NVIDIA Scandal & AI Selloff In the US: India’s Neocloud Boom Faces Twin Tests

The alleged espionage scandal involving NVIDIA CEO Jensen Huang and the growing scrutiny of the chipmaker’s business dealings with China have added another layer of uncertainty to the global AI infrastructure market already facing questions about its financing model.  

At the same time, political and regulatory resistance to data centre expansion in the US could slow infrastructure investment, putting pressure on companies that have borrowed heavily to build capacity.

While unrelated, these developments present two separate risks to the investment model underpinning the AI infrastructure boom.

The first is already visible in the American stock market. Investors have been selling shares of US neocloud companies since September, with CoreWeave, Nebius and IREN witnessing sharp corrections. The selling intensified through October and into November, as concerns mounted over the sustainability of AI infrastructure spending and the debt these companies have taken on.

The selloff has only intensified talk of an AI bubble, and neoclouds in India face the risk of getting caught in the middle. 

Neoclouds – AI infrastructure providers that offer GPU-as-a-service – have emerged as an important part of NVIDIA’s AI ecosystem. They provide GPU compute to customers that may not have the scale or resources to build their own infrastructure. 

Neocloud business models are capital-intensive, with GPU procurement, data centre capacity, and financing arrangements determining how quickly they can scale up. NVIDIA has played a significant role in supporting this ecosystem through strategic investments and financing arrangements, helping expand the market for its chips beyond traditional hyperscalers. India is increasingly becoming part of this infrastructure buildout.

That buildout serves two demand pools. One is subsidised compute for domestic startups, researchers and institutions under the IndiaAI Mission, which had onboarded over 45,000 GPUs by June 2026. By August, it had approved 237 projects for subsidised compute, with 93.18 lakh GPU hours sanctioned. The other is commercial workloads from customers willing to pay for guaranteed capacity. The first highlights the scale of the government programme, but sanctioned hours do not necessarily translate into sustained usage. To justify their spending, India’s neoclouds are betting on the second.

Among the main Indian players, Yotta plans to spend over $7 Bn on AI infrastructure by the end of FY27 and is targeting a $1.5 Bn IPO, partly to repay debt. ESDS went to the market after the cost of the GPU servers it planned to buy rose from ₹25 Lakh to ₹1.8 Cr. E2E Networks, meanwhile, reported a 411% year-on-year jump in Q2 revenue.

Besides these, the likes of NxtGen, NeevCloud and Neysa are also building their neocloud infrastructure in India in anticipation of a big adoption wave. 

US giant CoreWeave is also entering India as the sole tenant of a planned 240 MW AdaniConneX data centre campus. However, any sustained pressure on its financing or expansion plans could have implications for the project. 

If concerns in the US around data centre build-up trickle down to India, neoclouds will be one of the first to get hit as a lot of their build-up depends on institutional funding commitment. The question is whether Indian companies will continue to enjoy the financial backing and attract customer commitments to withstand the headwinds if and when they come.

No Anchor Customers In India

The investment case for US AI infrastructure companies is underpinned by large customers such as Meta and OpenAI, whose scale and long-term compute requirements offer some visibility into future demand. These anchor customers make it easier for infrastructure providers to secure financing and justify large upfront investments.

Indian neoclouds, however, do not yet have comparable customer base to underwrite their expansion plans. International customers now account for 75-80% of Yotta’s customer base, partly because they are willing to pay more. Within India, however, the IndiaAI Mission is emerging as an important source of demand, although it utilised less than half its FY26 outlay of ₹10,000 Cr, which has directly impacted hardware procurement and deployment timelines, as Neysa mentioned in one of its blogs. 

A recent Moneycontrol report highlighted a mismatch in the Indian market: some GPUs are underused, while startups and researchers say fragmented capacity, allocation delays and the type of hardware on offer can make suitable compute difficult to obtain. A headline GPU count may therefore obscure both the quality of available clusters and the path from allocation to actual use. 

The mismatch makes domestic demand hard to read. The demand exists, but it does not always reach the neoclouds that have built capacity for it. Even well-funded companies are waiting. Juspay, which buys GPUs with its own cash, has been “given commitments that it will take three to six months” to receive them, co-founder and COO Sheetal Lalwani said at a media briefing. When large customers buy and run their own hardware, and smaller ones struggle to find suitable clusters, aggregate capacity figures say little about how much will turn into long-term, paying contracts for neoclouds. 

Can government-backed compute requirements offer the kind of long-term revenue visibility that hyperscaler contracts provide in the US? Or will Indian neoclouds remain dependent on international customers to sustain utilisation and justify their infrastructure investments?

GPU infrastructure requires substantial upfront capital, while the returns depend on sustained utilisation, predictable demand and long-term customer commitments. Notably, government-led regulations as well as citizen pushback could limit data centre build; it even emerged as a major factor for voters in the recent US midterm elections. For Indian neocloud players, which are increasingly catering to foreign and especially US-based clients, could suffer a trickle-down effect. Neocloud stocks, those of Coreweave, Nebius and Iren, among others, battered after the Democratic win in the recent elections. 

Closer home, ESDS and E2E stocks are already suffering from the second order impact of the sharp selloff in US tech stocks, following the mid-term results. On November 9, ESDS closed at approximately ₹1,147.25, against a September 24 high of ₹1,864.35, while E2E Networks ended at around ₹490.50, significantly lower than its October 14 high of ₹769.

The NVIDIA Question

NVIDIA’s role in the neocloud ecosystem goes beyond supplying GPUs. Its investments, financing arrangements and partnerships with infrastructure providers have helped bring new compute capacity online. But the relationship varies across Indian companies.

NeevCloud said its engagement with NVIDIA covers technical planning, GPU availability and deployment readiness. Procurement, financing and customer contracts are separate arrangements. The company acknowledged the concentration risk of relying on a single technology provider, but said it ties investments to customer commitments and available financing.

“The market correction reinforces the need for discipline. We assess each project on its commercial viability and phase investments accordingly,” Narendra Sen, founder of NeevCloud and RackBank Datacenters, told Inc42. 

ESDS said it has no direct commercial or financing relationship with NVIDIA. The company buys systems from OEMs and system integrators such as Supermicro, Dell, Lenovo and WWT. It has also signed a $1.25 Bn, seven-year agreement with Australia’s Sharon AI. Sharon AI owns and operates an 8,000-GPU cluster, which ESDS uses to serve its domestic and global customers. ESDS plans to explore similar partnerships.

“I would separate technology dependence from financial dependence,” said Piyush Somani, founder of ESDS. He pointed out that all of ESDS’s deployed GPU capacity runs on NVIDIA. 

Financing is a different matter. ESDS commits capex only after a customer has paid at least 35% of the project cost, and only then does it place orders with OEMs or system integrators, Somani said.

Both companies say financing for neocloud centres is getting more selective. Lenders and investors now favour projects with confirmed customers, predictable revenues and substantial upfront advances. Sen added that older GPUs could face more pricing pressure as newer generations arrive.

This is where the economic model gets complicated. NVIDIA has offered neoclouds worldwide a revenue-sharing and credit-support model in July, but later paused some of the deals, as reported by The Wall Street Journal. 

Yotta is exploring similar structures where partners fund and own the GPUs, while the company installs and operates them in its data centres and signs customer contracts. 

Gorilla, the Nasdaq-listed British tech company, is one such partner. In April, the two companies deepened their partnership, adding another $2.8 Bn to an agreement that was initially valued at $500 Mn. 

Such partnership models depend on GPU rental prices holding up. If compute prices fall or older GPUs lose value faster than expected, who absorbs the loss: the neocloud, the customer or the financing partner?

The Jensen Huang Overhang

NVIDIA’s chips, software stack and partnerships shape how quickly new compute capacity comes online. That is one of the reasons The Economist has called it the “central bank of AI”. This also highlights growing concerns about the AI ecosystem’s dependence on NVIDIA.

On November 8, NVIDIA’s board launched an independent investigation into reports that Huang was the target of a foreign intelligence effort. Huang has acknowledged an extramarital affair with a woman alleged to be spying for the Chinese government. The NVIDIA board, however, has maintained that it has no reason to believe national security or company secrets were compromised. 

Questions about her alleged links to Chinese intelligence, and what they mean for NVIDIA’s China business, remain open.

For Indian neoclouds, the direct exposure is less about a change in NVIDIA’s leadership and more about their dependence on globally-sourced GPUs and NVIDIA’s software ecosystem. The Government of India has said its compute ecosystem currently relies on globally-sourced GPUs and any change in chip availability, export controls, pricing or allocation could affect deployment schedules and project economics.

There is no evidence yet that the controversy has affected NVIDIA’s relationships with Indian neoclouds. But any fallout for its leadership, China business or financing strategy would hit an ecosystem that leans heavily on NVIDIA’s technology. There are concerns regarding what a potential wider restructuring of NVIDIA might mean for the neocloud business, especially if the build-up of data centres slows down in the US. 

Incidentally, NVIDIA is set to announce its third quarter earnings on November 17, 2026, and this will reveal much more about the underlying numbers for the data centre business. Will the corporate governance issues have an overhang on the stock till then? 

In February this year, the fear was that AI would trigger a SaaSpocalypse. Since then, models have advanced and the opportunities have grown bigger, but the concern now is whether the companies building AI infrastructure can carry the debt and capital commitments needed to keep this boom going.

The post NVIDIA Scandal & AI Selloff In the US: India’s Neocloud Boom Faces Twin Tests appeared first on Inc42 Media.