Inside The AI Hype Cycle: What's Next For Enterprise AI?

Forbes

13 hours ago

You might have read about AI—the next technology revolution that is boosting the stocks of some technology leaders such as Microsoft and Nvidia to new all-time highs. There is no doubt that the world is adopting AI at an unprecedented pace—but for a full embrace from the business community, more work needs to be done.
Google's Gemini AI tells me with data sourced from S&P Intelligence that aggregate AI revenue is projected to reach $85 billion by 2029, marking a substantial increase from the estimated $16 billion achieved in 2024.
OpenAI alone reports that its annual recurring revenue is expected to reach $12.5 billion in 2025, which is a remarkable growth rate for a 10-year-old startup.
As we reach new highs in the hype cycle for AI, it's important to keep things in perspective: The full impact and return on investment in AI is as yet unknown. That's why it's important to track exactly how it's being adopted and where the strongest uptake is happening so far.
The Nuances of Enterprise AI Adoption
Dig beneath the surface, and AI adoption is nuanced. Many consumers are still in the discovery phase of AI. Others are just worried it's going to take away their jobs.
On the enterprise side, there are many challenges. Not all enterprises (businesses) know how to implement AI safely. There are cases in which AI service deployment has had massive success, yet other cases where there have been disappointments and failures.
The enterprises we have studied and spoken with confess they are struggling to adopt AI because of specific challenges—safety, data sovereignty, and cost, to name the most highly cited challenges. They want to be careful about how AI is deployed, because it could produce a mission-critical error, result in a lawsuit, or expose critical data.
Will There Be a Dot-Com Correction for AI?
With AI-related valuations in the stratosphere, it's hard to imagine that this great AI explosion won't experience some sort of setback—just as the Internet did in 2001-2002—as investors digest the reality of living up to enormous expectations built into the market.
The Internet correction wasn't a bad thing. As the technology matured and got cheaper in price, it paved the way for hypergrowth of some of the best companies in the world, such as Amazon and Google (Google actually didn't go public until 2005). The Internet correction lowered costs and allowed more time for the market to develop. The same trajectory is likely for AI.
For investors, the sweet spot of AI has been in the infrastructure market, where the "picks and shovels" strategy has been the best opportunity for investors. As the explosion of generative AI services and startups blossoms, demand for new datacenters and compute power has boosted players such as Nvidia, AMD, and Arista Networks—to name a few—which are providing the hardware to mine for AI gold.
Based on recent capital spending projections, Futuriom expects at least $300 billion to be spent building out AI-related datacenters this calendar year—with a trillion expected over the next three years. But the rate of the buildout could depend on how fast ROI is delivered in the AI segment.
Winners and Losers in Enterprise AI
The hyperscaler segment has also enjoyed great success, with Amazon, Google, and Microsoft all spinning out new tools and services that enable AI. The neoclouds such as CoreWeave and Nebius are not far behind, building out specialized GPU clouds for AI-related applications.
We have noted that AI stalwarts such as Microsoft and Nvidia have probed new highs in their stock-market valuations, but not everybody is faring as well.
Others have not fared as well for investors. For example, Apple's AI strategy has underwhelmed consumers and markets, and its shares remain down 20% on the year. The current rumors are that Apple is willing to pay upwards of $20 billion for a startup to fix its AI strategy—potentially its largest acquisition ever.
In another example, Adobe shares are down 13% on the lack of pull-through for its new AI services, which had at one point last year been built into the share price. In the case of Adobe, AI might represent a competitive threat rather than a boost. For example, customers could seek out new AI-enabled graphics tools and alternatives spawned by the advent of GenAI.
Key Use Cases for Enterprise AI
We have studied more than 100 enterprise deployments, and what we are finding are pockets of success concentrated in specific use cases. For example, we are seeing that AI deployments are been favored in specific verticals, such as financial services/insurance, healthcare, and retail—the top three verticals identified in our sample of deployments.
Many of the key applications are specific use cases, such as documentation automation and customer service that replaces human workers. So far, enterprise AI is a productivity play. According to our analysis of more than 100 deployments, the top benefits cited include the following:
One of the largest revelations of our early data analysis is that custom, proprietary AI platforms garner lots of interest. Again, this demonstrates the nuances needed to deploy AI in the enterprise—with data sovereignty, safety, and security looming large. Most enterprises are more comfortable building their own AI solutions and 'owning' the platform, according to our research.
In the 100+ enterprise use cases we examined, proprietary AI platforms were the most common, with Google Vertex, Microsoft Azure AI, and Amazon Bedrock being the next most common platforms used by enterprises, in that order. The most common AI model across all use cases and verticals was OpenAI's ChatGPT.
When you dive into the top three segments we looked at—financial services/insurance, healthcare, and retail—once again proprietary AI platforms were the most popular.
So what is the takeaway for AI investors? AI adoption is messy at best, and success will come down to execution for specific use cases. One emerging theme we have distilled is that enterprises are going to be particularly sensitive to the security, safety, and privacy of their data—leading to highly customized solutions on private infrastructure.
An AI shakeout may have already started. And the technology companies going after enterprise dollars will need to finely tune their offerings around operational efficiency while delivering high levels of safety and data security—at the same time figuring out the sweet spot for user adoption.