- The AI ecosystem is ripe for consolidation
- Consolidation will likely hit smaller AI startups first
- Meanwhile, the frenzy to build neoclouds could “get ugly,” according to one consultant
Bill Major, CEO with the fiber infrastructure company FiberLight, has predicted for the last couple of years that there was going to be major consolidation in the fiber industry, and his predictions have been correct. Now, Major is predicting consolidation in the AI space.
FiberLight has insight into the AI ecosystem because it lays fiber for the major hyperscalers including Microsoft, Meta, Google and Amazon, and a lot of that fiber is deployed to connect AI data centers.
However, while AI is booming — and bringing fiber along with it — companies in the AI space are struggling to garner profits, he told Fierce Network. “As you look at enterprises and consumers, the adoption is happening, but there's not an ROI for that investment today,” said Major.
At the top of the heap are the big AI platforms such as Open AI’s Chat GPT, Google’s Gemini, Anthropic’s Claude and Microsoft’s Co-pilot to name a few. And there are also big open-source AI platforms such as Meta’s Llama.
But, "not everybody's going to have an AI platform that's going to be successful," said Major.
Sudeep Suman, partner and managing director at the consulting firm AlixPartners, agreed. “So the question everybody is asking is: how does this stack up against the ROI curve?” said Suman. “Is it going to be end-consumer driven, enterprise-driven, ad-driven? That's the struggle everybody's trying to figure it out.”
Neoclouds are in line for consolidation, too
While it's too early to predict the winners and losers at the highest levels of AI, the next level offers a glimpse into areas ripe for consolidation, including neoclouds that provide GPU-as-a-service.
There are quite a number of neoclouds currently raising capital, renting space and buying chips to provide GPU-as-a-service.
“These companies will have a higher cost of capital, higher cost of land and higher cost of energy than any hyperscaler in the market,” said Andrej Danis, partner and managing director with AlixParters. “So today, they will be successful in providing the services because there is an absolute constrained supply,” said Danis. “But, as more capacity comes into the market, these players may get into a challenge, and it may become ugly for them.”
He said the neoclouds own the specialized chips, which he called “perishables” because their lifespan is only two or three years — and better chips that use less energy are quickly becoming available.
“[The specialized chips] are losing value immediately,” said Danis.
He said there will definitely be distress in the neocloud space followed by consolidation, sooner rather than later.
Smaller AI players
There are a lot of AI companies trying to create new applications, and those companies may be ripe for consolidation. In fact, a recent Wall Street Journal article said there are more than 1,000 startups with valuations of more than $1 billion.
Major said that relatively small players are under the gun to show returns on their investments quickly. “Inherently, they're going to get, you know, gobbled up,” he said.
Generally, when a new technology is creating a lot of frenzied investment, about 90% of the startups will fail or be consolidated, but 10% might succeed wildly.
“The AI native vendors who can deliver ROI are the ones who are going to be able to persist,” said Mario Ribera, partner and managing director at AlixPartners. He added that CIOs are experimenting with a bunch of AI software right now, handling about 60 AI tools on average.
“That's a lot,” said Ribera. “Which are the ones that really deliver value? Those are the ones that are going to stay. The rest, they're going to have to look at strategic options we think over the next 12 to 24 months.”
The AlixPartners consultants also noted that a lot of the AI startups use large language models (LLMs) that they don’t own. And many of them are focused on a single use case.
They said that data is key, and it’s ideal if AI startups own their own data and have real domain expertise.
“Data companies who are building applications around that will become winners over somebody just taking an LLM and trying to build a quick, shiny new object, which can be copied by anybody,” said Suman.
In terms of consolidation among the many, smaller AI startups, the AlixPartners consultants expect to see a lot of that soon.
“The next two years, I think customers are picking winners,” said Ribera. “Investors are getting more selective. They've been subsidizing a lot of the spend and the buildout. So [the startups] need to deliver ROI and results, or they're going to end up being consolidated into bigger platforms that can do that.”
Consolidated in FTTH, too
The consolidation in the AI space may be reversed from the consolidation we’ve seen in the fiber ecosystem.
It appears that AI consolidation will happen first among the many small start-ups, followed by the neoclouds. It may be several years before we see consolidation among the super huge platforms.
In the fiber-to-the-home (FTTP) space, we’ve seen big deals such as Verizon buying Frontier Communications and AT&T acquiring the Mass Markets fiber-to-the-home business of Lumen Technologies.
Now, we’re likely to see more deals among mid-size and smaller players such as Ezee Fiber’s purchase of Tachus Fiber Internet.