On June 10, 2026, Oracle delivered an earnings report that forced Wall Street to rethink the logic behind the AI investment cycle. The legacy database giant posted total revenue of $19.2 billion for FY2026 Q4, up 21% year-over-year, with cloud infrastructure revenue surging 93% to approximately $5.2 billion. Yet, these impressive figures paled in comparison to another metric—Remaining Performance Obligations (RPO). At the end of Q4, Oracle’s RPO reached $638 billion, up 363% year-over-year, with newly signed AI infrastructure contracts this quarter alone totaling $67 billion.
Just a year ago, Oracle’s RPO stood at $138 billion. In less than 12 months, it has grown nearly fourfold. This figure far exceeded analysts’ expectations of around $180 billion, compelling the market to reassess the true duration and depth of the AI infrastructure investment cycle.
Oracle is not an outlier. In 2026, the combined capital expenditures of the world’s five hyperscale cloud providers—Amazon, Microsoft, Alphabet, Meta, and Oracle—reached roughly $750 billion, up about 67% year-over-year, marking the third consecutive year of growth above 60%. According to the latest report from CreditSights, these companies’ capex-to-revenue ratios have climbed to unprecedented levels: Oracle at roughly 86%, Meta at 54%, Microsoft at 47%, Alphabet at 46%, and Amazon at 25%.
The relationship between capital expenditures and the AI infrastructure investment cycle has been a recurring topic over the past year. A central debate remains: will this massive spending translate into sustained revenue, or will it become a bubble of excess capacity? Oracle’s data provides a crucial validation signal—the sharp rise in RPO shows that the pace of actual contract signings on the demand side has not slowed; in fact, it’s still accelerating.
Cloud Infrastructure: The First Layer of AI Compute Demand Conversion
The most noteworthy detail in Oracle’s earnings isn’t just the RPO figure itself, but its composition. Of the $638 billion in RPO, about 12% is expected to be recognized as revenue within 12 months, and roughly 34% within 13 to 36 months. This means that over the next one to three years, Oracle has about $220 billion to $290 billion in additional committed revenue set to flow onto its income statement—a substantial growth driver for a company with annual revenue of around $67 billion.
At the same time, Oracle’s global GPU utilization rate hit 97.5% in Q4. This indicates that, at least for now, the market’s concern about "overbuilding not yet being utilized" does not hold. There is a structural time lag between supply-side expansion and demand-side consumption—a defining feature of the AI infrastructure investment cycle: first, invest in physical assets, then gradually convert that capacity into revenue through contracts.
Oracle isn’t the only beneficiary. Google Cloud, under Alphabet, saw Q1 revenue soar 63%, with its cloud backlog exceeding $460 billion. Microsoft’s AI business reached $37 billion in annualized revenue, up 123% year-over-year. Amazon AWS posted about $150 billion in annualized revenue, growing 28%. Collectively, these numbers point to a key conclusion: AI cloud services are transitioning from pilot projects to scaled deployments, with Oracle’s RPO data serving as one of the most direct indicators of this trend.
From an investment transmission perspective, cloud infrastructure providers are the first to benefit from AI capital expenditures. They convert capital investment into service capacity, then lock in future revenue streams through contracts. The efficiency of this chain directly impacts the sustainability of the entire AI infrastructure investment cycle.
Chips and Advanced Packaging: The Core Recipients of Capital Expenditure
The most direct recipients of AI data center construction are the semiconductor supply chains. According to a June 2026 report from Omdia, global AI infrastructure spending for the year will exceed $600 billion, with a significant portion flowing into GPU clusters, custom accelerators, and core compute components for data centers.
NVIDIA stands out as the most representative player in this chain. In Q4 2026, NVIDIA’s data center revenue reached $62.3 billion, up 75% year-over-year, while its networking business revenue skyrocketed 263%. More importantly, NVIDIA’s ecosystem is expanding into broader application scenarios. At WWDC in June 2026, Apple, NVIDIA, and Google jointly announced that Apple will use NVIDIA’s Blackwell GPUs in Google Cloud to support server-side inference for Apple Intelligence. This marks a pivotal customer win for NVIDIA in the emerging secure AI inference market.
The chip industry’s beneficiaries extend well beyond NVIDIA. TSMC, which dominates about 70% of the global chip foundry market, profits from the physical manufacturing of every AI chip. Advanced packaging is another critical bottleneck, with BE Semiconductor (Besi) and ASMPT holding the core equipment for AI chip advanced packaging. ASMPT’s research indicates that as long as the long-term trend for global AI data centers and AI PCs/phones holds, demand for high-end equipment will remain robust. UBS forecasts that memory semiconductor revenue will reach about $961 billion in 2026, with the DRAM market continuing to expand, driven by AI training and inference needs.
It’s important to note that the benefits of the AI chip supply chain are spreading beyond NVIDIA to a broader range of suppliers. However, due to the high concentration in upstream design and materials, risks are also concentrated at a few critical nodes—including supply bottlenecks, geopolitical shifts, and high customer concentration.
Power and Infrastructure Bottlenecks: From GPU Shortages to Energy Constraints
The AI infrastructure investment cycle is entering a new stage: bottlenecks are shifting from GPU supply to power and infrastructure. The U.S. Department of Energy projects that by 2028, data centers will account for 12% of U.S. electricity demand. This structural shift means the ceiling for AI infrastructure investment is moving from chip production capacity to grid capacity and power supply stability.
There are already clear examples of this transmission path. In June 2026, Fluence Energy announced a partnership with Siemens and NVIDIA to jointly develop power and electrical architecture solutions for AI data centers. Fluence will integrate its products to meet the grid stability requirements of AI workloads, including voltage and frequency fluctuation management, grid-independent restart capability, and AI load smoothing functions. Following this announcement, Fluence’s stock price jumped nearly 44% in a single day.
