在旧金山举行的 HumanX 会议上，Nvidia 执行长 Jensen Huang 敦促商业领袖以长远眼光看待 AI，而不是追逐短期 ROI。他表示，应以耐心、开放以及愿意尝试的态度来面对 AI，并把它的扩散比作工业革命，主张公司不应让短期市场压力遮蔽更大的变革。他的言论把 AI 描述为一种技术：前期支出可能不会立刻带来回报，但仍可能发掘出意想不到的价值。
Huang 的讯息传出之际，Nvidia 的业务持续飙升：其营收有望在本财年结束时达到 2023 年水准的 10 倍以上，而公司正把每 1 美元销售额转化为 75 美分的毛利。文章也提到 Anthropic 公布的营收运行率超过 30 billion 美元，以及超过 1,000 家商业客户每家每年支出超过 1 million 美元；在 Anthropic 承诺购买大量由 Google 设计、由 Broadcom 制造的晶片后，这一数字也推动了 Broadcom 的股价。这些数字显示，尽管投资人对回报提出更严格的问题，AI 热潮依然强劲。
核心的保留意见在于，仅靠营收成长本身并不能证明经济模式可持续，尤其对那些尚未公布完整损益表或资产负债表的私营 AI 公司而言更是如此。Nvidia 能透过季度报告为其 4.5 trillion 美元市值辩护，但 AI 营收背后的成本仍不明朗，文章指出，审视是有益的，因为许多较小公司将无法存活。Jon Peddie Research 的一项预测显示，在目前 135 家制造或计划制造 AI 处理器、并获得 28.8 billion 美元融资支持的公司中，预计到 2030 年只有 25 家能够存续，突显出这个产业可能变得多么集中且风险高。
At the HumanX conference in San Francisco, Jensen Huang, Nvidia's CEO, urged business leaders to take the long view on AI rather than chase near-term ROI. He said AI should be approached with patience, openness, and willingness to experiment, comparing its spread to the Industrial Revolution and arguing that companies should not let short-term market pressures obscure the larger transformation. His comments framed AI as a technology where early spending may bring no immediate return but could still uncover unexpected value.
Huang's message came as Nvidia's business continued to surge: its revenue is on track to end the current fiscal year more than 10-fold above 2023 levels, and the company is turning each dollar of sales into 75 cents of gross profit. The article also points to Anthropic's disclosed revenue run rate of more than $30 billion and more than 1,000 business customers each spending over $1 million a year, a figure that helped boost Broadcom after Anthropic's commitment to buy a large amount of chips designed by Google and made by Broadcom. These numbers show why AI enthusiasm remains strong even as investors ask harder questions about payback.
The central caveat is that top-line growth alone does not prove sustainable economics, especially for private AI firms that do not yet publish full profit-and-loss statements or balance sheets. Nvidia can defend its $4.5 trillion market value with quarterly reporting, but the costs behind AI revenue remain unclear, and the article says that scrutiny is healthy because many smaller firms will not survive. One forecast from Jon Peddie Research suggests that only 25 of 135 companies currently manufacturing or planning to manufacture AI processors, backed by $28.8 billion in funding, are likely to remain by 2030, underscoring how concentrated and risky the sector may become.