Post

Google borrowed $25 billion, and 4 AI geniuses left — this race finally reveals its true face Yesterday, Alphabet did something. Issued $25 billion in bonds. Divided into 10 tranches, ranging from 2 years to 40 years, with the longest tranche yielding 1.3 percentage points higher than Treasury bonds. And then? Orders flooded in totaling $115 billion. More than 4 times oversubscribed. The market went crazy for AI bonds, but — Alphabet’s stock price fell 1.4% that day. The money was raised, but the market didn’t buy it. You might ask: $25 billion, is that a lot? Yes. But compared to Google’s AI bill, it’s not enough. At the end of July, Alphabet just raised its 2026 capital expenditure forecast to a record $205 billion. More than double 2025’s. What’s the cost? For the first time since its 2004 IPO, Google’s free cash flow turned negative in a single quarter. All the money earned was poured back in, and it still wasn’t enough. So they issued bonds. $25 billion. Not enough? Issue more. Since early 2025, Alphabet has raised over $114 billion in bond financing. In the first half of this year alone, debt financing exceeded $50 billion, and nearly $85 billion in stock was issued. One company, two years, borrowed $200 billion. All poured into AI. But that’s not the most painful part. On the same day as the bond issuance, another event happened. Google’s Chief Scientist Jeff Dean, after 27 years, left. He took three top researchers with him. They founded an AI company called Discovery Loop. On the same day, DeepMind’s CEO and Nobel laureate Demis Hassabis stepped down from daily management to become Chairman and Chief Scientist. Nominally a promotion, but in reality — the researchers were pushed out of the decision-making table. Google split its AI department into three paths: product chasers stay, future thinkers sidelined, and the free spirits leave. Once the news broke, Alphabet’s stock price dropped more than 5% over two days. Putting these two events together, the picture becomes clear — On one side, $25 billion in bonds issued; on the other, 4 AI geniuses leaving. On one side, frantically pouring money into building data centers, buying computing power, stacking models; on the other, core talent is draining away. Money flows in, people flow out. What does this mean? The AI race has officially entered its second phase. What was the first phase? Technological breakthroughs. Whoever makes the best model first wins. Google has Transformer, DeepMind, Jeff Dean — the world’s top technical reserves. But the second phase competes on two things: capital and talent. Money, Alphabet has — $205 billion capital expenditure, $114 billion bond financing, and cash on hand. But talent? In the past six months, Google’s AI talent has been leaving in batches. Gemini co-lead went to OpenAI, core researchers went to Anthropic. Now even Jeff Dean left. Who will use the computing centers you spent $200 billion building? Who will iterate the large models you spent $205 billion training? Will AI competition in the future rely more on technological breakthroughs or financial investment? My answer is — neither. The future depends on a virtuous cycle of "having money to keep people, and people to spend money." Money can buy computing power, but computing power needs people to manage it. People can produce technology, but technology needs money to build. Missing one is a vicious cycle. Alphabet’s current situation is — plenty of money, but people are leaving. Jeff Dean is not the first, nor will he be the last. When a company’s AI strategy becomes "just throw money at it," what will the real AI experts think? "My value is not on your balance sheet." $25 billion can buy servers, but not 27 years of technical faith. $205 billion can build computing power, but not the next Transformer. $GOOGL $MSFT $AMZN #谷歌母公司发债250亿美元，AI投入压力升温