We're in an AI bubble, just a different one than you think. BofA ranked the biggest bubbles of the past 190 years by how completely a single theme took over the market at the peak. We’re back at those exact peak levels. The "AI Big 10" (the Magnificent 7 plus Broadcom, AMD, and Micron) is at 41% of the US market today, level with the dot-com peak. No one wants to hear we’re living in a bubble; they’d much rather be reassured that this time is different. I think it is a bubble, but I also think this time really is different. Let me explain why. I think the froth is in the labs, and the infrastructure trade is intact. The funny part is that the labs aren't even in this chart. OpenAI and Anthropic are private. That 41% is in chipmakers and cloud giants, highly profitable businesses still trading at reasonable valuations. These are not wagers on any one lab winning. They are a broad bet that AI, as a foundational technology, takes off, and the odds of that are super high. I think the labs are the weak spot. The model is becoming the brain, and the value is moving to the harness around it: access to company data, memory, security, permissions, workflow automation, and the ability to take actions inside other apps. Meanwhile, the brain is getting cheaper. Open-weight models now account for roughly a third of routed AI traffic, and in some datasets that slice is growing even faster. Businesses are gravitating toward them because they can tailor the models, keep workloads in-house, and avoid shelling out top-tier API fees on every single job. When intelligence gets cheap and easy to swap, OpenAI and Anthropic can't live on selling model access alone. So they have to move down the stack, sell finished products for coding, cybersecurity, legal research, finance, sales, and customer support. OpenAI is pitching its enterprise platform as the layer that controls and governs a company's agents, on top of the model underneath. Anthropic is doing the same with Claude Cowork and industry tools that do work people used to do inside legal, finance, and enterprise software. That puts both labs in direct competition with their own customers. Wall Street has already shown how it reacts. Anthropic's enterprise agent tools helped set off a software selloff that erased about $1 trillion in market value. The next breakout launch could reignite that trade. Investors should be asking whether OpenAI and Anthropic will swallow software’s most valuable layers and slash the number of subscriptions businesses need. However, they will compete against very well-integrated players who already own the workflows, the customer relationships, the compliance guardrails, and the internal champions. These incumbents are deeply embedded in day-to-day operations, making switching costly, slow, and politically difficult. That’s why I’m not so sure about the future growth trajectory for the frontier labs or Anthropic’s expected IPO at $2T. The labs carry the bubble risk. Their core product keeps getting cheaper, so they're forced to move up the stack and fight software companies for the customer. The chips, power, and data centers don't need to pick a winner. Every model runs on them, no matter which one wins. I’m modeling how a market bubble bursting could impact my portfolio. What worries me is the knock-on effect. If those two giants stumble, a lot of the market could get pulled under with them since so many firms orbit around their success. If you agree, join Milk Road PRO to see how I’m trying to hedge against it:

Tesla's FSD Supervised is at 14.9 billion miles and will pass 15 billion by month's end. The chart is about to go vertical, and it matters more than you think. The first billion took about 3.5 years. The billion from 1B to 2B took 202 days. The latest billion, from 14B to 15B, is taking about 29 days, roughly 34.5 million miles a day. The pace has gone up about 7x in two years, and it will keep accelerating. Tesla also claims 7x fewer collisions than human drivers. From the start, Tesla diehards were convinced FSD could outdrive the average human, but to be fair, the evidence just wasn’t there in sufficient volume. Given how quickly things have moved lately, the dataset is now sizable enough, and it is Tesla’s strongest card to play with regulators, with Europe shaping up as the next proving ground. This is what's being decided. In April 2026, the Netherlands approved FSD Supervised for its own roads. It used a special exemption because EU rules don't cover a system like this yet. Since then, seven countries have recognized it nationally: the Netherlands, Lithuania, Estonia, Denmark, Belgium, Slovenia, and Czechia. Together they cover only about 12% of the EU's population. EU-wide approval needs a qualified majority in the EU's vehicle committee (the TCMV). That means at least 15 of 27 member states, representing 65% of the population. The math makes Germany, France, Italy, and Spain the swing votes. Without them, there's no path to 65%. The committee was expected to vote in October. It now won't vote until December, per Reuters. It has discussed the Dutch request at least twice, including in late June, but has not called a vote. The agenda gives no official reason, but reporting points to three concerns. - First, FSD sometimes exceeds posted speed limits to keep up with traffic. - Next, regulators question whether driver-attention monitoring holds up in dense city driving. - Last, most of Tesla's data comes from US roads, and regulators want to know whether it transfers to European cyclists, motorbikes, and old city layouts. Europe really excels at hitting the brakes on adopting new technology. - Tesla's FSD in Europe doesn't enable Mad Max mode or anything like that, so it won't speed up. - The second and third are really about software capabilities. If they still assume we have different cyclists, motorbikes, or layouts, Tesla will fail to adapt. I don't know what to tell them. France just woke up and only started its own on-road tests with UTAC in early September and wants those results before it takes a position. I think it’s not just Europe being cautious. Lobbying likely plays a role too. European carmakers have dominated for a long time, but they’re starting to lose ground as Chinese competitors expand, and Tesla, especially with FSD, would be a formidable competitor, too. Elon's reaction to the delay was one word: "sigh." The bigger point is that Tesla adds a billion miles every month, and every month the "US-only data" objection gets weaker as 7 European countries have already approved FSD anyway. Long TSLA, short EU. If you like my posts, consider following for more insights and market commentary.

