OpenAI is upping the ante massively - FT中文网
登录×
电子邮件/用户名
密码
记住我
请输入邮箱和密码进行绑定操作:
请输入手机号码,通过短信验证(目前仅支持中国大陆地区的手机号):
请您阅读我们的用户注册协议隐私权保护政策,点击下方按钮即视为您接受。
观点 OpenAI

OpenAI is upping the ante massively

Calculated display of self-confidence is intended to signal it is still leading the charge to superintelligence

If your head is spinning after OpenAI’s recent flurry of megadeals, you’re not alone. Hundreds of billions of dollars are at stake — some of it promises future investment from Nvidia, some capital spending on data centres where construction has yet to start and some payments under a giant cloud computing contract with Oracle.

Look past the details of this spate of announcements, though, and there’s one overriding message: OpenAI is massively upping the ante. Weeks after the biggest tech companies increased their own capital spending plans, the company that started the generative AI gold rush wants everyone to know that it has put the first pieces in place for the biggest bet of all.

It all looks like a calculated display of self-confidence. OpenAI wants customers, employees and partners across the tech industry to know that it is still leading the charge to superintelligence. That is particularly important at a time when it is involved in a messy overhaul of its corporate structure, and as the once clear technology lead it had in large language models erodes.

This has also served as the opening salvo for what promises to be an epic assault on the financial markets, as OpenAI and its partners try to line up the financing for their giant bets. Among the questions yet to be answered: who will be on the hook for the data centres built in the hope that AI demand will continue to boom, and where will the cash flow come from to support the mountains of debt that are sure to be involved?

A back-of-the-envelope tally suggests OpenAI has reached broad agreements that will cover capital spending of about $1tn, some it undertaken by partners in the Stargate project it announced in the Oval Office earlier this year. This includes a deal for Nvidia to invest up to $100bn of equity, supporting data centre spending that could top $500bn, as well as the announcement of five new data centres built by Stargate partners SoftBank and Oracle that might cost another $400bn.

“Back of the envelope” is the operative term here, and it’s far from clear how much of this capacity will ever be built. While the five data centres are projected to be completed over the next three years, the Nvidia deal is open-ended.

There are also plenty of practical challenges to building and powering such a huge expansion in data centre capacity. Oracle, which is taking on the lion’s share of the development under the Stargate project, is a relative minnow in this world and is already carrying significant debt.

Not surprisingly, the many open questions have helped to feed suspicions that the AI market is entering a new and more bubbly phase. How likely is it that OpenAI’s business will support the $300bn in cloud payments it is due to pay Oracle by 2030, or that Nvidia will see the $350-400bn in new chip sales that analysts project from its OpenAI deal?

In deciding how much to price in to current valuations, the market has seemed somewhat schizophrenic. OpenAI’s deal with Oracle sent the software company’s shares into the stratosphere, despite the many open questions. But the initial stock price gains this week for Nvidia — which, regardless of the timing of the data centre builds, at least secured an important commitment from OpenAI to remain a key customer — quickly dissipated.

OpenAI CEO Sam Altman says his company will lay out its detailed financing plan later this year, while Wall Street analysts are hoping that Oracle will reveal its own financial blueprint at a company event next month. Showing OpenAI and Oracle have serious backers in the wings would go a long way to convincing the markets about their plans.

Nvidia’s commitment of up to $100bn in new equity is critical for OpenAI as it looks to outspend a group of giant tech rivals with powerful balance sheets. But the first $10bn slug of this isn’t due until the first data centre is deployed, something that isn’t expected until the second half of 2026, with the rest only pencilled in for future projects.

Ultimately, it will take significant — and sustained — revenue growth from OpenAI to support this whole endeavour. The big jump in capacity from Stargate data centres won’t come online until later this decade. Before then, the company that sparked the AI boom needs to show not only that demand for data centre capacity is continuing to outpace supply, but that customers are willing to pay serious money for its services.

richard.waters@ft.com

版权声明:本文版权归FT中文网所有,未经允许任何单位或个人不得转载,复制或以任何其他方式使用本文全部或部分,侵权必究。

弗兰克•奥哈拉与“美国世纪”的终结

这位诗人兼策展人曾站在美国对外文化攻势的最前沿,力图让“高雅艺术”的输出与军事力量一样强硬——而如今,这种理念似乎已经走到尽头。

NBA最差球队为何心甘情愿输球?

华盛顿奇才队一直在“摆烂”。这或许能帮助他们锁定篮球界的下一位超级新星。

股市的狂欢终究难以为继

投资者情绪正转向炽热的极度亢奋。

QVC债权人正饱受“买家懊悔”折磨

这家曾经家喻户晓的家庭购物电视频道已申请破产保护。

为何共享单车应用Lime的首次公开募股并非“烂摊子”

按讨论中的20亿美元企业估值计算,这家由优步支持的公司的企业价值将相当于去年营业利润的28倍。

本周图表:英国国债,其实没那么糟

英国政府债券并不像从表面上看那样异类。
设置字号×
最小
较小
默认
较大
最大
分享×