Tech
toy soldier standing with dollar
Getty Images

Welcome to the OpenAI, Anthropic, and Google price wars

It’s the clearest signal yet that AI models are becoming commoditized.

In a matter of days, the narrative surrounding the artificial intelligence boom has violently shifted from performance breakthroughs to a brutal, margin-crushing price war.

This week, Google slashed its entry-level consumer AI Plus subscription from $7.99 to just $4.99 a month. Now, The Wall Street Journal is reporting that soon-to-be-public OpenAI is exploring drastically cutting its token pricing to defend its enterprise turf against soon-to-be-public Anthropic, which is poised to do the same.

It is the clearest signal yet that the foundational models powering the AI revolution are becoming a commodity.

For the past few years, tech giants and startups alike could charge a premium for the sheer novelty of generative AI. But as mainstream models achieve relative parity on common enterprise tasks, business customers are questioning the hype and the costs. Chief information officers are realizing that these tools are at least somewhat interchangeable, and they are starting to push back against the exorbitant costs of “tokenmaxxing” — the unchecked burning of AI compute as a proxy for productivity. When companies like Uber report maxing out their entire 2026 AI budgets by early summer, brand loyalty disappears. Enterprises simply swap to whichever model is cheapest that day, and the vendor lock-in evaporates. When the magic fades, you compete on price.

Corporate AI adoption chart
Sherwood News

While it’s normal for companies to lower prices to undercut their competition and gain market share, the glaring issue here is that the product these companies are selling is incredibly expensive to produce and is already being sold at a steep loss.

Unlike traditional software-as-a-service products, which boast low marginal costs, generative AI is astonishingly capital-intensive to operate. Every query and line of code demands a massive, continuous burn of specialized silicon and power. OpenAI and Anthropic are already losing billions of dollars annually just to keep the lights on and the servers humming.

Of course, Google has the deepest pockets and the most resilient ad revenue engine of the bunch, allowing it to treat AI as a loss leader. But for pure-play AI startups relying on token volume to survive, a price war could be a bloodbath.

As OpenAI and Anthropic head toward the public markets, Wall Street is about to find out if they’re backing a high-margin tech revolution or just a heavily subsidized utility.

More Tech

See all Tech
tech

Report: OpenAI and Nvidia in talks to team up for 10-gigawatt data center in Ohio

Fresh off scaling back ambitious plans for its Stargate data centers, OpenAI may be moving forward with a new plan: a 10-gigawatt data center in Ohio powered and backed by Nvidia.

According to a report by The Information, the new data center, built on federal land, would dwarf the largest data centers being built today in terms of computing power.

The facility would cost about $500 billion to build, and OpenAI would would own the equipment and be on the hook for 20 years of lease payments, which Nvidia would provide a backstop for, per the report.

If this sounds familiar, Nvidia and OpenAI did announce a similar deal back in September. Nvidia said it would invest as much as $100 billion in what CEO Jensen Huang called “the biggest AI infrastructure project in history,” which never came to fruition (though Nvidia did invest $30 billion in OpenAI). Per the report, this potential deal is a new plan.

OpenAI’s Stargate partner SoftBank is part of the plan as well. SoftBank’s SB Energy is providing financing for the project, and broke ground on the facility in March. The land on which the data center would be built is owned by the Department of Energy.

The facility would cost about $500 billion to build, and OpenAI would would own the equipment and be on the hook for 20 years of lease payments, which Nvidia would provide a backstop for, per the report.

If this sounds familiar, Nvidia and OpenAI did announce a similar deal back in September. Nvidia said it would invest as much as $100 billion in what CEO Jensen Huang called “the biggest AI infrastructure project in history,” which never came to fruition (though Nvidia did invest $30 billion in OpenAI). Per the report, this potential deal is a new plan.

OpenAI’s Stargate partner SoftBank is part of the plan as well. SoftBank’s SB Energy is providing financing for the project, and broke ground on the facility in March. The land on which the data center would be built is owned by the Department of Energy.

A robotics system is demonstrated during LogiMAT 2026, highlighting advances in warehouse automation. (Photo by Leonardo Gerzon/NurPhoto via Getty Images)

The robots are coming... to help small businesses, actually

Labor shortages, not bots, are the bane of so-called blue-collar businesses.

Patrick Sisson6/10/26
tech
Rani Molla

Amazon just secured a massive $17.5 billion line of credit

Amazon has landed a $17.5 billion line of credit arranged by Citibank, according to a new SEC filing.

While the filing says the money is for general corporate purposes, the company is clearly on a global borrowing spree to fund its massive AI infrastructure investments, with $200 billion in planned capex this year. For perspective, that budget is larger than the entire GDP of most countries. This giant credit line comes shortly after Amazon shattered the record for issuance in Canada’s “maple bond” market.

The spending is so aggressive that credit rating agency S&P recently warned Amazon’s leverage will increase substantially and it will likely report negative free operating cash flow over the next two years to support the data center build-out. Yet, Amazon is rushing to borrow anyway, hoping to service a massive $364 billion cloud backlog.

69
Rani Molla

I didn’t make this up: Tesla currently has authorization for 69 unsupervised Robotaxis in Texas, according to the state’s database. That’s up from 42 — perhaps a reference to 420 — last month. While that represents growth, it’s far from the scale that CEO Elon Musk had promised.

And having permission to be on the road doesn’t mean the vehicles are actually in service.

The number of unsupervised Robotaxis has actually declined recently, despite the company’s highly publicized expansion, according to data from Robotaxi Tracker. The site has tracked 32 active unsupervised Tesla Robotaxis in the last month and just 23 in the last week.

Tesla and Musk, who once threatened to take the company private at $420, have long been fans of sophomoric numerology. You can’t actually tip in the Robotaxi app, but as a joke the company suggests tips of $0.69 or $4.20 — and if you tap them, it brings up a “just kidding” graphic.

Latest Stories

Sherwood Media, LLC and Chartr Limited produce fresh and unique perspectives on topical financial news and are fully owned subsidiaries of Robinhood Markets, Inc., and any views expressed here do not necessarily reflect the views of any other Robinhood affiliate, including Robinhood Markets, Inc., Robinhood Financial LLC, Robinhood Securities, LLC, Robinhood Crypto, LLC, Robinhood Money, LLC, Robinhood U.K. Ltd, Robinhood Derivatives, LLC, Robinhood Gold, LLC, Robinhood Asset Management, LLC, Robinhood Credit, Inc., Robinhood Ventures DE, LLC and, where applicable, its managed investment vehicles.