Investing in AI is high-stakes risk, but so far investors and industry watchers in the space have little sway over whether or when AI will break.
OpenAI recently closed its latest round of $122 billion funding, with continued support from the usual suspects Amazon, Nvidia, Microsoft, and SoftBank.
Nvidia isn’t just backing OpenAI. It also sells chips that AI companies need to advance their own technology. Such arrangements have led to criticism that the AI sector is like a money pit, supported by investor capital while companies are still trying to figure out profitability. This may be typical for startups, but expectations are placed so high on AI that failure can have far-reaching consequences.
The question for CIOs is whether a self-reinforcing funding cycle in which investors back companies and those companies become customers can sustain the vendor ecosystems and pricing models that companies rely on to drive their profits. AI initiatives from pilot to production.
Even further away, Public opposition to the construction of large-scale data centers It is intended to support AI, which raises questions about ongoing costs and technology advancements. In municipalities in Tennessee, Missouri, Indiana, New Jersey and other states, residents are challenging plans to build or expand data centers in their areas. Maine recently advanced legislation that would put a moratorium on large-scale data center construction across the state. The bill has not yet been signed.
The funding cycle driving AI reminds Craig Everett, assistant professor of finance at Pepperdine Graziadio School of Business, of the fiber-optic ramp-up of the 1990s. At the time, he said, the telecommunications industry was “hustling” to connect the world with fiber-optic cables and building far too much equipment.
“They weren’t making equity investments with each other, they were doing so-called capacity swaps, which was really disgraceful,” Everett said. He is also a director of the Pepperdine Private Capital Markets Project.
Everett said some carriers are buying each other’s capacity in-kind. For the companies involved, the impact on actual costs was virtually zero, but on the books the profits of both companies would increase because the physical transactions would be recorded as revenue. “It was a little shady,” he said.
keep money honest
While current deals and financing with AI may also raise eyebrows, Everett said the way AI is being handled seems to be beyond the norm. “It’s definitely a merry-go-round of funding…you invest in a company and that company buys your product. It tends to have an upward spiral effect until, of course, the music stops,” he said.
On the surface, he said, these seem like legitimate investments. “The fact that they are also customers is a nice side effect.”
AI is often framed as a tool that CIOs can deploy to achieve efficiency and internal creativity, but not every idea that an AI company generates has legs. Even a well-funded bet can go wrong: OpenAI Quit the generated video app “Sora” Later this month, APIs are expected to be published in September, highlighting how quickly expensive AI initiatives can be reevaluated. Sora’s death also ended a $1 billion licensing agreement with Disney. Sora’s operating costs and copyright issues appear to outweigh its short-term benefits.
Pursuing military contracts can also be a source of income for AI players, but such relationships are risky. Anthropic’s insistence on putting guardrails in place for military use of its AI clashed with the Department of Defense, which barred it from contracting. OpenAI has also sought to improve defense contracts to prevent its technology from being used for surveillance or other purposes.
Is there an income stream?
Will AI then primarily survive on its funding rather than its actual revenue? Daniel Docter, managing director at Dell Technologies Capital, said similar questions surfaced in earlier technology cycles, such as telecommunications in the early 2000s. He cited Enron and WorldCom’s frauds that led to both companies going bankrupt. “Isn’t the money that’s flowing here just to turn things around and buy equipment and fiber optic and bring it back here? Hey, something’s going on. Obviously, something was going on,” he said.
What Docter sees as different this time around is the underlying demand for AI, which he says shows no signs of slowing down yet. “The important words are still“I haven’t seen anything yet,” he said.
Doctor said a large number of companies in the AI space will need to do the heavy lifting to build infrastructure (chips, computers, networking, data centers) that consumes new capacity as soon as it comes online. “It’s consumed quickly. It’s like, ‘It’s ready. Raise your hand if you want it,'” he said.
Rethinking what it takes to fund innovation
According to Steven Waterhouse, founder and general partner at Nazaré Ventures, the AI funding cycle can be misunderstood or misdiagnosed. He’s been building in technology and the internet since before there were web browsers. Recalling that Yahoo and other dot-com companies went public, Waterhouse said there were questions about the money that went into those companies and the revenue they generated. “During periods of rapid expansion with new technology, you’re going to see strange funding,” he said.
While deals like Nvidia’s investment in OpenAI and Microsoft’s investment in Anthropic may garner attention, there are other AI players and investors in the broader ecosystem that continue to grow, supported by what he calls real returns. “Currently, our portfolio includes 16 companies around the world, including Europe and the United States. This is not just a Silicon Valley phenomenon I’m talking about,” Waterhouse said.
In particular, he said he is seeing an acceleration from proof-of-concept to production revenue as companies plan long-term contracts for either compute, applications, and agent workflows.
Despite that potential, the cost of building AI capabilities remains a clear issue, said Greg Zorella, principal analyst at Forrester. “There are limited supplies of things like data centers that support scaling AI use cases across the enterprise,” he said.
Additionally, AI costs may rise in the near term as many companies move from proof of concept to scale-up from mid-year onwards. Limited supply naturally means companies need to dig deep into their pockets. “If you don’t have the capacity to handle the exponential growth in AI adoption, someone is going to end up paying more for their AI adoption,” Zorella says.
The other shoe you might drop
He cautioned that companies may not be taking into account the very complex economics of how much AI actually costs, especially since market dynamics can drive prices up.
It remains to be seen how long investors will be willing to spend money in the AI sector, as costs remain significant for all involved. Zorella said that even after end-user companies understand their cost model, they still need to understand what those costs will be in two to three years.
“How much does it cost to hire an agent, given all the cloud fees, LLM fees, and all these other types of fees you might not have thought of?” Zorella said.
