Cointreau. Businesses derated on pandemic inventory issues, GLP-1 fears, and generational consumption anxieties, even though they’re long-duration brands with real scarcity and pricing power.

GUY: Which reminds me of something from Acquired. Over on Acquired’s Ferrari episode, the whole lesson was scarcity economics. Ferrari sells only around 14,000 cars a year, caps the SUV or FUV mix near 20%, allocates roughly 80% of cars to existing owners, and therefore admits fewer than 3,000 new customers per year. That scarcity is why Ferrari can have a market value above Ford, Volkswagen, Honda, Stellantis, even Mercedes in some windows... despite tiny unit volume.

AVA: And that’s actually a beautiful contrast with AI. Ferrari gets paid for restricting abundance. AI gets paid, at least for now, on the promise of enormous future abundance. One reason Ferrari-style economics are so durable is that they’re already controlling scarcity today, while AI names are being valued on expected future ubiquity.

GUY: That’s a great point. Investors may be paying peak prices for the dream of future intelligence while underpricing businesses that already have pricing power because they own something scarce, tangible, and hard to replicate.

AVA: There’s also a behavioral layer here. Over on The Compound and Friends, Haley Saxs talked about “learned financial helplessness,” and Josh Brown argued that younger investors increasingly turn to BNPL, options, sports betting, and speculation because long-term compounding feels inaccessible when everything costs a thousand dollars.

GUY: That line from Brown was dead-on... “everything costs $1,000.” One home repair, one medical bill, one car issue, and your savings discipline is gone. That changes how retail participates in markets. Not as patient owners of productive assets, but as people reaching for asymmetry.

AVA: Meanwhile, over on Capital Allocators, the institutional world is moving the opposite direction... more engineered, more customized, more sophisticated. They talked about separately managed accounts being used to access single-PM talent and replicate pod-shop infrastructure more efficiently. Walleye was described as a $12 billion multistrategy business, Docside as working with more than 60 managers, and UTIMCO as using the model for about a $7 billion GMV alpha pool inside an $11 billion zero-beta hedge-fund portfolio.

GUY: That contrast is kind of incredible. Retail gets more gamified and thematic. Institutions get more private, more modular, more optimized. So public markets end up with a strange mix... speculative retail flows chasing stories while serious capital quietly builds custom structures away from the crowd.

AVA: Which may be one reason concentration persists. Because the institutions that might otherwise arbitrage some of this are increasingly doing their best work in bespoke structures, not in broad public benchmarks.

GUY: And over on Odd Lots, Jamie Dimon made the country-level version of that argument when he talked about the UK. He said growth needs proper tax policy, proper investment policy, consistency of law, and an explicit strategy for sectors like financial services, biotech, and tech. The deeper message was simple: capital goes where the rules are durable.

AVA: Right. National competitiveness is capital allocation too. If policy credibility is weak, the country gets a higher risk premium. If it’s strong, you attract long-duration investment. Which is directly relevant when every major power is now competing for data centers, fabs, grid investment, and supply-chain relocation.

GUY: So the full picture today is... AI capex is both growth engine and fragility concentrator. Oil and shipping are not side stories; they are power-input stories. China is both rival and balancing item. And market concentration is no longer just a valuation concern; it’s a macro vulnerability.

AVA: Exactly. So let’s finish with what we’re watching, because the calendar matters a lot here.

GUY: First, over on Macro Voices and Geopolitical Cousins, late May into early June is the big physical-market checkpoint. If Hormuz disruption persists, that’s when inventories and already-in-transit barrels may stop masking the shortage. Watch crude, watch freight, watch inflation breakevens, watch whether risk assets finally have to price physical scarcity instead of just geopolitical theater.

AVA: Second, over on Macro Voices and Geopolitical Cousins, watch the June oil path specifically. If crude stalls in the $120 to $130 range, that’s painful but maybe manageable. If it starts pushing into $150 to $200, that’s a different regime. Then you’re talking about recession odds, policy panic, and potentially forced demand destruction.

GUY: Third, over on Geopolitical Cousins, the Trump-Xi summit may matter more than the Iran headlines themselves. If it produces even a partial détente... around trade, sanctions enforcement, industrial inputs, semis, whatever... that could ease inflation pressure across multiple fronts. If it fails, you could combine oil stress with tariff stress and supply-chain stress, which is exactly what markets do not need.

AVA: Fourth, over on RenMac Off-Script and Forward Guidance, the next one to two U.S. jobs reports and the summer wage prints are critical. If hiring broadens beyond AI-linked construction and durable manufacturing, that would make the economy look less one-engine. If wages stay around 3.5% and monthly prints stay near 0.2%, that supports the “stable but narrow” interpretation. If wages reaccelerate, the Fed gets tougher.

GUY: Fifth, over on Big Technology Podcast and Invest Like the Best, watch the Anthropic-Colossus 1 rollout over the next several quarters. The first visible effect is already there: higher Claude Code limits, fewer peak-hour reductions, bigger API ceilings. But the real question is monetization quality. Does this new capacity translate into durable paid enterprise usage... or did it just reveal how much recent demand was budget-slack token burning?

AVA: Sixth, over on Excess Returns, I’m watching the 2026 to 2027 AI capex accounting cycle very closely. Useful lives, depreciation schedules, utilization disclosures, impairment language... all of that. If hardware obsolescence outruns accounting assumptions, some of today’s apparent profitability in the AI stack could be overstated.

GUY: Seventh, over on Macro Voices, I’m watching the RMB and Asian yield equities. If Gave’s call of 5 to 8% annual RMB appreciation over a three-year window starts to look plausible, then you probably also get renewed interest in Asian dividend and hard-asset names... PetroChina, China Mobile, maybe selected insurers. That would be a signal that markets are moving toward a more pragmatic U.S.-China regime.

AVA: Eighth, and this is more of a market-structure watch... over on Macro Voices, keep an eye on breadth. If semis keep leading and the S&P keeps printing highs while breadth stays stuck around 50 to 55%, that tells you concentration risk is still increasing. If breadth improves, that’s healthier. If it worsens, indexes may be saying “all clear” while internals are saying “single point of failure.”

GUY: And ninth, over on Excess Returns and Acquired, I’m watching the anti-bubbles. Not because I think AI collapses tomorrow, but because crowded winners often create neglected opportunities elsewhere. Energy, old-economy suppliers, selected consumer brands, scarcity businesses... places where the narrative has maybe overshot the underlying economics.

AVA: Yeah. The big mistake right now would be treating this as a simple binary... either AI changes everything or it’s a bubble. Both can be true in different time horizons. The better question is: which pieces have durable economics, which pieces are overfunded, and which ignored assets benefit from the distortion?

GUY: That’s the whole game. Not “is AI real?” Of course it’s real. The real question is duration, winners, and constraints. How long does demand stay strong? Who actually captures the profits? And what physical bottlenecks hit first... power, chips, shipping, oil, financing, politics?

AVA: And if you get the answers wrong, you can be directionally right and still lose money. Which, honestly, is the most 2026 thing imaginable.

GUY: That’s wild, but true. All right... that’s the show.

AVA: We’ll be watching the oil buffers, the jobs data, the Trump-Xi setup, and whether AI demand starts to look more like durable utility usage or more like enterprise sugar high.

GUY: Thanks for starting your morning with us. I’m Guy.

AVA: I’m Ava. See you tomorrow on Morning Signal.