AI bubble and impact on the digital infrastructure space

AI bubble in the context of digital infrastructure dynamics

We are witnessing an unprecedented capital deployment cycle, where hundreds of billions of dollars are being invested in AI compute, energy, and networking capacity. The core tension lies between massive infrastructure buildout, driven by enterprise commitments to spend 10–40% of budgets on AI, and the current revenue reality of roughly USD 20 billion across AI-native companies worldwide.

However, this apparent mismatch may reflect measurement limitations rather than bubble dynamics. The critical infrastructure challenge ahead involves energy constraints potentially outpacing data centre construction, with blackout risks posing systemic threats. The key question is not whether demand exists, but whether the infrastructure stack can scale sustainably to meet the requirements of embodied AI and AI-driven solutions that are not yet priced into current projections.

This document aims to connect the main drivers of the alleged AI bubble with the dynamics of the digital infrastructure industry and present our view on whether, from this perspective, it may represent a bubble in the making.

The debate over whether Artificial Intelligence (AI) constitutes a bubble is an ongoing discussion that emerged prominently in 2023 following the launch of ChatGPT. The explosion of Artificial Intelligence (AI), particularly generative AI (gen-AI), has triggered a wave of market capitalisation and enthusiasm that draws direct parallels with the tech bubble of the late 1990s. The central question facing financial analysis is whether this rapid escalation represents a fundamental and sustainable technological transformation, or merely a wave of speculative euphoria driven by hype.

Although AI’s valuation has evolved to unprecedented highs, it remains below the price multiples observed at the onset of previous bubbles in terms of the price multiple to the bubble start. In fact, as of the cut-off date for this document, no single KPI or conclusive indicator confirms that AI represents a financial bubble. Meanwhile, a surge in Google searches for “AI Bubble” since August, along with the thousands of articles and comments on the topic, signals an unprecedented level of widespread public concern.

Beyond this debate, it is also important to recognise that technology has become an essential component of modern economies. Its exclusion from regional growth strategies would mean ignoring not only its positive contributions but also the expectations of a prosperous future associated with it.

A closer examination of the underlying components of the entire value chain, particularly digital infrastructure, reveals key insights that either support or contradict concerns about AI bubbles. This report will examine main points around the AI bubble debate (economic strain, investment dynamics, valuations, ROI, hardware requirements, circular investment flows, and monetisation models) and link each of them with the economics drivers and characteristics of the digital infrastructure space. By weighing both supporting and opposing arguments, we expect to provide a clear view on whether AI is a bubble from a digital infrastructure angle.

In light of these dynamics, Fide Partners supports investors and operators in identifying real value across the digital infrastructure landscape and navigating the strategic implications of AI with clarity and confidence.

Note 1: The cut-off date is November 2025. Given the pace of developments and AI-related announcements, some aspects discussed may be outdated by the time this document is read.

The investment paradox of AI: Circular flows, inflated valuations, and real fundamentals

AI circularity reveals a risk of interdependence: if one fails, could all fail?

Why it might be a bubble

The recent boom of circular agreements between OpenAI and the largest chip providers (NVIDIA and AMD), as well as with the infrastructure principal players (Oracle, Microsoft, CoreWeave), has turned on the alarms on possible speculative excess in the AI market. As an example, NVIDIA has announced that it will invest USD 100 billion in OpenAI, while later will purchase more chips from NVIDIA. Simultaneously, AMD publicly disclosed a strategic multi-year partnership with OpenAI that includes issuing warrants allowing OpenAI to acquire up to 10% of AMD’s equity, alongside multi-billion-dollar purchases of AMD’s AI accelerators.

This pattern is also evident in the partnership between OpenAI, Thrive Capital and Thrive Holdings, where Thrive Capital, which invests in OpenAI, has launched Thrive Holdings and granted OpenAI equity for supplying models and technical expertise to its portfolio companies, forming a circular flow in which OpenAI powers the value creation that underpins Thrive’s returns. These examples suggest that chip providers are implicitly “subsidising” their own buyers and have led some analysts to identify similarities with the dot-com bubble.

Similar to the above, current capital investments are clustering around a limited number of AI model developers and infrastructure providers that rely heavily on hyperscale cloud operators. This has created a self-reinforcing ecosystem where hyperscalers fund start-ups, start-ups depend on their infrastructure, and valuations rise simultaneously on both sides. Such interdependence magnifies vulnerability. Should hyperscaler investment slow or compute access tighten, valuations across the ecosystem could adjust sharply.

Neocloud providers, such as CoreWeave, Crusoe, and Nscale, play a pivotal intermediary role within the AI investment loop, positioned between GPU manufacturers and hyperscalers. The next diagram illustrates how these operators both acquire GPUs from NVIDIA and receive indirect capital or off-take support from it, forming a reinforcing cycle of funding and hardware dependency. This structure accelerates GPU deployment but also concentrates financial exposure, as many neoclouds’ margins and liquidity remain tied to NVIDIA’s pricing and product cadence.

Moreover, the financial structure is worrisome: Microsoft reported a USD 3.1 billion hit to its net income in Q3 2025 from its equity-method investment in OpenAI, financing over USD 27 billion in liabilities linked to OpenAI’s funding rounds.

Exhibit 2.1: AI ecosystem financial flow

This implies an estimated loss for OpenAI of around USD 11.5 billion in that quarter, though the figure remains a derived estimate rather than an officially disclosed amount, according to disclosures within Microsoft’s earnings filings. Additionally, OpenAI has projected a cumulative cash burn of up to USD 115 billion through 2029. These figures underline that the funding model relies heavily on new capital commitments and circular investment (rather than on robust FCF generated by the business itself), which is a hallmark of earlier speculative bubbles.

Beyond the record-breaking capital expenditure, the market’s current valuations further illustrate the speculative tension surrounding AI. OpenAI, for instance, is now valued at nearly USD 500 billion, up from USD 157 billion a year ago, despite projecting multi-billion-dollar annual losses through 2028 and not expecting to become profitable until 2030. This reinforces the disconnect between valuation and near-term fundamentals, particularly as its CEO, Sam Altman, has stated that profitability is not a priority at this stage while investment requirements continue to scale toward the USD 1 trillion mark.

Meanwhile, the IMF and the Bank of England have both warned that risk assets, particularly those linked to AI, are “well above fundamentals” and show “stretched valuations.” According to Morgan Stanley, just five AI-driven companies accounted for roughly 75% of the S&P 500’s gains, raising concerns about a potential “Cisco moment”; a concentrated collapse if demand fails to meet projections. These dynamics echo the speculative excesses of past bubbles.