As 2026 gets underway, it's clear that AI's momentum hasn't stalled in the least. If anything, it's picking up speed, especially with the recent surge of interest in AI-powered agents, often referred to as agentic AI, that are beginning to reshape how the technology is used.
Bob O'Donnell is the founder and chief analyst of TECHnalysis Research, LLC a technology consulting firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on Twitter @bobodtech
Capital spending on AI projects is surging, not just among the largest hyperscalers but across businesses of nearly every size and industry. For companies supplying the hardware and software that underpin AI systems, that investment wave has largely been good news. Yes, we've seen (and will undoubtedly continue to see) stock market bumps along the way, but there's no denying that the bigger picture trajectory continues to be incredibly positive.
At the same time, businesses and consumers are still working out how they actually want to use these tools. Enough compelling results have emerged to encourage experimentation, and there's real excitement about what might be possible next – even if that excitement is often tempered by a sense of unease about what widespread AI adoption could mean.
"Tokens" may be a convenient unit of measurement, but they remain an awkward stand-in for real customer value.
Despite all of this progress, one critical piece is still missing: business models that make economic sense and allow AI developers to sustain themselves long enough to deliver lasting impact. The tension is easy to describe, if difficult to resolve. The dilemma is straightforward, but not simple: customers want predictable spending and clear business value, while AI suppliers face variable – and in many cases rising – costs to deliver increasingly sophisticated capabilities. "Tokens" may be a convenient unit of measurement, but they remain an awkward stand-in for real customer value.
It's worth noting that the AI ecosystem is far from uniform. Some companies driving major advances are financially healthy. Many others, however, remain stuck in a cycle of heavy investment with little revenue to show for it. In some cases, their long-term viability appears to rest more on optimism than on durable financial fundamentals.
That imbalance has fueled growing fear of an AI bubble or crash because of the uncertainty around how these future plans translate into durable revenue. While some of the fallout has been visible for a while, those concerns now seem to be spreading across the broader tech industry (and even into the global economy) in ways that aren't always rational. Whether the fears are justified or not, their impact is real, and it's unlikely we've seen the last reaction driven by those doubts.
Because of all this, I remain convinced that 2026 is the year we need to start seeing more realistic types of business models coming to the AI side of tech industry. While the modern oracle of AI, Nvidia CEO Jensen Huang, loves to talk about how generating more tokens leads directly to generating more dollars, this argument feels less convincing than it used to. Building complete, easy-to-deploy AI solutions for enterprises while also figuring out how to monetize consumers who are happy to consume massive amounts of AI output for free is pushing early monetization strategies to their limits.
While the modern oracle of AI, Nvidia CEO Jensen Huang, loves to talk about how generating more tokens leads directly to generating more dollars, this argument feels less convincing than it used to.
Some experiments are underway, though their effectiveness remains uncertain. OpenAI's plan to introduce ads into ChatGPT has already faced skepticism, along with some pointed (and humorous, watch below) responses from competitors like Anthropic. Enterprise-focused offerings appear more straightforward, yet even there the shift toward outcomes-based pricing instead of model access raises difficult questions. Paying for results sounds appealing until issues of measurement, accountability, and risk enter the picture.
Plus, while there's been a lot of recognition around the largest, multi-purpose frontier models, some of the most dramatic success has been with smaller companies creating more targeted models that are optimized for specific industries and applications. In those cases, the value proposition is often clearer, and clarity matters when budgets come under scrutiny.
Per-seat pricing models, such as those used for Microsoft's Copilot, initially looked promising. Over time, however, concerns have emerged around uneven usage: heavy adoption by some employees and minimal engagement by others. Metered usage could help address that imbalance, but implementing it effectively is still an open question. Compounding the issue is the lack of meaningful training in many organizations, which significantly limits how these tools are used. There's also a quiet irony at play: companies deploying AI to reduce headcount may ultimately shrink the very pool of seats they're being asked to pay for.
Finally, one other issue that needs to be addressed is trust, but not necessarily in the way you may first think.
Trust in output quality, governance, safety, and security is essential, but so is trust in the longevity of the vendors themselves. Given how potentially impactful and far-reaching the influence of AI tools can be, if a company who wants to deploy a cool new AI technology can't be certain that the supplier of that technology is going to be around for a while, they simply won't do it. In that sense, financial stability becomes a prerequisite for trust.
There's little doubt that 2026 will serve as the backdrop for a fascinating period of discovery, as organizations and individuals start to discover more of the "art of the possible" when it comes to AI and agents. But for these developments to move beyond trendy hype and tech industry navel gazing and reach mainstream businesses and consumers, there's got to be more validity to the business propositions being put forward.
There are already enough technical, social, and even political battles to overcome when it comes to AI adoption. Building a solid base of business viability is going to be essential if the next step in AI industry evolution is going to occur.