Bloom Energy also occupies a crucial position in AI power infrastructure. Oracle has signed contracts with Bloom to procure up to 2.8 gigawatts of fuel cell capacity for multiple data center projects.
Hydropower infrastructure and power distribution equipment suppliers are becoming important beneficiaries in the AI investment cycle. However, it’s important to note that these projects have much longer execution cycles than chip manufacturing. Whether grid upgrades can keep pace with data center construction is a key physical constraint facing AI infrastructure investment.
Humanoid Robots: The Application Layer Extension After AI Compute Scales
AI infrastructure investment isn’t limited to data centers themselves. Once compute capacity reaches a certain scale, its application scenarios begin to expand horizontally—demonstrated by the current investment boom in humanoid robots.
In June 2026, German humanoid robotics company Neura Robotics completed a Series C funding round, raising up to $1.4 billion with participation from NVIDIA, Amazon, Qualcomm, Tether, Bosch, Schaeffler, and the European Investment Bank, among others. The company’s valuation reached about $7 billion. This round included milestone-based payment terms, meaning funds will be disbursed only as the company hits specific targets—indicating investors are taking a more cautious, long-term approach.
A notable macro backdrop: Dealroom data shows that global robotics funding reached a record $55.8 billion in 2026, nearly double the previous year’s total. While Neura Robotics’ raise is impressive, it accounts for less than 3% of the global robotics investment pool. This suggests that capital inflows into humanoid robots have ballooned rapidly in the early stages of commercialization, but the industry remains in a highly uncertain, competitive phase.
Investments at this application layer are both an extension of AI infrastructure buildout and a long-term window for realizing AI’s commercial value. However, humanoid robots have yet to establish clear, scalable business models or profit paths, making investment returns highly uncertain.
Validation Signals and Risk Assessment in the Investment Cycle
The core data underpinning the current AI infrastructure investment cycle can be summarized at several key points. Oracle’s RPO for FY2026 Q4 hit $638 billion, up 363% year-over-year, with more than a third of contracts set to be recognized as revenue over the next 13 to 36 months—currently the most direct evidence of long-term demand in the market. CreditSights estimates that the combined capex of the top five hyperscale cloud providers will reach about $750 billion in 2026, up 67% year-over-year, covering GPU clusters, custom accelerators, data centers, and supporting power and cooling systems. Macquarie’s Viktor Shvets recently stated that the AI infrastructure investment cycle has become a global bubble, but is unlikely to burst in 2026 or 2027. Meanwhile, JPMorgan notes that as of the end of May 2026, AI-related debt accounted for about 15% of the entire corporate bond market.
Beneath the optimistic data, risk signals are also clear. Sequoia’s David Cahn calculates that there is an annualized gap of about $600 billion between hyperscale cloud providers’ AI capital expenditures and the actual revenue generated by the AI ecosystem, and this gap is widening in 2026. Allianz Research points out that the divergence between AI capex and revenue growth has reached about 46%, surpassing the 32% deviation seen during the 2001 telecom bubble.
Another signal worth watching comes from CoreWeave. After its IPO, shares of this AI data center operator rose more than 150%, but since the end of the lock-up period, its three co-founders have collectively cashed out about $2.3 billion, with Chief Strategy Officer Brian Venturo alone selling over $1.1 billion. Major institutional investor Magnetar Financial also sold about $5.5 billion in shares, halving its stake. While founder and early investor selling at high valuations doesn’t necessarily signal a deterioration in fundamentals—CoreWeave’s revenue is still up 111% year-over-year—it does serve as a warning that market sentiment may shift when valuations are elevated.
Conclusion
In summary, the current AI infrastructure investment cycle can be distilled into three relatively clear takeaways.
First, the scale of capital expenditure is still accelerating. The roughly $750 billion in capex by the top five hyperscale cloud providers in 2026 is quantitative proof that the AI infrastructure investment cycle has yet to peak. Oracle’s $638 billion RPO shows that contract signings on the demand side remain robust, with about a three-year window to convert physical buildout into contract revenue. This means that even in the most aggressive phase of supply expansion, revenue support is being built in parallel.
Second, the distribution of beneficiary assets is shifting from concentration to diffusion. From cloud service revenue growth, to chip design, advanced packaging, and power supply, all the way to capital flowing into application areas like humanoid robots, the AI infrastructure investment cycle now features a multi-layered transmission chain. The chip supply chain remains the most certain beneficiary, but capital attention is rising rapidly for power equipment, infrastructure hardware, and application-layer companies.
Third, the risks of supply-demand imbalances cannot be ignored. Behind the current scale of investment, there is indeed a "scissors gap" where capital expenditure growth far outpaces revenue growth, and a structural concern as AI-related corporate debt ratios rise sharply. Investors participating in this cycle need to closely monitor demand-side validation indicators such as RPO conversion rates and GPU utilization, rather than focusing solely on the absolute value of capital expenditures.
For investors looking to participate in this AI infrastructure investment cycle, Gate’s recently launched U.S. stock trading feature offers a new entry point. Through a strategic partnership with Alpaca, users can directly invest USDT on the Gate platform in over 10,000 stocks and ETFs listed on the NYSE and NASDAQ. From chip leaders like NVIDIA and TSMC, to power infrastructure players like Fluence and Bloom Energy, to cloud service providers like Oracle, and even publicly traded companies in the still nascent humanoid robot sector—all can be accessed in one interface.
The underlying logic of the AI infrastructure investment cycle is grounded in the real-world pace of physical buildout, not just market narratives. The answer to whether this cycle is genuine lies not in the numbers on an earnings report, but in every data center coming online and connecting to the grid.
