Alibaba's AI Spending Spree, Concerns of Circular AI Financing

There is a lot going on: tech shares are down for a fifth session, Treasury yields rebounded after buyback news, and Brent nears $93 amid the Iran war. Alibaba is the headline, spending hard on AI as profit takes a hit.

Alibaba committed $10 billion to AI. Revenue grew 9%, but profit fell 75% and free cash flow hit a $6.6 billion outflow while it builds infrastructure and pushes Qwen. Cloud is bright; e-commerce is difficult. AI and AGI are Eddie Wu’s north star, but returns are not near-term.

AI demand reshaped memory and semiconductors, but good news is embedded in valuations. Spending remains an existential hedge, even as higher real yields reflect term premium and rate uncertainty. Software’s AI selloff lets investors separate durable businesses from old valuations. Nvidia is hold-your-breath, though disappointment is not my base case.

Prediction markets went from essentially nothing at our 2024 election launch to 6 billion July contracts, 20 times last year’s level. Sports ignited liquidity; financial and crypto contracts broadened it. We see federally governed products; states disagree. Courts will draw a boundary, and we will adapt and argue for the fullest outcome.

Meta paying Microsoft for Azure is the circular AI economy. Returns require use to spread, yet revenue remains concentrated among perhaps 10 or 20 giant AI companies. Coding assistance helps explain Meta’s API ambitions: cut vendor costs and sell service. Alliances and rivalries arrive together.

Chinese models are closing benchmark gaps, though benchmarks are not real-world performance. In a coffee-shop website test, Anthropic was expensive and excellent; cheaper models could be good enough. Buyers increasingly ask cost per task, not token price.

Our Series C lets Castellion move at maximum speed. Blackbeard is headed toward production next year, but nothing matters until we make thousands at low cost. North Star, the larger system, and interception are urgent. SpaceX taught us vertical supply chains and cost reduction can deliver scale and capability.

AI is a hundred-X opportunity; we still back paper-and-pencil ideas. Semiconductors need veterans, models need researchers, and agents draw younger builders. We make eight to 10 selective bets a year, seeking unicorns. At inception, it is the jockey, not racetrack: A-plus founders can pivot. We keep funds small.

Massachusetts is considering tough AI safeguards: labs would publish safety plans and assess catastrophic biological, nuclear, and cyber risks. The Senate has passed it; the House and governor remain. The state also wants to keep Harvard and MIT talent from heading west, while industry divides over the bill.

That is it for Bloomberg Tech. Treasury intervention has not stopped the Nasdaq 100’s fifth straight down session, and markets are still asking whether it is only a temporary fix. Four days down, one to go.