AI Trends

NVIDIA Earnings Call: Q3 Revenue of $115.84 Billion, CEO Search Underway and Acquisition of Hugging Face Announced

East MoneyUpdated Nov 18 10:03

AI summary

On its fiscal 2027 third-quarter earnings call, NVIDIA disclosed quarterly total revenue of $115.84 billion, more than doubling year-over-year, with data center revenue of $108 billion, up 21% quarter-over-quarter, and raised its fiscal 2028 revenue growth guidance to approximately 75%. Board Chairman Harvey Jones said there is no intention to reinstate Jensen Huang and that the company has engaged Heidrick & Struggles to conduct a global CEO search; interim CEO Colette Kress said business continues as usual. The company also announced it has signed a definitive agreement to acquire Hugging Face, with the Vera Rubin platform contributing nearly $20 billion in revenue in its first full quarter of mass production.

Why read this

SelectedAI score 86

The earnings call disclosed key information including the management arrangements following Jensen Huang's resignation, the Vera Rubin ramp-up, and the acquisition of Hugging Face.

Topics

Tags

Body · AI translation

Main body EM_StockImg_Start

EM_StockImg_End

After the bell on Tuesday, US Eastern Time, Nvidia delivered its first earnings report of the "post-Jensen Huang era." The report showed that in the third quarter of fiscal 2027, both the company's revenue and profit significantly beat analyst expectations, while it also raised its growth guidance for the new fiscal year; however, multiple uncertainties, including personnel turmoil, new US export control rules, and temporary pressure on gross margin, still hang overhead.

On the earnings call, the board made clear it had no intention of reinstating Jensen Huang, while interim CEO Colette Kress emphasized that business was proceeding as usual, that the company's employees worldwide were working as usual every day, "seamlessly accomplishing what we needed to accomplish," and that "we have communicated with our customers, suppliers, and partners, and they support us 100 percent."

Looking ahead, although after the US midterm elections' "blue wave," computing demand may face potential suppression from US policy and regulation, Kress stressed: "Whether from 2027 or extending into 2028, demand driven by agents will continue, and we will sustain the volume of demand before us."

Key points from the earnings call

I. Personnel and governance

1. Board Chairman Harvey Jones said in his opening remarks: the board has no intention of restoring Jensen Huang to the CEO role and has engaged Heidrick & Struggles to conduct a global CEO search, considering both internal and external candidates, and will not make a hasty appointment just for speed.

2. A National Security and Export Control Oversight Committee was established, composed of independent directors, external advisers, and legal counsel, to strengthen communication with US regulators.

3. Colette Kress said that during the transition "business is as usual," with customers, suppliers, and employees remaining stable; core supply chain partners such as TSMC continue deep collaboration, and the technology roadmap is unaffected by the personnel changes.

II. Core financials and business guidance

1. Q3 results (fiscal 2027 Q3): total revenue of $115.84 billion, more than doubling year over year; adjusted EPS of $2.68, both far exceeding expectations; non-GAAP gross margin of 74.9%, above guidance thanks to product mix optimization. Data center business revenue was $108 billion, up 21% quarter over quarter.

2. Q4 guidance: revenue of $130 billion, plus or minus 3%; Q4 marks a temporary bottom in gross margin (71%-72%), expected to recover to 72%-73% in fiscal 2028. Q4 includes an extra week, and transition-related costs push up operating expenses.

3. Raised fiscal 2028 outlook: revenue growth year over year raised from 70% to 75%, based on a higher fiscal 2027 base, supported by recovering end-market AI demand and supply chain capacity expansion.

4. Record shareholder returns: Q3 returned a total of $28 billion to shareholders; an additional $150 billion share repurchase authorization was added, with $227 billion remaining under the repurchase authorization, planned to be completed by the end of fiscal 2028.

1. The Vera Rubin platform is ramping faster than expected, contributing nearly $20 billion in revenue in its first quarter on the market, the fastest volume ramp of any product in Nvidia's history; the next-generation Vera Rubin Ultra will launch next year, with Feynman scheduled for 2028. The DSX POD modular AI factory architecture has been deployed, significantly shortening data center deployment cycles.

