While enthusiasm for the space still abounds, as evidenced by the continued growth in AI funding through early 2024 and the run-up in GenAI chip-maker NVIDIA’s stock price, many are starting to wonder: Is this investment environment sustainable or will we see sequential declines in venture financing over the next few years? And if the VC financing environment slows down, will GenAI have a second act?

Expectations are beginning to temper

While big funding rounds for GenAI startups are still happening, we anticipate that as a whole, funding will decline in the next couple of years. Despite last year’s great expectations, scaling GenAI in real-world business applications has proved challenging, and venture firms are beginning to take a more discerning approach, wanting to see that startups have clear, achievable paths to profitability and scalability.

As the funding environment tightens, signs are emerging that some GenAI startups are running out of capital, including prominent “unicorns.” Amid mounting money problems, Stability AI, the maker of the popular text-to-image model Stable Diffusion, laid off 10% of its staff in April and was recently recapitalized by an investor group. And Inflection AI recently made a deal with Microsoft for $650 million, a small fraction of the $4 billion valuation at which the chatbot startup raised $1.3 billion last year.

This could be a sign of things to come. While we expect valuations for best-in-class GenAI companies will continue to hold, last year’s stratospheric valuations of 50-100X revenue will most likely begin to fall back to Earth, and we may see even more small and mid-sized GenAI companies seek the shelter of an incumbent acquisition.

Predicting GenAI’s next move

Breakthroughs in foundation models (FMs) such as large language models (LLMs) ushered GenAI onto the world stage last year. However, FMs are increasingly capital-intensive to develop and maintain, requiring thousands of specialized chips to handle their computing power and massive amounts of electricity to run their data centers.

As infrastructure costs skyrocket, some startups are partnering with—or even ceding the field to—Big Tech incumbents that can continue to pour billions of dollars into FMs without needing to generate revenue immediately due to the profitability of their adjacent businesses, such as cloud computing, search, or digital advertising.

With these challenges in mind, many AI investors are starting to ask: What’s beyond foundation models? We see a few likely candidates.

First, we predict increasing uptake among enterprises of AI-first, third-party software-as-a-service (SaaS) providers and infrastructure companies, especially as these startups mature and enterprises see initial return on investment (ROI). In fact, we’re already observing increased enterprise adoption of SaaS AI solutions across verticals such as knowledge management, marketing, sales, legal, and customer service.

On the infrastructure side, we foresee the rise of pick-and-shovel tools that help enterprises go from pilot to production in a scalable way, including prompt management, model security and routing, output evaluations, interference and cost optimization, and extract, load, and transform (ETL) processes on unstructured data.

However, we think the hottest new area of venture investment will be autonomous AI agents, software agents that can simulate human behavior and plan, make decisions, and execute tasks in complex environments without human intervention or supervision. Autonomous AI agents are just emerging from research and development, and as of now, their funding pales in comparison to the funding for GenAI’s core infrastructure. While we don’t predict that they will make up that ground anytime soon, AI agents could start to see significant funding growth in the next few years.

GenAI’s globality means more funding opportunities

As venture investors representing the world’s most global bank, we at Citi Ventures have always taken a broad perspective on AI. While most tech markets have started in the U.S. and rippled outward, GenAI has been an international phenomenon from the get-go.

Europe has long been at the vanguard of AI innovation, giving rise to success stories such as Hugging Face, Mistral AI, and Aleph Alpha (which last year raised one of the continent’s largest AI funding rounds ever). Asia-Pacific is also emerging as a pivotal player in the AI ecosystem, with China home to a fourth of the world’s top AI experts and Southeast Asia coming up as a leader in the “vertical AI” market, which focuses on niche B2B applications.

Given GenAI’s many epicenters, we see increasing opportunities to invest in regional companies applying GenAI to local use cases. For example, we’re interested in the development of LLMs in non-anglophone countries that have strong digital infrastructure but can’t use GenAI apps geared toward English speakers, such as the recent work in India to develop multilingual language models that can understand and respond in local languages.

Optimism is still high for the future of GenAI

Despite our prediction that GenAI funding will decelerate this year after the initial burst of enthusiasm, we still believe the industry’s recent breakthroughs are among the most profound technological advancements of our time.

Foundation models, especially LLMs, have the potential to change every aspect of our lives—from the way we work, create, and experience art to how we make scientific discoveries and more. But they have significant challenges to overcome before they can be adopted broadly within enterprises, and solving these challenges will require extensive amounts of capital and time for research and development.

Meanwhile, we remain optimistic about GenAI as innovation moves to the application layer and as regional startups apply it to local use cases. The GenAI revolution may slow down as it proves its worth, but it’s here to stay.

More must-read commentary published by Fortune:

• Inflation, housing, immigration, taxes: The Harris-Walz economic policy scorecard
• The ‘sustainability recession’ will end soon—and not by choice
• AI development is being hijacked by Big Tech and rich nations, UN report warns
• The most underrated leadership skill, according to Jake Sullivan

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.