Article complete

NVIDIA's Crisis Has Only Just Begun, and There's No Jensen Charm to Save It

NVIDIA, without Jensen Huang, now has to deal with slowing AI infrastructure spending, tighter export controls, resistance to data centre expansion, hyperscaler competition, and mounting financial pressures on neocloud providers.

NVIDIA

Key takeaways

  • NVIDIA faces challenges like slowing AI infrastructure spending, tighter export controls, and competition from hyperscalers.
  • Despite impressive revenue growth, NVIDIA's stock has declined significantly, reflecting investor concerns about future capital expenditure.
  • The company will not reinstate Jensen Huang as CEO, leading to uncertainty in its leadership and strategy.
  • Analysts have raised earnings expectations but lowered the average price target, indicating mixed investor sentiment.
  • There are doubts about the sustainability of AI capital expenditure growth beyond 2027, impacting future valuations.

Takeaways drafted with AI from the article text.

Even as interim CEO Colette Kress reaffirmed that the company would not reinstate Jensen Huang, NVIDIA’s third-quarter fiscal 2027 results left little room for other complaints. 

Revenue more than doubled year-on-year to $115.8 billion; data centre revenue grew 110% to $107.6 billion. The company also raised its fiscal 2028 revenue growth forecast from 70% to 75%.

Yet NVIDIA shares fell 2.9% post-earnings, extending their decline to 36% from their October 6 record high. The stock had initially risen 5.2% in after-hours trading following the results, but the optimism did not hold.

Analysts also raised their average estimate of NVIDIA’s adjusted net income for the next fiscal year to $393 billion from $380 billion following the results. 

However, the average price target fell from $265.70 to $251.74, according to the LSEG analyst consensus data, per Reuters. 

Investors are effectively assigning a lower valuation to a company whose earnings expectations continue to improve. But why? 

The Deceleration in Capex

Richard Clode, Portfolio Manager at Janus Henderson Investors, wrote in an analyst report that the earnings beat and improved guidance had met investor expectations.

However, the debate has shifted towards 2028 AI capital expenditure. Much of the data centre construction planned for 2027 is already committed, leaving investors to question whether hyperscalers and AI companies will continue spending at the same pace for years ahead.

“This reminder of NVIDIA’s extraordinary financial prowess highlights the even more depressed valuation, which already embeds significant concern around future AI capex and a post-Jensen discount,” wrote Clode. 

The company’s immediate demand outlook is strong, but investors are wary about its durability. UBS estimated in July that capital expenditure by Microsoft, Amazon, Alphabet, and Meta would rise 76% to $673 billion in 2026, but growth would slow to 25% in 2027 and just 6% in 2028. 

A slower expansion in hyperscaler infrastructure budgets could eventually constrain orders for GPUs and other data centre products, particularly as major cloud providers invest in their own custom accelerators.

The company is also facing a problem with a recent financing plan. 

According to a recent Wall Street Journal report, a $500 billion financing initiative led by Huang, involving NVIDIA and six major Wall Street institutions, has yet to fund a single deal. The initiative was intended to help AI infrastructure companies finance GPU purchases, but negotiations have stalled over disagreements about who should bear the financial risks.

Investors are also weighing the long-term implications of Huang’s departure, including its potential impact on NVIDIA’s product roadmap, ecosystem partnerships, and financing initiatives, many of which were shaped by his personal relationships and direct involvement.

Counterpoint Research is looking out for the departure of senior leaders as the company searches for a permanent CEO. “Moving to a traditional hierarchy will test the organisation as long-time lieutenants adjust to new reporting lines and compete for leadership roles,” according to its research note.

Fresh Regulatory Scrutiny 

NVIDIA’s international distribution is poised to become more expensive, complicated, and difficult to scale, while exposing the company to greater regulatory and political risk. Even as the US earlier authorised sales of certain NVIDIA GPUs to China, investigators are examining the scale of purchases by its Asian customers of export-restricted chips that may have subsequently been diverted to Chinese buyers.

A Wall Street Journal report said Singapore-based Megaspeed (which also employed Alice Huang—allegedly linked to Chinese intelligence and to Jensen Huang) arranged nearly $12 billion in NVIDIA chip orders for Chinese companies between March 2024 and June 2025. 

Reports also suggest that Malaysia-based Speedmatrix continued receiving more than $800 million worth of NVIDIA technology for eight months after NVIDIA placed Megaspeed on its internal no-sell list last June.

