Milk Road Macro (@MilkRoadMacro)

AI costs are now outpacing human labor costs at big tech companies. Uber's CTO already blew through his entire 2026 AI budget in 2025. It cost more than the human workers in that department while tartup founders are bragging about their AI bills as a badge of seriousness. VCs are asking about token spend as a signal of commitment. The macro number: worldwide IT spending is expected to be up 13.5% this year to over $6 trillion. Most of the incremental spend is going straight to token costs and enterprise contracts with OpenAI, Anthropic, Google and others. The original premise was simple: AI cuts costs. Replace expensive humans with cheaper compute. The actual pattern emerging is the opposite. The companies most committed to AI are spending more. High token usage has become a signal of AI seriousness. Two possible interpretations: One: AI is delivering productivity gains faster than expected and the ROI already justifies the cost. Uber's CTO blew his 2026 budget in 2025 because AI was producing results worth more than what was budgeted. Two: AI has become a competitive arms race where you have to keep spending just to stay relevant, regardless of near-term ROI. Either way, the beneficiaries are the same: the companies selling the tokens. OpenAI, Anthropic and the infrastructure beneath them are capturing every dollar of this spending surge.

AI systems are teaching themselves skills they were never trained to have. Here's the example that Eric Schmidt explained: A Google AI was prompted in Bengali. A language it was never trained on. With only a small amount of prompting, it suddenly could translate the entire Bengali language. Nobody programmed that. It just emerged on its own. Schmidt calls this "the black box problem." You don't fully understand what the model learned. You can't always tell why it got something right or why it got something wrong. The field has theories but the honest answer is: we turned it loose on society before we fully understood it. His defense: "We don't fully understand how a human mind works either." This isn't just a safety conversation. It's a business one. The companies that solve interpretability, not just raw performance, are going to be worth an enormous amount. Understanding why a model does what it does is becoming a regulatory requirement and eventually a customer trust issue. Right now the AI race is won on benchmark scores. The next phase of the race gets won on explainability.

Bitcoin is the first new asset class in 170 years. The last new asset class before Bitcoin was oil which was discovered in the 1850s. @ricedelman believes that $BTC is the oil of our era. Here's why: Modern portfolio theory says you want as many non-correlated assets as possible. And over the last 16 years, Bitcoin hasn’t consistently moved in sync with stocks, bonds, or gold. That's exactly the property you want from a diversifier. You don't need to believe in Bitcoin to own it. You just need to believe in diversification.

Kevin O'Leary says Steve Jobs was one of the brutal guys in business. O'Leary was selling educational software to Apple in the late 80s. The brand manager at Apple wanted a $12 million research budget to survey teachers on what updates they wanted. Jobs exploded on her: "They don't know what they want until I tell them what they want." O'Leary pushed back. Called him an asshole. Said she had a point and maybe they should listen to the market. Jobs turned to him: "Is Apple not your number one OEM? Are you not making more money with us at higher margins than any other channel you have? Shut up and go do the work." It sounds like arrogance. But Jobs was right about the outcome every single time. He was right about the Mac. Right about iPod. Right about iPhone. Right about every product people said nobody asked for. The lesson O'Leary took away: The truly great product people don't survey the market. They understand something about human desire that the market hasn't articulated yet. And they execute on that vision relentlessly. That's a moat no competitor has been able to replicate in 40 years of trying.