The AI industry’s structure resembles a complex web of financial interdependence.
At its core, companies like OpenAI and Nvidia invest heavily in suppliers, customers, and cloud partners—such as AMD, Intel, Microsoft, Amazon, and Oracle—creating a network where hardware, software, and service providers have a vested interest in each other’s success.
Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs) are specialized processors critical to AI and machine learning. GPUs, originally developed for video game graphics, excel in parallel computations like matrix multiplications, fundamental to deep learning and generative AI. TPUs, developed by Google, are optimized for high-throughput machine-learning workloads, boosting AI task efficiency. This hardware ecosystem underpins the industry’s growth but requires significant financial coordination.
This interdependent financing model supports business continuity for investors. For example, Nvidia, a leading chip manufacturer, reportedly holds a small percentage stake in CoreWeave, a company that rents computing power and purchases Nvidia’s chips, aligning with Nvidia’s strategy to secure demand. Similarly, Amazon has invested several billion dollars in Anthropic, the developer of the Claude AI models, which relies on Amazon Web Services (AWS) for model training and inference. Google’s multi-billion-dollar investment in Anthropic ensures access to Google’s TPU capacity. These arrangements blur the line between genuine market demand and financially incentivized purchases, potentially inflating revenue projections.
The primary financial concern is the massive capital expenditure (CAPEX) required relative to current industry revenue. According to McKinsey & Company, an estimated $5.0 trillion in CAPEX will be needed by 2030 for AI-equipped data centers. To justify this, the AI industry must generate significant revenue, estimated in the tens of billions. OpenAI, a key player, reportedly generated $3.0–4.0 billion in annual revenue in 2024, driven by enterprise subscriptions and API usage. To fund its infrastructure, OpenAI has secured a multi-billion-dollar revolving credit line, potentially reflecting optimism in growth projections, or an underestimation of ecosystem challenges.
The scale of infrastructure required is substantial and critical. OpenAI’s “Stargate” project, a collaboration with Microsoft, is a $100 billion plan to build AI data centers with significant gigawatt capacity, likely spanning several years. This energy demand poses constraints: even if chips are delivered on schedule, power infrastructure may lag due to permitting issues, construction delays, and reliance on commodities like copper and rare earth metals. To address this, tech giants like Amazon and Google are investing billions in advanced nuclear power, including small modular reactors (SMRs), to secure reliable, carbon-free energy for data centers. These investments aim to stabilize long-term energy costs, reduce dependence on volatile traditional power markets, and utility companies. Alternatively, in Alberta, the province’s extensive gas and pipeline system is being positioned as a backbone and key enabler for off-grid data centers and AI facilities. For example, Pine Cliff Energy Ltd. will supply natural gas for 25 years to a planned off-grid Alberta data center adjacent to its extraction and production facility, beginning when the center is commissioned. Importantly, China is cutting energy bills at major data centers by as much as 50%, accelerating its semiconductor drive and sharpening its AI rivalry with the U.S, underscoring the global race to secure affordable, reliable power for next-generation AI infrastructure.
This intricate investment structure resembles post-war cross-shareholding models in Japan and South Korea, such as keiretsu, where interlinked corporate ownership secured supply chains. However, these systems were later criticized for masking financial flows and risks. If AI demand or monetization underperforms, or if access to energy and commodities falters, this web of cross-holdings could become highly vulnerable. China’s advancements in AI, including hardware development, have accelerated dramatically in 2025, positioning it as a formidable challenger and threat to U.S. dominance in the sector, and the U.S. AI ecosystem itself.
The AI industry’s investment ecosystem faces significant geopolitical risks, particularly in the supply chain for GPUs, TPUs, and rare earth metals. U.S. export controls on advanced semiconductors, such as restrictions on Nvidia’s high-end chips to China (which accounted for approximately a fifth of Nvidia’s 2023 revenue), threaten to disrupt global supply chains and limit market access. Reliance on commodities like copper and rare earths from geopolitically sensitive regions further exposes the industry to price volatility and supply shortages. For instance, trade disputes or mining disruptions could delay data center buildouts, compounding energy and infrastructure constraints. To mitigate this, tech companies are exploring diversified sourcing and domestic manufacturing, though these solutions require time and investment.
Regulatory pressures add further complexity. The EU’s AI Act, set to enforce strict compliance by 2026, and U.S. policies, including executive orders on AI safety, may increase operational costs by an estimated 5–10% for AI firms, diverting capital from infrastructure expansion. Regulations mandating model transparency or data usage restrictions risk uneven global adoption, with stricter regions lagging in AI monetization and threatening the revenue needed to justify the industry’s $5.0 trillion CAPEX forecast. Major players are lobbying for balanced regulations, and their financial strength provides some resilience, but navigating this evolving landscape remains critical to sustaining the interdependent investment model.
Early signs of market stress are emerging. For instance, rental prices for older AI chips reportedly dropped to $0.40–$2.00 per hour as of late 2024, below break-even for many operators. If infrastructure demand softens, stranded assets could result. Monetization remains uncertain: OpenAI reportedly has millions of weekly active users, with only a small percentage of paying customers. Large corporations experimenting with AI face challenges, including understanding technological limitations, managing expectations, training workforces, and optimizing data pipelines. However, successes like enterprise adoption of generative AI tools for customer service automation and predictive analytics suggest monetization potential.
The financial strength of major players provides reassurance. U.S. mega-cap tech firms are projected to generate substantial free cash flow in 2026, even after CAPEX, reflecting robust balance sheets. As of 2024, AI leaders trade at approximately 35 times forward earnings, lower than the 60 times multiples during the late-1990s internet bubble, and these companies generate tangible earnings. The market is betting on AI’s future profitability, not infinite growth. While the current cycle echoes the dot-com bubble, it appears grounded in stronger financial fundamentals and actual revenue generation. However, the intertwined nature of this ecosystem means that any underperformance by a major actor is poised to trigger significant consolidation across the sector and its collateral industries.
Globas Group is your trusted partner in navigating the complex landscape of the digital and AI industries. We guide you in overcoming technical and financial challenges while transitioning to a digital future. Our partner, duMonde Group, brings its expertise in communications, strategy, and policy.
Sources: Barron’s, Bloomberg, Financial Times, Goldman Sachs, Harvard Business Review, McKinsey Global Institute, Reuters, The Information, The Verge, The Wall Street Journal.