The regulatory response is also expanding. BIS’ November 19 interim final rule requires licences for covered advanced-chip exports, re-exports, and domestic transfers to or within Indonesia, Malaysia, Singapore, Thailand, and Brazil, subject to exemptions for qualifying US-headquartered customers. It also extends controls to remote computing access. 

The impact is already visible. Chinese AI startup Moonshot AI scrapped a $150 million deal to rent NVIDIA Blackwell Ultra GPUs through an Indonesian data centre following the announcement. 

Will the Neocloud Bubble Pop? 

The other concern is NVIDIA’s customer mix and which customers to prioritise.

Its first set of customers are hyperscalers such as Google, Amazon, Microsoft, and Meta. These are NVIDIA's biggest customers, accounting for $60.1 billion in revenue in Q3, up 71% year-on-year. 

However, these companies are simultaneously developing their own custom silicon to reduce their dependence on NVIDIA. Anthropic trains and runs Claude across AWS Trainium, Google TPUs, NVIDIA GPUs, and AMD GPUs, using a proprietary system to route workloads based on price and performance.

OpenAI has partnered with Broadcom to develop Jalapeño, its first custom AI processor designed specifically for LLM inference, which it plans to deploy by the end of 2026.

Google, meanwhile, uses its seventh-generation Ironwood TPUs for both training and inference, including workloads involving Gemini. Its TPUs also power other frontier models, including Anthropic's Claude. 

Then there is NVIDIA's second customer base, which could be particularly vulnerable to a leadership change.

NVIDIA’s broader AI Clouds, Industrial, and Enterprise category generated $47.5 billion in Q3 revenue, up 18% sequentially and nearly three times its year-ago revenue. It includes neocloud providers, enterprises, and industrial customers.

Unlike hyperscalers, neoclouds and frontier AI labs are not developing competing accelerators at scale. They are therefore important to NVIDIA’s efforts to diversify its customer base. But many lack the balance sheets to finance enormous infrastructure deployments on their own.

Huang understood this dependency and actively supported the ecosystem. In an earlier interview with Dwarkesh Patel, he said, “If we didn't help CoreWeave exist, they would not exist.” 

He extended the observation to Nscale and Nebius, saying NVIDIA had supported their growth as well.

Despite this, neoclouds are already facing financing and execution pressures. 

CoreWeave reported $46.9 billion in debt, despite having a $132.5 billion revenue backlog, and expects adjusted profitability only in the first half of 2028. Nebius, meanwhile, cut its 2026 revenue forecast after delays in third-party data centre capacity, even as it raised its 2027 capital expenditure plans to $40–45 billion.

During NVIDIA’s Q3 earnings call, Kress said the company would remain “disciplined and selective” in supporting the ecosystem. Partners must demonstrate that they have secured financing and built operational data centre capacity before NVIDIA’s conditional guarantees take effect.

Venture capitalist Tomasz Tunguz believes that NVIDIA, by becoming more selective, could also weaken one of the customer segments it needs most to defend its long-term market position. In his blog, he wrote that a more conservative financial backstop could moderate the growth of smaller, capital-constrained cloud providers. 

“A diversified buyer base is NVIDIA's best defence against hyperscaler custom silicon,” said Tungusz. 

If neoclouds struggle to finance their expansion, NVIDIA could become increasingly dependent on Google, Amazon, Microsoft, and Meta—companies that are simultaneously investing billions in developing their own accelerators.

NVIDIA may struggle to replicate Huang’s charm offensive in securing its supply chain. He had personally negotiated with major suppliers, securing memory capacity in South Korea and chip manufacturing capacity in Taiwan.

“That stranglehold is personal. Jensen was the one who won these deals, CEO to CEO, with the major supply chain companies around the world," Semiconductor analysis firm SemiAnalysis wrote in a note.

Without Huang, suppliers may no longer prioritise NVIDIA to the same degree, making capacity more accessible to competitors. This would allow competitors such as AMD, Broadcom, and hyperscalers to secure greater access to previously constrained supply.

Published

Spotted an error? Tell our editors. Read our editorial standards and corrections.

More in this story

data centre · 3 stories
  1. AI Job Apocalypse or Productivity Boom? Silicon Valley Can’t Decide
  2. Telangana Targets 5 GW Data Centre Capacity by 2029 as Microsoft Opens Hyderabad Cloud Region
  3. You are here: NVIDIA's Crisis Has Only Just Begun, and There's No Jensen Charm to Save It
About the author

Senior Journalist

All stories
Belamy · Daily

The day’s AI news, before your first meeting.