20VC with Harry StebbingsThe New VC Game | CoreWeave Sell-Off Creates Panic | Revenue Multiples Out of Control
20VC with Harry Stebbings
VC has grown selective 0:00
Big funds now want billion-dollar positions, and that has narrowed what counts as interesting. Companies with strong growth, real revenue, and good teams can still get ignored if they do not fit the right story. The speakers say venture money is concentrating on a small set of hot names, while many other companies look unfundable even when they have solid numbers.
AI trade and CoreWeave 10:31
CoreWeave is described as a risky neocloud with heavy capex and debt, so its stock moves hard when the market doubts AI growth. The speakers say last week looked like an AI slowdown week, helped by noise around Anthropic numbers. But if end-user demand stays strong, they think the business can still work, since customers like Microsoft and OpenAI keep pulling for compute.
B tier assets 14:01
CoreWeave is treated as a B tier asset because it exists only while Nvidia and similar firms need it. The point is that B tier assets can still make money, but they should not be held as if they were the core winners. Oracle is placed in the same bucket because it is renting excess capacity and buying GPUs rather than selling enough software of its own.
Anthropic and routing 17:00
Anthropic’s numbers are read as strong but choppy. Revenue is growing fast, yet the pace is slowing from a very high base, and that makes later forecasts uncertain. The talk then shifts to Replit’s choice to make OpenAI the default model because it is cheaper for similar quality. That is taken as a sign that customers are cost sensitive and that model routing matters more than brand.
Build to prod 28:00
Code review is treated as the key software category, not code generation. The real problem is getting work from built and working in dev into production. People are building all day in tools like Cloud Code and Cursor, but reviewing, testing, and making it production-ready still takes far more time. The next wave of products may sit around memory, testing, and sandboxes, because that is where software delivery now changes.
Inference and valuation 31:30
Modal is described as a stronger bet than CoreWeave because it adds value beyond bare GPU rental, and its revenue growth looks striking. The inference market is said to be expanding fast, with a few providers already at several billion in run rate revenue and much more room to grow. Town is treated as a smaller, earlier bet. The speakers keep returning to the gap between private prices and public multiples, and whether growth can justify them.
Revenue multiples reset 41:30
The discussion turns to whether the market is pushing revenue multiples too far. Companies in marketing and AI are still growing, but public peers trade at low levels, often under three times revenue. The worry is that AI versions of weaker businesses may settle quickly into the same pricing as slower peers, even if growth stays strong for a while.
Cash and control 43:02
Bending Spoons is treated as a software version of Berkshire, buying modest assets and holding them for a long time. That leads into a broader point: 37signals showed that a bootstrapped company can throw off serious profit, and some founders may be better off raising less or none at all, because buying out investors later can be costly and awkward.
Returns and Tesla 55:03
The talk opens on venture and private equity returns, with Harvard and Cambridge named as strong cases, but the point is that the market still has not reached the wild 2021 level. It then turns to Tesla as a final standalone show, with Elon Musk’s stunts, valuation premium, and share giveaways seen as a way to soften criticism and build loyalty.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
