The AI trade is expected to shift capital expenditure from general infrastructure towards core chips, greatly benefiting Nvidia due to the shorter lifespan of GPUs compared to data center components. JPMorgan forecasts significant growth in chip financing and estimates over $3 trillion in AI chip investment over the next five years, with Nvidia poised to dominate GPU shipments. Despite concerns about overall AI return on investment and modest productivity gains, Nvidia’s strong market position and extensive partnerships suggest continued near-term leadership.
The massive rollout of artificial intelligence has heavily relied on foundational infrastructure, encompassing the construction of data centers, networking hubs, and power plants to support power-intensive algorithms. However, this trend is anticipated to evolve, with physical infrastructure likely to command a smaller proportion of capital expenditures in the coming years. Instead, the core component of this hardware — chips — is set to capture a larger share of investment.
This shift is primarily driven by the differing lifespans of components. Data centers, while substantial, can cycle through chips in just a few years, whereas larger infrastructure elements like plants can endure for decades. This dynamic presents a significant advantage for Nvidia, the leading manufacturer of Graphics Processing Units (GPUs).
According to JPMorgan strategist Tarek Hamid, the financing curve for data centers is expected to stabilize around 2028, but chip financing will likely continue its ascent into 2030, particularly as demand for replacements grows. Hamid projects that spending on GPUs and other AI-specific chips could rise from approximately 50% to 60% of total annual spending by 2030, in contrast to more general "data center box capex." The bank also forecasts over $3 trillion in financing for AI chips and essential hardware over the next five years, with silicon spending soaring to about $800 billion in four years from $340 billion in 2026.
Nvidia is already experiencing significant benefits from the AI boom, reporting $81.6 billion in revenue for the fiscal first quarter alone, an 85% increase year-over-year. CEO Jensen Huang emphasizes the company's "uniquely positioned" role at the heart of the AI transition, a position that will be further cemented by the growing allocation of capital expenditure towards GPUs. Nvidia CFO Colette Kress echoed this sentiment in May, projecting annual AI spending to reach $3 trillion to $4 trillion by the end of the decade, a forecast frequently cited by Wall Street analysts.
Despite this optimistic outlook, Nvidia's stock has lagged behind some other AI chipmakers this year, partly due to increasing demand for Central Processing Units (CPUs). While Nvidia has seen a more than 12% rise in 2024 (note: the original article says 2026, assumed typo for 2024 based on context), CPU maker AMD has more than doubled.
The shifting capital expenditure mix is intrinsically linked to component lifespan and depreciation. Data center components can last up to 30 years, as noted by UK computing services firm Infiniti, while a GPU's lifespan can be a mere tenth of that. This shorter lifespan, coupled with robust AI demand, is expected to drive increased purchases of Nvidia's graphics chips.
Overall AI capital expenditure forecasts continue to climb, with JPMorgan now anticipating total spending of $5.5 trillion through 2030, an increase from its November prediction of $5.1 trillion. However, questions persist regarding the return on investment (ROI). While hyperscalers like Amazon and Microsoft are reporting substantial revenue growth from their cloud services, sector-wide productivity gains, crucial for justifying the massive AI investments, remain unproven.
The U.S. experienced a modest 0.3% labor productivity growth in the first quarter, according to the Labor Department. Economist Dean Baker of the Center for Economic Policy Research highlighted concerns, stating that without blockbuster output growth, another quarter of weak productivity growth would ensue, questioning whether "job-killing AI is yet another economic myth, or the AI is much smarter than we think, and is hiding from the statistical agencies." Economists generally hold modest expectations for long-term AI-driven productivity gains, with MIT's Daron Acemoglu estimating no more than a 0.71% increase in total factor productivity over the next decade. Whether these gains will translate into the boosted profits necessary to justify increased spending is unclear, a concern growing among industry insiders.
David Linthicum, former chief cloud strategy officer at consultancy Deloitte, told CNBC that he expects actual AI capacity building to be about half of what was planned, citing construction starts data as a key indicator.
Nevertheless, in the near term, conditions are aligning for Nvidia to maintain its leading position in the AI trade. As the AI buildout progresses and investment pours into the sector, Nvidia's extensive reach across various facets of the industry sets it apart. JPMorgan projects Nvidia to ship 8.9 million GPUs this year, significantly outnumbering Google's 4.5 million comparable TPUs and Amazon's 1.9 million Inferenta and Trainium chips. While Google TPUs are expected to gain ground quickly, reaching 8 million shipments next year, Nvidia is still forecast to ship 9.9 million GPUs. The company also boasts a broad portfolio of hardware deals and partnerships, including those with OpenAI, Anthropic, Amazon, and Microsoft, as highlighted in a June 17 JPMorgan note.
