AI Trends

NVIDIA Quarterly Revenue $115.8 Billion: After Jensen Huang's Departure, China's Compute Market Accelerates Its Divergence

36Kr · AIUpdated Nov 18 09:51

AI summary

NVIDIA released its financial report for the third quarter of fiscal year 2027, with revenue of $115.8 billion, up 103% year over year, and net profit of $65.3 billion, up 105% year over year; data center business revenue was $107.6 billion, accounting for 93% of total revenue. This was the first earnings report after former CFO Colette Kress took over as interim CEO, and Chairman Harvey Jones explicitly stated on the conference call that the board does not intend to restore Jensen Huang's CEO position. Before the earnings release, nine companies in the industry chain that NVIDIA has invested in, including CoreWeave and Nebius, had cumulatively lost about $111.8 billion in market value; after the earnings report, some share prices recovered in after-hours trading.

Why read this

SelectedAI score 79

NVIDIA's earnings report and Jensen Huang's departure overlap; the original article sorts out the divergence in China's compute market and the stock-price transmission across the industry chain.

Topics

Tags

Body · AI translation

Author | Qiu Xiaopei

Editor | Yuan Silai

Amid the storm stirred up by Jensen Huang's departure, Nvidia delivered what may be its most closely watched earnings report in recent years.

At dawn Beijing time on November 18, Nvidia released its financial results for the third quarter of fiscal 2027. This was Nvidia's first report card since former CFO Colette Kress took over as interim CEO.

The earnings data showed that Nvidia's third-quarter revenue reached $115.8 billion, up 103% year over year; net profit reached $65.3 billion, up 105% year over year, both hitting record highs. Among this, data center business revenue reached $107.6 billion, up 110% year over year, accounting for 93% of total revenue.

Among this, based on the location of customers' headquarters, Nvidia's quarterly revenue from mainland China and Hong Kong was $11.05 billion, up about 272% year over year. This does not mean that the China AI chip business has recovered. In its financial guidance for the next quarter, Nvidia still did not include data center computing business revenue from China.

In just five days since the turmoil began, Nvidia's market value evaporated by about $800 billion, equivalent to a drop of 3.5 Alibabas. The strong earnings report injected a shot of confidence into the market. After the report was released, Nvidia's stock price rebounded in after-hours trading, rising about 4% at one point.

However, even after delivering an earnings report with both revenue and profit doubling, doubts in the capital market have not fully dissipated. The market subsequently gave back some of the gains, then rebounded again to about a 3.4% increase, showing that market sentiment remains volatile.

Nvidia's financial data for the first three quarters of this year, chart by HardKr

At least judging from the financial data, Huang's departure has not immediately shaken Nvidia's growth fundamentals. But the capital market's concerns have already gone far beyond attention to one company's own performance. The fate of Jensen Huang personally and of Nvidia as a company has become strongly tied to the future course of AI in China and the United States.

Over the past few years, Nvidia has leveraged its technological advantage in high-end GPUs to gradually build a vast industrial network spanning chip manufacturing, cloud computing, and AI model companies. And Huang himself was the weaver standing at the very center. He was a technology leader and, in a sense, a key bridge. His sudden departure, along with the continuously changing geopolitical environment, is bringing new uncertainty to this network.

In China, some cloud providers and AI companies have begun stepping up reserves of overseas computing resources, while others are accelerating adaptation to domestic GPUs. The uncertainty of high-end computing power supply is redistributing the interests of the domestic computing power market.

Facing rising geopolitical risks, Nvidia has also begun adjusting its internal governance structure. On the eve of the earnings release, Nvidia's board announced the creation of a new standing body, the "National Security and Export Controls Oversight Committee," to coordinate national security policy, export control compliance, and global trade risk management at the board level.

This also sends a clear signal: the model of relying heavily on Huang's personal mediation between Washington and global customers is being dismantled into a more institutionalized board-level oversight structure.

And in the U.S. capital market, a group of cloud service providers, optical communications companies, and chip manufacturers that Nvidia has been deeply tied to over the past few years have also suffered sharp stock declines one after another. The market has begun to re-examine the industrial relationships Nvidia built in the past.

Huang's latest remarks on the 17th

China's computing power amid the turmoil

Recently, HardKr's conversations with several people in China's computing power industry found that, in the face of Nvidia's uncertain future, the industry has formed a relatively consistent expectation: it may become more difficult for Chinese companies to obtain Nvidia's high-end computing power in the future. "If Nvidia is subsequently taken over by other management, the difficulty of getting cards will likely increase, and overseas scrutiny of the supply chain will continue," a cloud provider source told HardKr.

According to a media report on November 12, some Chinese cloud providers and AI companies have begun locking in overseas Nvidia GPU resources in advance, and some small and medium-sized model companies are also contacting computing power leasing providers in Southeast Asia, trying to lock in future computing resources as much as possible before supply policies tighten further.

But the situation turned faster than they expected.

Just four days later, overseas computing power channels also came under new regulatory pressure.

On November 16, the U.S. Department of Commerce issued new rules to further strengthen oversight of third-country data centers and remote access to advanced AI chips. This means that U.S. restrictions on computing power to China are extending from physical chip exports to overseas computing services. Even if GPU resources are locked in advance, they may not be usable on a sustained basis in the future.

Nvidia

However, tightening computing power supply is not bad news for all vendors. For cloud providers that have stockpiled large amounts of GPU resources in the past, changes in supply and demand may instead bring new business opportunities.

"The computing power shortage is actually a good thing for us personally. Business will definitely be easier next year. Recently, customers on hand are all rushing to ask for quotes and grab computing power," a cloud provider salesperson told HardKr.

