KPMG has indicated that graphics processing units or GPUs, long associated with gaming and high-performance computing, have rapidly evolved into a cornerstone of artificial intelligence infrastructure. Their parallel processing capabilities make them indispensable for training complex models, running inference tasks, and powering data-intensive applications across industries. As demand for AI surges, GPUs are transitioning from specialized hardware into a distinct alternative investment category.
Backed by sustained structural demand, chronic supply shortages, and revenue-generating leasing arrangements, they offer institutional and high-net-worth investors a pathway to participate directly in the AI economy.
KPMG’s recent analysis, the third in a series examining GPU dynamics, draws on a proprietary survey of 120 investment professionals and high-net-worth individuals.
It reveals strong appetite for technology-themed alternatives, with 31 percent of respondents having allocated capital to tech assets in the past three years—outpacing real estate (21 percent) and private equity or venture capital (20 percent).
Among those engaging with GPUs, the dominant motivations are clear: 70 percent seek meaningful capital appreciation, while 54 percent value portfolio diversification beyond traditional markets.
Innovation appeal (47 percent) and long-term wealth preservation (46 percent) also feature prominently.
Current allocations remain modest, reflecting the asset class’s relative novelty.
Many portfolios hold under 5 percent exposure, yet forward-looking intentions signal growth.
Nearly half of respondents anticipate scaling commitments substantially over the coming years, with optimism running high: 51 percent describe their outlook as very positive and another 24 percent as somewhat positive.
Pricing trends reinforce this confidence. NVIDIA’s latest Blackwell-series GPUs have risen 15-23 percent in recent months, with Ada models up 5-10 percent and lead times stretching to three-to-seven months amid fierce competition from hyperscalers, enterprises, and governments.
Leasing models provide the yield component that appeals to infrastructure-style investors.
Private equity funds and specialized operators are increasingly financing GPU clusters, colocation facilities, and “GPU-as-a-service” platforms. Similar themes appear in reports from other leading firms.
McKinsey highlights a multi-trillion-dollar race to expand AI-ready data centers, where GPU operators, colocation providers, and hyperscalers capture value through utilization rates and power-efficient designs.
Deloitte notes that AI chip sales could reach hundreds of billions by 2027, with financial services showing the fastest GPU adoption for applications such as fraud detection.
Goldman Sachs underscores hyperscaler commitments approaching $1 trillion by 2027, alongside the rise of “neocloud” providers building dedicated GPU capacity.
Risks remain front-of-mind. Sixty-one percent of KPMG’s respondents view GPUs as carrying modestly higher risk than typical alternatives, citing hardware price volatility, rapid technological obsolescence (chips often have four-to-six-year economic lives), regulatory scrutiny, energy consumption, and liquidity constraints in nascent vehicles.
Barriers to wider adoption include limited awareness (58 percent) and the need for clearer benchmarks and ESG-aligned structures.
Market maturation is underway. Investors are calling for more standardized products—ETFs, dedicated funds, and partnerships with established financial institutions—to improve accessibility and trust.
As AI permeates every sector and model complexity continues to escalate, GPUs are expected to mature into a stable, diversified holding akin to digital infrastructure or renewable energy assets. GPUs represent more than hardware; they embody the physical backbone of the AI evolution.