Excellent episode on @theallinpod. @friedberg analyzes the root cause of Europe's problem. We focused too much on immigration into social welfare systems from countries with an incompatible culture. That’s what’s now clashing in France, Brussels, and Italy. What we didn’t do, and what was common practice before 2000, is focus immigration on skilled workers in underserved areas. Now the system is fully backward. Taxpayers in trouble get no support from the social system because it is optimized for sugarcoating migrants. Our own population, when they can’t find jobs, get mostly sarcastic comments that they’re not trying hard enough, while the same system has no problem defending terrorists and criminals. It should be the other way round and needs to be fixed. Link: open.spotify.com/episode/5YN…

88% of the entrepreneurs behind VC-backed companies walk away with nothing. The top 1% take 79% of all exit value. The number has been circulating on LinkedIn since The Economist's Callum Williams shared the new paper by Eduardo Azevedo and Kent Smetters at Wharton, with Florian Scheuer and Min Yang in Zurich. Many readers take it as a verdict on venture capital. To me it shows the power law at work, and the pattern looks the same wherever money meets companies. Public markets follow it too. Hendrik Bessembinder tracked some 26,000 US stocks from 1926 to 2016 and found that 58% of them failed to beat Treasury bills over their lifetime. 1,092 companies, about 4% of the total, created the entire $35 trillion in net wealth, and some 90 of them produced more than half of it. Venture capital applies the same rule to the people who write the cheques. Blake Jackson and Ilya Strebulaev followed more than 100,000 professionals at US VC firms. Among the 12,151 VCs whose returns they could measure, about 120 people, the top 1%, produced 57% of all net profits, and the top 5% generated 90%. Only 38% of VCs with at least one investment ever recorded a success. Whatever the asset class, most companies create little value or destroy it. About one in ten drives the result, and a handful within that group delivers the majority of returns. VC funds build their portfolios on this arithmetic. The partners invest in ten companies, double down on the winners and stop funding the rest. One in ten is the planning assumption, which means nine companies in every cohort lose their backing. That is the most plausible reason 88% of these entrepreneurs see nothing at exit. Their investment case did not create the value the model requires. Entrepreneurs who pitch to VCs should understand this strategy before the first meeting. A term sheet is a contract to grow fast. It leaves room for two outcomes: win fast or fail fast. One company in the portfolio drives the result, and the fund discards the rest of the deck. Liquidation preferences, vesting and access to secondaries are all worth negotiating, and every entrepreneur should negotiate them. Better terms change who takes the money in a weak exit. They leave the rule of the game untouched. Understand the rule, and you get in and out of venture with less frustration. If you have raised venture money, at what point did you realise your investors planned for one winner in ten?

Nobody fails at 30 messages a day. They fail at the first three. New Year's Eve is coming and we will for sure see the next wave of big goals. "I want to run a marathon in March" is likely amongst them. And I already hear the laughter: "But you haven't run one mile since high school. How will you do that at 52?" Nobody will take it as an insult. The bubbly tastes good, people are happy, and most of the resolutions will be forgotten on January 2. And almost none of them survive February. It is the same dynamic in business. Take the nasty job of scaling sales in a business. Every company that wants to grow revenues needs one thing: more customers. Your personal network goes only this far, and even with all your friends and old contacts involved, there is a limit of potential customers. The only way to grow a business is the belittled and mostly hated cold outreach. Fun fact: Did you know that one of the early key investments in Coinbase is the result of cold outreach to investors? In March 2012, Brian Armstrong sent Garry Tan 0.05 bitcoin and a mail about his young company, Bitbank. Tan read on. His early checks, $300,000 in total, grew to more than half a billion dollars. But same as with running marathons, big revenue goals need a big number of cold messages and relationships. And that has a natural reflex: procrastination. When the goal is way too big for where the person stands today, the mind and emotions shut down, and the person does something else. Running marathons teaches that. In every marathon comes a breaking point where the mind simply cannot comprehend that there are still 10k left. Telling it "only one quarter left, three are done" doesn't work here. What's the antidote that gets the aspiring elite marathon runner to the finish line? Focus on small steps. When the goal is too big, like 100 million revenue from a base of 0, break it down. In marathons it means: OK, I focus now on the first 10 steps. When that works, 100 meters ahead, and once that is cleared, 200. You get the hang of it. In sales it is the same. Focusing on writing 100 cold mails, and after 8 hours of doing everything else, finding yourself having another coffee with another colleague and pushing the 100 cold mails to tomorrow, to the bigger goal of 200, is the best way to not get anything done. Do it like in a marathon at the point of a mental breakdown: Start small. From a base of 0, with 3. Three are easy. But they get you started. And when 3 every workday works for a week, go to 6 the next week, then 9, and so on. At that pace, 30 a day is ten weeks away. After 1 to 2 months the habit is formed and scaling becomes easier. In big endeavours it is not a shame to start with a small first step. On the contrary. It helps establish the habit that can later be scaled to big outcomes. What was the first small step on your last big goal?

