Dan Gallagher
Spending money to make money is taking on a whole new meaning in the artificial-intelligence world.
The public is already familiar with the exorbitant capital spending of the world's largest tech companies as they race to build up their AI capabilities. But the two companies that compel much of that investment have sizable bills of their own. OpenAI and Anthropic expect to spend nearly $65 billion combined this year just on the costs to train and operate their AI models, according to financial documents obtained by The Wall Street Journal.
That exceeds what the two companies are generating in revenue. And those two cost items are expected to soar to a combined $127 billion next year and reach nearly $250 billion by 2029.
These are costs the two companies themselves are projecting in financial documents they have circulated to their private investors. OpenAI expects its costs for training and inference, the latter refers to the process by which AI models respond to user queries, to exceed its revenue until 2029. Anthropic expects to cross that line next year, though other costs are expected to keep the Claude chatbot parent in the red on a pretax earnings basis also until 2029.
Reality could play out differently. For instance, both companies could end up growing revenue at an even more explosive rate than currently projected. But costs in the AI market so far have proven to move in only one direction -- up.
And both OpenAI and Anthropic are competing head-to-head with some of the world's largest companies that are themselves developing their own advanced AI models. A key difference: they have lucrative core businesses that can pay for their AI ambitions. Google-parent Alphabet and Facebook-parent Meta Platforms are expected to generate a combined $334 billion in cash from operations this year, according to consensus estimates from FactSet.
Will public investors really be ready to underwrite such expensive endeavors? Both OpenAI and Anthropic are reportedly planning initial public offerings for later this year. Admittedly, signs of AI's runaway costs have been apparent for a while, given the blowout spending of major cloud network operators and the huge growth of key component suppliers like Nvidia.
Also, OpenAI and Anthropic would hardly be the first major tech companies to go public while still deep in the red. Some of those who have turned out quite well for investors. Amazon showed annual operating losses and burned cash for three years after its IPO in May of 1997, according to data from S&P Global Market Intelligence.
But even Amazon wasn't as much of a relative stretch. The e-commerce upstart went public at a market value of around $430 million, which was less than 0.01% of the S&P 500's value at the time.
By contrast, OpenAI and Anthropic are together valued at more than $1.2 trillion based on their most recent funding rounds, according to PitchBook data. That equates to over 2% of the S&P 500's current value. The ability of the largest AI labs to control their costs will matter much more to a wider swath of investors than Amazon's willingness to sell Harry Potter books at a loss.
This is an edition of the WSJ AI & Business newsletter, a weekly digest to help you make sense of AI's impact on business with news, insights and data from our global team of technology journalists. If you're not subscribed, sign up here.
Anthropic Wants to Drive More Business Adoption of its AI Tools
Anthropic is planning to invest $200 million in a new venture with major private-equity firms that aims to sell AI tools to their portfolio companies, continuing the AI company's push for business customers. The new company would serve as a consulting arm for Anthropic that teaches businesses how to incorporate the startup's AI tools in their operations.
The Number
The amount of computing capacity Broadcom has contracted to provide Anthropic beginning in 2027. That capacity will come in the form of Google's TPU chips.
What the Humans Are Saying
AI in Charts
Meta's Ray-Ban smartglasses have been a hit, with 7.2 million units shipped last year according to IDC. Meta sees the glasses as a portal to its AI technology, but its partner EssilorLuxottica is benefiting in a big way as well and may be a cleaner way for investors to play this angle of Mark Zuckerberg's AI vision.
AI in the Wild
Maine looks poised to become the first state to freeze the building of new data centers with legislation that could pass this spring. And community backlash against these properties is spreading across the country. Lawmakers in more than 10 states have proposed temporary bans on data-center construction this year. Dozens of county and city governments have already passed such measures.
Other Highlights From the Week in AI
-- What to know about OpenAI's ideas for a world with 'Superintelligence.' -- The workers opting to retire instead of taking on AI. -- One company's effort to make an AI-ready catalog of everything we buy. -- Intel partners with SpaceX, Tesla to operate new chip plant.
About Us
WSJ AI & Business is a weekly look at AI's transformation of the business world. This newsletter was curated and edited by Dan Gallagher and Asa Fitch. Reach them at dan.gallagher@wsj.com and asa.fitch@wsj.com (if you're reading this in your inbox, you can just hit reply). Got a tip for us? Here's how to submit.
(END) Dow Jones Newswires
April 07, 2026 12:10 ET (16:10 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments