$750 billion AI infrastructure investment wave: Strategies and risks of NVIDIA, Google, and Oracle

Industry Background

Global investment in AI infrastructure is entering an unprecedented expansion cycle. Based on the latest financial reports from multiple tech giants, the market currently estimates that this round of investment has reached a magnitude of $750 billion. This figure is not an estimate but rather comes from the actual financial guidance of three representative companies: NVIDIA's revenue guidance for the second quarter of fiscal 2027 at $91 billion, Alphabet's capital expenditure guidance for 2026 at $180-190 billion, and Oracle's remaining performance obligations (RPO) of $638 billion (up 363% year-over-year). The aggregation of these three sets of data reflects that capital investment across the entire chain of AI infrastructure—from chips to data centers to cloud services—is accelerating.

NVIDIA founder Jensen Huang calls this process "the building of AI factories, the largest infrastructure expansion in human history." The driving force behind this expansion comes from strong enterprise demand for generative AI, large model inference, and AI agents, as well as the computing power arms race among hyperscale cloud providers.

Market Impact

The stock prices of these three companies have shown significant divergence in 2026. As of July 1, NVIDIA had risen 6.07% year-to-date, Alphabet had risen 14.2%, while Oracle had fallen 26.02%. The market's valuation divergence across different business models is enormous.

NVIDIA has demonstrated its absolute dominance in the AI chip space. In the first quarter of fiscal 2027, revenue reached $81.61 billion, up 85.2% year-over-year, with the data center business contributing $75.25 billion and data center networking revenue growing 199%. Non-GAAP gross margin remained at 75.0%, net profit grew 210.63%, and quarterly free cash flow reached $48.55 billion. The company's board approved an additional $80 billion stock buyback plan. However, supply-related commitments totaled $119 billion, indicating that capital expenditures will remain high.

Alphabet's Google Cloud business grew 63% to $20.03 billion, with backlog nearly doubling sequentially to over $460 billion. The company's first-quarter earnings per share were $5.11, far exceeding the market expectation of $2.63. The Gemini API processes 16 billion tokens per minute, up 60% quarter-over-quarter. However, free cash flow fell 46.63% due to more than doubled capital expenditures. CFO Anat Ashkenazi stated bluntly: "We are seeing unprecedented demand for AI computing resources, both internally and externally."Oracle is at the sharpest end of this investment wave. Its RPO reaches $638 billion, providing multi-year revenue visibility, with $75 billion involving prepayments or customer-supplied GPU arrangements, alleviating some capital burdens. Multi-cloud AI database revenue grew 404%. However, in order to catch up with infrastructure investment, Oracle plans to raise approximately $40 billion through debt and equity in fiscal year 2027, adding to its existing debt of $218.7 billion, putting pressure on its balance sheet. Free cash flow in 2026 was negative $23.69 billion, causing a sharp drop in stock price.

Competitive Landscape

Three companies play different roles in the AI infrastructure industry chain, forming a complementary yet competitive relationship.

NVIDIA is the "shovel seller", supplying GPUs and network equipment to all cloud vendors with a 75% gross margin. Its competitive advantage lies in its hardware ecosystem and CUDA software barriers, but it faces competition from AMD and custom chips (such as Google TPU, Amazon Trainium). NVIDIA's value lies in the scarcity and irreplaceability of its products, but in the long run, declining computing costs will compress gross margins.

Alphabet is the "miner", both renting computing power to enterprises through Google Cloud and using its own AI stack (Gemini model) to boost advertising and cloud service revenue. Its advantage lies in its massive user base and data flywheel, but it needs to balance internal AI R&D investment with capital expenditure returns. An operating margin of 36.1% indicates its relatively high monetization efficiency.

Oracle is the "follower", rapidly expanding cloud infrastructure capacity through aggressive debt leverage to secure long-term contracts from large enterprise customers. Its RPO growth is impressive, but negative free cash flow and falling stock price reflect market concerns about the sustainability of debt financing. Oracle's success depends on its ability to convert backlog orders into revenue and improve capital efficiency.

In addition, Microsoft, Amazon, Meta, etc. are also making large-scale investments, forming a broader competitive landscape. In the next 12-24 months, whoever can most quickly convert capital investment into revenue and profit will gain an advantage in the next phase.

Enterprise Insights

For enterprise decision-makers, the choice of AI infrastructure needs to balance performance, cost, and the risk of vendor lock-in. NVIDIA's GPUs remain the mainstream choice for large model training and inference, but consider multi-vendor strategies and the trend of custom chips; Google Cloud and Oracle Cloud offer different tiers of enterprise services. Oracle's multi-cloud AI database growth of 404% indicates strong demand for hybrid cloud architectures. Enterprises should pay attention to:1. Capex Pace: AI infrastructure investment is in full swing, but enterprises should not blindly follow the trend. Assess the ROI of your own AI application scenarios to avoid excessive reservation of computing power. 2. Supplier Financial Health: Oracle has high debt levels, requiring monitoring of its delivery capability and service continuity. NVIDIA and Alphabet have strong cash flows and are more resilient. 3. Contract Flexibility: Long-term contracts lock in prices while also locking in technology pathways; it is advisable to retain some flexibility.

Future Outlook

12 Months: NVIDIA's Blackwell architecture and next-generation GPUs will further enhance performance, with data center revenue likely to maintain high growth; Google Cloud benefits from the commercialization of the Gemini model, but capital expenditure pressure persists; Oracle will report Q1 FY2027 earnings (September 10, 2026) and hold an Investor Day (October 28), where RPO conversion efficiency needs to be validated. If Oracle fails to deliver on guidance as expected, its stock price may face further pressure.

24 Months: The proliferation of custom chips (e.g., Google TPU, Amazon Trainium) may weaken NVIDIA's monopoly, but the substitution effect will be limited in the near term. Demand for inference computing power driven by AI agents and multimodal models will sustain cloud revenue growth. If Oracle can reduce its debt leverage and achieve positive free cash flow within two years, its stock price may recover.

3 Years: AI infrastructure will shift from training to inference dominance, changing the structure of computing power demand. Companies with full-stack capabilities (chips + cloud + models), such as Alphabet, may prevail. NVIDIA needs to prove its competitiveness in the inference market, while Oracle must achieve breakthroughs in cloud market share. Global regulations (e.g., AI Act, data governance) may increase compliance costs but will not fundamentally alter the direction of investment.

Overall, the $750 billion wave of spending is no longer a prediction but a reality unfolding. The three companies are participating in different ways. Investors and enterprises should closely monitor their respective capital efficiency, contract quality, and actual revenue conversion capabilities.