Toma转发了
History doesn't repeat, but it rhymes. Today's AI boom is unfolding a lot, just as car production did in the early 20th century. ????? Hear me out. ?? Startup Explosion → Brutal Consolidation In the 1900s, over 3,000 automobile companies emerged in the US. By 1929, the "Big Three" (GM, Ford, Chrysler) controlled 75% of the market. Today's thousands of AI startups (including us at Toma) will *probably* face the same mathematical reality. ?? Technology Wars On top of that, early cars saw fierce competition between electric, steam, and gasoline technologies. Gasoline won decisively in the 1910s. The current debates between home-grown vs. foreign models, LLMs vs. SSMs, and open- vs. closed-source will eventually have a decisive victory. ?? Capital Intensity Ford's assembly line innovation required massive capital but reduced the Model T's price from $850 to $290. Similarly, training today's frontier AI models requires billions in compute. The winners aren't just those with the best ideas, but those with sufficient capital to scale and iterate. ?? Market Cycles Matter The Great Depression ruthlessly culled auto manufacturers, leaving only the strongest standing. With fears of recession, a similar shakeout looms for AI startups that raised funds recently but haven't found sustainable revenue models. ?? Speed of Change What took decades in the automotive era happens in months with AI. ChatGPT reached 100M users in two months; the auto industry needed decades to reach similar penetration. This acceleration means both opportunities and threats materialize faster, giving companies less time to adapt. ?? Core Difference: Digital vs Physical Unlike cars, AI models can be distributed globally instantly at near-zero marginal cost. This fundamentally changes scaling dynamics and competitive moats. The barriers aren't manufacturing capacity but data, talent, and compute access. ?? Prediction By 2030, we'll look back at 2022-2025 as the "Model T moment" for AI – when the technology transitioned from novelty to necessity. Most AI startups from this era will be footnotes in history, acquired for talent or defunct. The handful of winners will become the economic engines of the next half-century, just as the auto giants defined the 20th century. Sounds gloomy. What can we learn from all this? Execution trumps innovation. Having the first AI breakthrough is like building the first car – historically interesting but commercially irrelevant if you can't build something people truly want.