The AI sector is expected to divide in 2026.
The final quarter of 2025 witnessed a tumult of tech sell-offs and recoveries, as circular transactions, debt offerings, and inflated valuations raised alarms about an AI bubble.
This kind of instability may indicate early signs of the evolution of AI investment as investors become more discerning about who is spending and who is profiting, as stated by Stephen Yiu, chief investment officer at Blue Whale Growth Fund.
So far, retail investors, particularly those exposed to AI via ETFs, have generally not distinguished between companies that have products but lack business models, those incurring losses to support AI infrastructure, and those benefiting from AI expenditures, Yiu told CNBC.
Currently, “every company seems to be succeeding,” yet AI is just beginning, he noted. “It’s crucial to differentiate” among various types of companies, which is “what the market might begin to do,” Yiu added.
He identifies three distinct groups: private companies or startups, publicly listed AI spenders, and AI infrastructure firms.
The first category, which encompasses OpenAI and Anthropic, attracted $176.5 billion in venture capital during the first three quarters of 2025, according to PitchBook data. In the meantime, major tech players like Amazon, Microsoft and Meta are the ones financing AI infrastructure providers like Nvidia and Broadcom.
Blue Whale Growth Fund assesses a company’s free cash flow yield, which is the cash a business generates after capital expenditures, relative to its stock price, to determine whether valuations are reasonable.
Many companies within the Magnificent 7 are “trading at a considerable premium” since beginning substantial investments in AI, Yiu remarked.
“When I evaluate valuations in AI, I prefer not to invest — even if I believe in the transformative potential of AI — in the AI spenders,” he elaborated, indicating that his firm prefers to be “on the receiving side” as AI allocations are likely to further affect company bottom lines.
The AI “fervor” is “concentrated within specific sectors rather than being widespread throughout the market,” Julien Lafargue, chief market strategist at Barclays Private Bank and Wealth Management, stated in a CNBC interview.
The greater risk resides with firms obtaining funding from the AI surge that have yet to turn a profit—”for instance, certain quantum computing-focused companies,” Lafargue noted.
“In such instances, it appears investor sentiment is driven more by hope than by substantial results,” he added, insisting that “differentiation is crucial.”
Read more
The requirement for differentiation also signals a shift in the business models of Big Tech. Once asset-light enterprises are becoming more asset-intensive as they acquire the technology, resources, and real estate necessary to support their ambitious AI initiatives.
Corporations like Meta and Google have transitioned into hyperscalers that invest significantly in GPUs, data centers, and AI-based products, altering their risk profiles and business approaches.
Dorian Carrell, head of multi-asset income at Schroders, remarked that valuing these entities as software and capital-expenditure-light models may no longer be viable — particularly as firms continue to figure out funding for their AI strategies.
“We’re not asserting that it won’t succeed, we’re not claiming that it won’t manifest in the coming years, but we are suggesting that paying such a high multiple with lofty growth expectations integrated may be questionable,” Carrell told CNBC’s “Squawk Box Europe” on December 1.
This year, tech firms turned to the debt markets to finance AI infrastructure, but investors were wary of overreliance on debt. Although Meta and Amazon have raised funds this way, “they’re still in a net cash position,” Quilter Cheviot’s global head of technology research and investment strategist Ben Barringer informed CNBC’s “Europe Early Edition” on November 20 — a critical distinction from companies that may have tighter balance sheets.
The private debt markets “will be extremely intriguing next year,” Carrell added.
If expanding AI revenues do not exceed those costs, profit margins will tighten and investors will question their return on investment, Yiu stated.
Moreover, the performance differences between companies could widen as hardware and infrastructure depreciate. AI spenders will need to consider this in their investments, Yiu noted. “It’s not reflected in the P&L yet. Starting next year, gradually, it will complicate the figures.”
“Hence, there will be increasing differentiation.”