6 million GPUs ordered. Nvidia can build 3 million. That is the gap behind Nvidia's new RTX PRO 5500 workstation card. Chinese customers have ordered more than 6 million units, while Nvidia expects to produce no more than 3 million worldwide in 2027. ByteDance, Alibaba and Tencent alone placed orders for over 3 million cards, worth more than $35 billion at list price. Nvidia has told some partners to expect 5 to 10 percent of what they asked for. Is AI a bubble? It depends. I am a huge fan of looking at supply and demand. Trump and Xi have decided to race for the data centre crown. Faster, bigger, better. Who will come out first and win the race? It reminds me of the space race between the Soviet Union and the US in the 1960s. This time the prize is neither orbit nor the moon. The prize is winning the AI race and owning the next generation of the internet. The LLMs are an early application, and I believe better models will replace them at some point, built on new algorithms that do more than predict the next token. Google wiped out a search engine industry that looked well established at the time. Who remembers AltaVista in 2026? All of these new ideas need one thing, supercompute. And the Reuters report shows that China's 2027 demand for this one Nvidia card is twice the global production capacity for the same year. When Nvidia cannot supply, I think buyers will go down the chain to other producers like AMD. Once those sell out, new suppliers will appear. Europe is not even competing in this race yet. It tries to regulate AI away and hopes the world will copy its superior rules. The world won't. At some point Europe will wake up and understand that whoever owns the data centres gives its own industries leverage and a competitive advantage. Those without them lose. Are we in a bubble? The supply and demand dynamics say no. Where do you think the overflow demand goes first?

Berlin, the Invest in Germany Summit. Jensen Huang shares the stage with Friedrich Merz. Nvidia invested almost a billion dollars in Germany over the last year. The Chancellor gets to ask the man who builds the hardware of the AI economy his own question. He asks how to regulate it. Huang answers like an engineer. Safety comes first, he says, and it is a matter of engineering. The industry must invent the tools that make AI safe, companies must follow best practices, and governments regulate the applications at the end of the chain if a gap appears. His clearest line: there is nothing to regulate if the industry does not thrive. Merz thanks him and gives his own view. He wants AI regulation at the European level, and better still worldwide. This exchange captures the state of Western European politics like no report I have read this year. The visionary and maker explains how to build. The Chancellor asks how to control what others build. The numbers show which mindset governs Europe today. Between 2019 and 2024, the EU enacted some 13,000 legislative acts, while the United States passed around 3,500. Europe has become world-class at writing the manual for machines that other continents build. Western European politicians used to think like Huang. In 1967, the governments of France, West Germany, and Britain agreed to fund a European airliner at a time when American manufacturers owned the market, and Franz Josef Strauss pushed Germany into the project that became Airbus. Twenty years later, telecom companies from 13 European countries signed the GSM memorandum in Copenhagen, and Europe set the mobile standard that the rest of the world adopted. Those politicians backed the builders first and wrote the rules once there was something worth protecting. If Western Europe wants to prosper again, its leaders need the mindset of Jensen Huang and need to drop the one Merz symbolises in this part of the conversation. Constant regulation, offered as the first answer to every new technology, chokes the European economy and society to death. Which European policy decision since 2000 would you put next to Airbus and GSM?