Nvidia GPU Access Diversifies Across Clouds and Neoclouds
Nvidia Corp. anticipates $108 billion in revenue for the October quarter, a figure management projects will represent an 89% year-over-year increase. This growth trajectory accompanies a significant expansion in how companies access the chipmaker's GPUs, which remain in high demand for artificial intelligence workloads. The company's stock recently reached a new record, lifting its market capitalization close to $6 trillion. The proliferation of access options has created a complex landscape for enterprises requiring immediate computing power.
Nvidia Corp. anticipates $108 billion in revenue for the October quarter, a figure management projects will represent an 89% year-over-year increase. This growth trajectory accompanies a significant expansion in how companies access the chipmaker's GPUs, which remain in high demand for artificial intelligence workloads. The company's stock recently reached a new record, lifting its market capitalization close to $6 trillion. The proliferation of access options has created a complex landscape for enterprises requiring immediate computing power.
Traditional cloud infrastructure providers, including Amazon.com Inc., Microsoft Corp., and Alphabet Inc.'s Google, have historically led Nvidia's customer base. However, the supplier is diversifying its client mix. A recent filing indicated that five clients accounted for at least 10% of Nvidia's accounts receivable in the July quarter, up from three clients in January. Industry research firm SemiAnalysis identified 323 Nvidia GPU providers as of September, a significant increase from 209 providers less than 11 months prior. Jensen Huang, Nvidia's chief executive officer, stated at a Goldman Sachs tech conference in San Francisco that a new crop of neoclouds with hundreds of billions of dollars in backlog is emerging.
Hyperscalers retain a reputation advantage for enterprise clients, particularly for those relying on Microsoft Azure or Amazon Web Services. Bindu Reddy, CEO of AI assistant startup Abacus, noted that enterprise customers often require their subprocessors to be reputable brands like Azure. Anthropic and OpenAI have committed to spending over $500 billion combined with Amazon and Microsoft. Gartner data shows these two companies controlled 59% of the cloud infrastructure market in 2025. Hardeep Singh, a Gartner analyst, explained that hyperscalers benefit from over a decade of full-stack capabilities, though they may not always possess the specific GPU quantities enterprises require. Andy Jassy, Amazon's CEO, told analysts in July that the company cannot serve all anticipated demand this year and expects this constraint to persist through 2027.
Neoclouds have emerged to address capacity gaps left by hyperscalers. Modal, a startup operating virtual sandboxes for AI agents, transitioned from hyperscalers to using 25 different neoclouds to secure sufficient GPU volume, according to CEO Erik Bernhardsson. Major cloud providers are also engaging with neoclouds; Google and Microsoft have tapped CoreWeave to handle customers they cannot service directly. Marc Boroditsky, chief revenue officer of Netherlands-based neocloud Nebius, confirmed that hyperscalers have approached his company to manage overflow demand. Reactor, a video generation startup, uses GPUs from both Nebius and hyperscalers to ensure low-latency user experiences, according to CEO Alberto Taiuti.
Accessing neocloud capacity often involves upfront payments and lead times of several months. CoreWeave's near-term capacity is essentially sold out, CEO Mike Intrator stated during an August earnings call. Chen Goldberg, an executive vice president at CoreWeave, noted that turning over 10,000 GPUs to a new customer on short notice is difficult. Smaller specialist neoclouds offer more flexibility and bare-metal options but require customers to manage more technical aspects. Zhen Lu, CEO of Runpod, described supplier relationships as closely guarded secrets due to tight current capacity.
Oracle Corp. has introduced a bring-your-own-hardware model for GPUs, a strategy driven by its higher debt levels and lower credit rating compared to Amazon or Microsoft. Oracle CFO Hilary Maxson stated that this structure preserves margins and improves return on invested capital. John DiFucci, an analyst at Guggenheim Securities, suggested that Advanced Micro Devices and Nvidia might utilize this model. Meanwhile, SpaceX has entered the GPU rental market, agreeing in May to rent GPUs to Anthropic for $1.25 billion per month through mid-2029. Bret Johnsen, SpaceX's finance chief, reported that current economics yield a less-than-one-year payback on new compute capital deployments.
On-premises installations remain common as companies balance capability with cost control. Lenovo's Infrastructure Solutions Group saw enterprise and small business revenue nearly double in the June quarter. Vlad Rozanovich, a senior vice president at Lenovo, reported increased enterprise interest in bringing AI into their own facilities. Dropbox CEO Ashraf Alkarmi cited supply chain connections as a structural advantage for maintaining on-premises GPU usage. Everpure CEO Charlie Giancarlo stated the company acquired its own GPUs to run open-weight AI models for internal engineering teams.
Market pricing reflects the sustained demand. Ornn index data indicates that the cost to rent an Nvidia B200 GPU by the hour has more than doubled since March.