When new high-end GPU supply is limited while demand for model training and inference is still growing, vendors that already hold existing computing power gain greater bargaining power. Conversely, companies lacking computing power reserves and still relying on new purchases or overseas leasing may face higher costs and supply risks.

On one side, the scarcity of existing computing power is being further amplified; on the other, uncertainty around new computing power and overseas leasing channels continues to increase. The Nvidia turmoil is intensifying the divergence in China's computing power market.

In fact, many domestic vendors had already sensed the shift in industry direction and adjusted their business direction in advance.

“We are already working on a second backup option, increasing the intensity of domestic adaptation to cope with potential market changes next year,” another cloud vendor source said.

The aforementioned cloud vendor source observed that before the turmoil, the players trying domestic computing power were basically central state-owned enterprises and state-owned enterprises, while startups' concern was time cost—startups that obtain financing need to quickly launch products, and without products they cannot continue to raise funds, so it is difficult for them to have sufficient patience for domestic adaptation. Therefore, domestic computing power was previously seen more as a supplementary option. But now, as uncertainty over overseas supply increases, domestic chips have also begun to assume the role of safeguarding more important matters.

One obvious change is that since the beginning of this year, the scope of domestic computing power adaptation has been expanding, gradually extending from a single model to multiple mainstream models and chip platforms.

In the past, domestic chips often needed separate optimization for different models. Now, in February this year, Zhipu GLM-5 has already adapted to multiple domestic chips such as Huawei Ascend, Cambricon, and Moore Threads; in the same period, Taichu Yuanqi has also become compatible with more than 30 mainstream large models; in September, Huawei Ascend entered PyTorch's official support system, meaning domestic computing power is trying to extend further from single-model adaptation to the underlying software ecosystem.

Jensen Huang

Policy signals have also been quietly released.

An industry source told HardKr that over the past two months, computing power subsidy programs in some localities have put forward stricter requirements on the proportion of domestic GPUs used—among them, some programs even require that the share of domestic accelerator cards must reach more than 80%.

An industry source judged to HardKr that the industry development process will inevitably encounter phased black swan events, but these shocks may not change the long-term trend. “Looking at it the other way, some phased adjustments may make the certainty of industry development stronger. There will be short-term pain, but it may be a good thing in the medium and long term, allowing this industry to develop more solidly.”

Nvidia after the Jensen Huang era

And outside Nvidia's headquarters in Santa Clara, USA, on the other side of the ocean, after Jensen Huang's departure, the white flowers people placed there have still not withered.

It is less that they are grieving for Jensen Huang than that they are mourning the passing of a beautiful world.

Someone placed flowers outside Nvidia's door to “mourn”

A “return and rescue” campaign around Jensen Huang has already begun.

According to foreign media reports, Jensen Huang's supporters are preparing to set up a “war room” involving Silicon Valley tech executives, trying to win White House support and push for his return. An online petition demanding his reinstatement has so far attracted more than 180,000 signatures.

For this chip giant founded by Jensen Huang and led by him for more than thirty years, the founder's unexpected departure is clearly not an ordinary management change. Behind it are more complex and massive contradictions and interests. Even if Jensen Huang ultimately manages to return, his style of operation will inevitably shift.

Before today's earnings conference call, Jensen Huang publicly stated that “if the board invites him, he is willing to come back to lead Nvidia.” In response, Nvidia Chairman Harvey Jones said firmly at the opening of the call that “the board does not intend to restore Jensen Huang's CEO position, and will not accept analysts' questions on this issue.”

Over the past year, Nvidia has completed a series of equity investments, procurement commitments, and technology cooperations.

From taking stakes in Intel, Synopsys, and Marvell, to investing billions of dollars in optical communications companies such as Lumentum and Coherent, to reaching capital and technology cooperation with Corning and Nokia, Nvidia's investment map has covered chip design, manufacturing, optical communications, and network infrastructure. On the downstream side, Nvidia has also injected capital into cloud service providers such as CoreWeave and Nebius to support the construction of computing power infrastructure.

In other words, Nvidia is no longer just a GPU supplier, but also plays multiple roles such as industry investor, important purchaser, and provider of customer capital. By locking up upstream capacity and supporting downstream computing power demand, Nvidia is trying to bring more companies into its own computing power ecosystem and build a massive semiconductor empire.

However, the deep binding that was once regarded as a competitive barrier has now instead become a source of crisis.

According to HardKr statistics based on public market data, from November 11 to 16, the stock prices of AI cloud service providers CoreWeave and Nebius, in which Nvidia has invested, fell cumulatively by about 17.9% and 30.7%, respectively; optical communications companies Lumentum and Coherent fell by about 15.7% and 12.3%, respectively, and Corning fell by about 11.9%; in addition, Intel and Marvell fell by about 11.2% and 13.9%, respectively; Nokia and Synopsys also fell by about 8.9% and 4.5%, respectively.

Before Nvidia's earnings conference call, the nine companies had lost a combined market value of about $111.8 billion, roughly equal to one Pinduoduo or two Meituan.

Declines of Nvidia supply chain companies before the earnings release, chart by HardKr

However, as Nvidia delivered better-than-expected earnings, market sentiment began to reverse. In after-hours trading after the earnings release, CoreWeave and Nebius rebounded by about 5% and 6%, respectively, and the stock prices of semiconductor companies such as Intel also recovered.

This round of market volatility, first falling and then rising, reveals another side of Nvidia's industrial layout: the industrial ecosystem Nvidia built in the past through capital and technology cooperation is becoming a network through which risk and return are transmitted to one another.

For Nvidia, the test at hand may not be how many more GPUs it can sell, but whether, after losing Jensen Huang, the company can continue, under the leadership of its newly appointed professional managers, to maintain the past pace of technological iteration, capital investment, and industrial cooperation.

end

Source

36Kr · AI