$BSP is turning into a high-conviction pick for me. The only question is whether $33 a share leaves enough upside. Let me walk you through my model. Start with what they already own. Many apps, including Evernote, Vimeo, WeTransfer, AOL, Eventbrite, and now Airtable and Miro. Q2 revenue was $704 million with a 54% adjusted operating margin. If they never bought another company and just ran these for cash, they're worth about $18 a share after debt. So at the current $33, investors are paying roughly $15 a share for everything else. The real asset is the "deal machine". Each year they spend $4B to acquire new businesses for a 2.5ish revenue multiple. Here are the latest deals: Airtable: $1.285 billion, about 2.7x recurring revenue Miro: $1.355 billion for about $600 million of recurring revenue, roughly 2.3x (it was valued at $17.5 billion in 2021). Then they bring that business to their platform, restructure the organization, make some optimizations, and increase operating margins to 45%. Management only signs deals it expects to return 25% (unlevered IRR) a year, before any debt. That's the return on the whole deal: the margin gains after they buy, the growth, and what the business is worth at the end. My model assumes deals get pricier over time and the last ones return only about 22% IRR. The math still works. In my base case, the yearly deal budget grows from about $4.0 billion today to about $15 billion by 2036. They run at about 3x leverage (2.4x at the end of Q2). Every $1 of new profit lets them borrow about $3 more. That money funds the next deal, the deal adds profit, and the profit supports more debt. It might sound scary at first, but remember: they buy existing businesses with clear product-market fit and huge user bases. As a result, the machine is worth about $28 a share. The market pays $15, which only makes sense if they do not execute. The real risk is that deals get more expensive as checks get bigger, or they can't expand margins as they once thought they could. Management admits this. The prospectus says it's plausible they'll have to lower their return hurdles as they scale. So I built it into the model. In my base case, purchase prices rise from 2.5x to about 3.4x sales, and the last deals return only about 22%, below their own 25% hurdle. Even with that drag, I get $46 a share: $18 for the existing portfolio plus $28 for the machine. That's about 40% above today. I'll share my full model and its assumptions on the @milkroaddaily podcast, so make sure to tune in. If you want to get notified whenever I make a trade, join Milk Road PRO here:

LIT is up 5.9x since May. Is it still a buy, or has the rally run its course? I’ve been championing Lighter for a long time because I truly believe their product is simply better than everything else on the market. And honestly, their team is the best. When you’re investing into startups, that’s really what you’re underwriting. It’s also why Lighter can point to heavyweight backers like Founders Fund, Dragonfly, Robot Ventures, and Robinhood, alongside other major names. That is exactly what pulled me in, too. But none of the product metrics were moving in the right direction, so I sold right before the token really started running. I highlighted the day when I sold on the chart below. Since that point, the token has nearly 6x. It’s tempting to chalk that up to booming business momentum, but the numbers don’t support it. The core metrics stayed basically stagnant until mid‑August, when OI jumped to nearly twice its prior level mainly due to price increases. It’s a welcome step in the right direction, but it still feels a long way from earning a sixfold price hike. My investing philosophy is that fundamentals eventually drag price where it belongs. Once I truly understand a business, I can judge whether its earnings power is set to expand and that’s how I build my thesis. When an asset is driven purely by sentiment, you’re just guessing. That’s why I ask myself: if no new buyers show up for the next three years, what return do I get anyway? When a company’s fundamentals are truly solid, profits keep rolling in, and assuming I didn’t overpay on the way in, the payoff on my investment should be very compelling. They might kick off share repurchases, boost shareholder dividends, or acquire other companies to fuel the next leg of expansion. In crypto, token holders often push for buybacks, but unless they scale up over time (they usually don't, they’re unlikely to create sustainable demand pressure and only deliver a short-lived price bump. I’m revisiting the leading crypto projects to see if anything stands out as a real opportunity, but so far it’s been a dead end. Drop your favorite token below that's worth exploring, and I will take a look. Follow me for more updates. Or see my real-time portfolio here:

The Milk Road Macro Index just hit a 5-month low at -0.6. Last time it did this, the S&P 500 dropped 9%. It hasn't printed this weak since April. Last time we were here, MRMI ran to about -2.5 and sat there for weeks. The S&P 500 dropped 9% through that window. We're back in the red now. The index just punched through -0.5. That's the line where this thing flips to risk-off. The main drivers pulling this index down are growth impulses and market breadth, both of which have deteriorated significantly over the last few days. So is this the time to cut risk? At minimum, it’s worth paying close attention: yields are exploding, consumer spending is trending down, oil is approaching $100, odds of rate hikes are at 50%, and the S&P 500 is less than 2% from ATH. I’m trimming a few positions to reduce risk and build up cash. But I'm not treating this like the March washout until I see follow-through. I'm watching the next few MRMI prints and reading comments from our macro expert @BitcoinJesusETH. If you are worried as well and not sure how to navigate through these times, join MR PRO to see how 5 analysts are reacting to this turbulence: