AI revenues skyrocket — and enterprise CIOs pay the bill

Leading AI vendors and infrastructure providers have invested more than $1 trillion in the emerging technology in recent years, and the check is coming due, with some of these companies now aiming to cash in on past spending.

Huge revenue projections from AI companies like OpenAI, Anthropic, and Nvidia have raised eyebrows from potential customers, with critics saying end users will be stuck with the bill after an orgy of spending over the past four years.

The cost of new technology to customers tends to decrease over time, but some observers suggest the recent AI gold rush may delay price cuts for several years. 

CIOs should ask who is supporting the record revenues that some AI vendors are projecting, says Calvin Cooper, COO at AI cost containment vendor NeuroMetric AI. Those revenues are driven by enterprise contracts, he suggests.

“Pricing power is shifting toward vendors, and the window to negotiate from strength is closing,” he adds. “You are not a passive beneficiary of AI progress. You are the revenue base underwriting it.”

While per-token computing costs have dropped, AI vendors are finding new ways to increase revenue, he says, including new agents, reasoning compute, and advertising embedded in LLMs, as OpenAI has recently announced.

CIOs should look to the recent past, with the cloud era as a cautionary tale, Cooper says, when IT leaders have complained about the costs of single-vendor cloud deployments for years.  “As AI goes deeper into operations and moves faster than cloud ever did, the time to build for optionality is now, before the lock-in happens and switching costs become material,” he adds.

Huge revenue growth

Cooper and other critics have raised concerns as OpenAI reportedly projects to triple its revenue in the next year. Competitor Anthropic recently announced its annual revenue has hit $14 billion, growing by 1,000% in each of the last three years. And GPU maker Nvidia’s 2025 revenue was $130.5 billion, a 114% increase from 2024.

At the same time, the AI spending glut may not slow down. Late last year, Nvidia CEO Jensen Huang predicted up to $4 trillion in AI infrastructure spending by the end of the decade, which would be wonderful news for his company. Just recently, Huang suggested that AI projects shouldn’t be judged by near-term ROI, or the lack thereof, but instead, companies should keep experimenting.

The suggestion that AI customers ignore ROI earned Huang some jeers.

Revenue red flags

The profit projections for large AI vendors shouldn’t automatically exclude them from consideration, but CIOs need to be aware of the implications, says Sam Gupta, principal consultant at digital transformation advisory firm ElevatIQ.

“It’s a red flag, as it generally indicates aggressive sales and profit targets,” he says. “These aggressive targets lead sales teams to adopt a short-term mindset to save their jobs, even if it means buyers losing their jobs due to under-delivery of overpromised claims.”

AI prices are likely to drop in the long term, but not before AI vendors pull out huge profits to recoup their infrastructure investments and energy costs, he adds.

“It’s the drug dealer analogy, where the goal is to get people and businesses addicted to the drug called AI,” Gupta says. “While AI is unquestionably powerful, once the real costs kick in, companies need to evaluate the value of AI against real costs, which would obviously not be as rosy as today.”

Some relief eventually

One piece of good news for AI buyers is that there’s still plenty of competition for the large vendors, even if consolidation may happen over the long run, Gupta says. “Right now, AI is really in that early-adopter phase, and because of that, millions of companies and solutions are popping up every day,” he adds. “The hype curve might settle down a bit as buyers mature.”

Additional forms of competition will emerge in response to customer concerns over costs, data sovereignty and other issues, says Behnam Bastani, CEO and founder at AI edge computing vendor OpenInfer.

Cheaper ways of running AI are coming, and new LLM competitors will take market share away from the big guys, Bastani says.

“Change is coming, which means profit margin probably goes down, and that change also means new creativity,” he says. “We’re going to see some companies have devaluations, and we’ll see certain companies pop up that bring new type of opportunities to be favored.”

Learning from the past

While the AI revenue projections may put off some customers, past IT breakthroughs have followed a similar trend, adds Eric Helmer, EVP and global CTO at software support provider Rimini Street.

“Mainframes, databases, and cloud all followed the same pattern: heavy capital investment, rapid enterprise adoption and ultimately a small number of big winners,” he says. “Large AI profit projections should not scare enterprise buyers or be a reason to avoid these vendors.”

Nevertheless, IT leaders should pay attention to the way they deploy expensive AI tools, he adds. “What gives us pause is how we use them, making sure abstraction and governance are built in from day one to avoid crippling dependency,” Helmer adds.

Still, enterprises will end up footing the bill for all this AI investment over the short term, he says. In most technology cycles, prices come down over the longer term, he adds.

“AI is no exception, but it is not a straight line,” Helmer says. “The difference with AI is the sheer scale of upfront investment. Someone has to pay for that, and in the early stages, it is largely enterprises, through usage-based pricing and long-term commitments.”

IT leaders should focus more on the value of the AI tools they deploy than the price, he adds. “If AI meaningfully improves productivity, decision-making and outcomes, the economics work,” he says. “If it does not, no amount of future price compression will justify the spending.”