AI computing power rental demand surges, US stock computing sector collectively strengthens
On July 28, 2026, the US AI computing power sector experienced a strong rally. By close, NVIDIA (NVDA) rose 4.2%, AMD (AMD) gained 3.8%, Oracle (ORCL) jumped 5.1%, and data center REIT Equinix (EQIX) climbed 2.9%. The Philadelphia Semiconductor Index (SOX) also rose 2.3% on the day, with market sentiment notably improving.
Computing power rental model: shift from buying to leasing cards
The core driver of this rally comes from the explosive growth of the AI computing leasing market. As models with hundreds of billions of parameters become standard for enterprises, the high capital expenditure and maintenance costs of building data centers have deterred most small and mid-sized tech firms. A July 25 report from industry research firm Gartner states that the global AI computing leasing market will reach $62 billion in 2026, up 78% from 2025. Enterprises use NVIDIA H200, B200, and other GPU clusters through quarterly or annual subscriptions, avoiding depreciation risks.
In its July 27 earnings call, Oracle revealed that its cloud infrastructure (OCI) division saw Q2 revenue grow 112% year-over-year, with over 70% coming from AI computing leasing. CEO Safra Catz said, "We are experiencing the strongest demand cycle in history, with capacity expansion plans extending into H1 2027." Spurred by this, Oracle's stock surged for two consecutive days, also lifting other cloud service providers.
Three catalysts converge: earnings, policy, and tech iteration
- Earnings beat expectations: Besides Oracle, several AI infrastructure companies reported strong results. AMD's data center revenue grew 95% year-over-year, driven mainly by orders for the MI300X and the next-gen MI400 series. Although NVIDIA has not reported yet, market expectations for its Q2 earnings in late August have been raised to approximately 140% revenue growth year-over-year.
- Policy tailwinds: On July 26, the US Department of Energy announced the "AI Energy Resilience Initiative," providing targeted subsidies and tax credits for large-scale computing centers, encouraging efficient computing combined with green energy. This move lowers long-term costs for computing operators, benefiting REITs like Equinix and Digital Realty.
- Accelerated tech iteration: In mid-July, NVIDIA announced full-stack liquid cooling adaptation for its H200 GPU, raising single-rack power density in data centers to over 150kW, further boosting margins on high-end leasing. AMD launched the cloud-native MI400X, supporting the latest UCIe 2.0 interconnect standard, with improved compatibility attracting more tenant migration.
Valuation and risks: points to watch
Despite the bright outlook for computing leasing, the sector's valuations are at historical highs. NVIDIA's trailing P/E exceeds 65x, AMD over 55x, and Oracle above 40x. Investors should be wary of the following risks:
- Supply glut: Since early this year, TSMC's CoWoS capacity expansion has exceeded expectations, with GPU foundry delivery times shortening from 24 weeks to 16 weeks. If demand growth slows, lease prices may soften.
- Geopolitical uncertainty: The US Department of Commerce is considering further tightening export restrictions on AI chips to China, potentially impacting cloud providers' overseas expansion plans. Oracle, Amazon AWS, and others have data center footprints in Asia; a policy shift could hit revenue expectations.
- Interest rate environment: The Fed left rates unchanged at its July meeting but hinted at a possible 25-basis-point cut in September. If rate cuts are delayed, financing costs for capital-intensive computing leasing companies will remain high.
Outlook: How to seize investment opportunities in computing leasing?
According to multiple institutions, the AI computing leasing track still offers deterministic growth in H2 2026. In a July 28 research report, Goldman Sachs noted that as more enterprises transition from "AI experiments" to "AI production," demand for flexible and efficient computing will continue to surge. It recommends focusing on three types of targets: firstly, core GPU suppliers (NVIDIA, AMD); secondly, leading cloud leasing platforms (Oracle, Microsoft Azure, Amazon AWS); and thirdly, data center REITs and supporting infrastructure (Equinix, Vertiv).
Meanwhile, smaller-cap companies like C3.ai and BigBear.ai, while bearing more risk, may generate excess returns if they can create differentiated leasing solutions in vertical industries (e.g., healthcare, finance). Investors can allocate according to their risk tolerance.
In summary, the AI computing leasing model is reshaping the industry landscape, and the rally in related US stock sectors may have just begun.
