世界上有很多钱。这对 AI 来说够吗？2026 年这场大规模支出狂欢中的一位参与者给出了两个截然不同的答案。
There is a lot of money in the world. Is it enough for AI? One player in the great spend-a-thon of 2026 provides us with two different answers.
阿波罗收购阿布扎比政府支持的资产管理公司穆巴达拉资本股份的交易，押注的是华尔街目前面临的真正短缺不是资本，而是投资渠道。阿波罗的退休金业务每周吸纳 10 亿美元资金，必须找到去处，但其发掘交易的引擎跟不上节奏。通过投资穆巴达拉，它获得了源源不断的交易流的优先认购权。结论是：过剩的资本在追逐过少的机会。
Apollo’s deal to buy a stake in Abu Dhabi’s government-backed asset manager, Mubadala Capital, is a bet that the real shortage facing Wall Street right now isn’t capital, but places to invest it. Apollo’s retirement arm takes in $1 billion a week that it has to find something to do with, and its origination engine can’t keep up. By investing in Mubadala, it gets first dibs on a steady stream of deals. The takeaway is that an oversupply of capital is chasing too few opportunities.
阿波罗向 SpaceX 提供的 400 亿美元芯片租赁则暗示了相反的情况。六天前我们关注 SpaceX 的融资时，风险最高的那一档似乎很可能以比美国国债基准利率高出约 3.5 个百分点的利差找到买家，利差虽高但尚在合理范围。到本周交易完成时，利差已扩大至 3.95 个百分点 —— 对于拥有投资级评级的 SpaceX 来说，这是惊人的高昂成本。这些令人咋舌的条款清楚地表明：AI 资金虽然存在，但其成本已比几周前高出许多。
Apollo’s $40 billion chip lease to SpaceX suggests the opposite. When we checked in on SpaceX’s financing six days ago, the riskiest piece appeared likely to find takers at about 3.5 percentage points over a baseline US Treasury rate, steep but reasonable. By the time it closed this week, the spread had blown out to 3.95 points — outrageously expensive for SpaceX, which carries an investment-grade rating. Those eye-watering terms make clear that while the AI money is there, it now costs a more than it did a few weeks ago.
资本提供方正在迅速认清 AI 交易的风险，即使他们有钱可烧。一位华尔街高管说：“我不说他们四周前很蠢，但……他们变聪明了，而且快得惊人。”
Capital providers are wising up to the risks of the AI trade, even if they have the money to burn. “I won’t say they were dumb until four weeks ago, but…” said one Wall Street executive, “they’ve gotten a lot smarter, fast.”