Benchmark, Y Combinator, General Catalyst and SoftBank don't bet on companies. They bet on statistics - nine failures for one jackpot.
Venture capital sounds boring and respectable, but it's actually the roughest bet in finance. A fund takes money from pension funds, university endowments and wealthy families, spreads it across dozens of startups, and accepts upfront that most will burn out. Nine out of ten die quietly, with no obituary. The tenth has to become the next big player and return more than was put in - enough to cover the losses from the other nine. It's not an investment strategy in the classic sense. The player writes the rules and then plays by them.
Y Combinator, a San Francisco accelerator, funds startups at their most fragile moment, when they have only an idea, in exchange for a small stake in the company. Every quarter it launches a new batch of founders, gives them a modest sum and a few months of intensive mentoring, then shows them to a room of investors in a single day. It's not looking for a finished company. It's looking for raw material for the next stage of the chain - raw material the bigger funds will later buy.
Benchmark is an old-school Silicon Valley fund, known for early bets on eBay and Uber - for decades it deliberately kept its funds small to stay attentive to every company. This year it broke its own rule and raised a growth fund for the first time. It bet early on Cursor, a code editor with built-in AI, and on Cerebras, an AI chip company whose stake was already worth over three billion dollars at IPO. One right call erased all the wrong ones - exactly as the model predicts.
General Catalyst got in early on Anthropic, the company behind the Claude chatbot, and builds its image as smart capital from there: a fund that doesn't just pay - it pushes its portfolio companies into hospitals, banks and insurers, places where a sale takes months of negotiation, not a single demo call. That's why its specialty is exactly there, in regulated industries, where AI struggles to get in without connections and trust.
SoftBank, a Japanese conglomerate and owner of the Vision Fund, is the biggest and most unpredictable player in the whole system. From its stake in OpenAI, the company behind ChatGPT, alone, it reported a profit of forty-six billion dollars in a single year - a number that dwarfs the budgets of entire countries. Unlike Benchmark, SoftBank doesn't pick carefully. It bets giant sums on a handful of players and counts on scale itself to bring the win.
What stays hidden
Everything I've built so far has come slowly, with what's on hand. That's why the venture capital model is foreign to my instinct - to me, growth should be stable, not a lottery. But I understand the logic from the inside. I've watched agencies burn one client's budget to impress the next one. The difference is scale and stakes. Here we're not talking about one client's budget, but billions poured into companies that have no right to grow slowly, because the fund behind them has to cover the nine dead bets.
So when you read news about yet another AI company burning cash or promising impossible growth - ask who's behind it. It's not the product dictating the pace. It's the fund, which has to prove to its own investors that this particular bet is the tenth one.