2. Announced the acquisition of Hugging Face, stepping up investment in the open-source large model ecosystem; agentic workloads account for about 80% of total token volume; GPU rental prices continue to rise, with Hopper and B300 rental quotes rising significantly, and computing capacity across the entire platform remains at full utilization.

3. Networking business performed strongly, with CPO co-packaged optics entering mass production; the Vera CPU has been adopted by multiple leading customers including Meta, OpenAI, and CoreWeave.

IV. Supply chain, guarantees, and customer financing

1. Expanding substrate and advanced packaging partner manufacturers to further strengthen the global supply chain; Q3 purchase commitments reached $390 billion.

2. Total guarantee scale of $118 billion; backstop guarantees for neocloud providers are subject to preconditions taking effect.

3. Vendor financing will prudently focus on the three major data center infrastructure elements of land, power, and buildings; customer procurement follows uniform pricing, and capacity allocation prioritizes customers' infrastructure readiness and funding availability, with no price favoritism.

V. Responses to market concerns and risks

1. Regarding market worries that policy changes in 2028 could trigger an AI downturn: management believes demand driven by agentic AI is sustainable, the current shortage of computing supply will continue, and the company will keep communicating with all US government departments.

2. On long-term pricing and gross margin: the fourth quarter is the low point for gross margin; relying on full-stack software and hardware synergy advantages, the company is confident in product pricing, but does not rule out dynamic adjustments in the future based on the competitive environment.

3. Frontier AI labs contributed about 15% of data center revenue this year; cloud business revenue growth at hyperscale cloud providers has accelerated for six consecutive quarters, and the cloud industry's backlog has reached $2.5 trillion.

Full transcript of Nvidia's fiscal 2027 third-quarter earnings call (AI-assisted translation, some content abridged)

Toshiya Hari (NVIDIA Investor Relations): Good afternoon, and welcome to NVIDIA's fiscal 2027 third-quarter conference call. Joining me today are Colette Kress, Interim Chief Executive Officer, Executive Vice President and Chief Financial Officer, and Harvey Jones, Chair of the Board of Directors. I will now hand the call over to Harvey for opening remarks.

Harvey Jones (Chair of the NVIDIA Board of Directors): Thank you, Toshiya. Good afternoon, everyone, and thank you for joining us today. As Chair of the NVIDIA Board of Directors, I will say a few words on behalf of the Board. NVIDIA is an exceptionally strong company, and its strength stems from outstanding people, breakthrough technology, and an exemplary execution record. Our financial performance speaks for itself.

On November 12, the Board announced that, following the disclosure of personal conduct inconsistent with the company's code of conduct, standards and culture, Jensen Huang voluntarily resigned from his positions as President, Chief Executive Officer and member of the Board of Directors. The NVIDIA Board unanimously accepted his resignation and has no intention of having Jensen Huang return as CEO. The Board has appointed long-serving and highly respected Chief Financial Officer Colette Kress as Interim CEO.

In addition, the Board announced that it has initiated a formal search for a permanent CEO, and we have engaged Heidrick & Struggles to lead this search. Although this is an unexpected leadership transition, the Board's actions have been deliberate and are in the best interests of the company's employees, customers, partners and shareholders.

This morning, the company announced that the Board has established a "National Security and Export Controls Oversight Committee," whose members will include certain independent directors, senior external advisors and legal counsel. This committee will strengthen communication with the US government, including all relevant agencies.

During today's call, we will not provide information or take questions regarding the circumstances of Jensen Huang's resignation or the ongoing investigation; the Board will provide further updates in due course.

Acting on behalf of shareholders, the Board is proactively addressing the immediate consequences facing the company and strengthening its oversight capabilities for the future. We have an outstanding leadership team, and we have full confidence in Colette's leadership, character and ability to serve as Interim CEO. For the past 13 years, Colette has been at the center of the company's extraordinary growth and has played a key role in executing its strategy, with a deep understanding of the company's business, customers, operations and financial position. The Board and the entire leadership team remain fully focused on driving NVIDIA forward, and our commitment to creating value for shareholders, customers and partners is unchanged. I will now turn the call over to Colette.

Colette Kress (Interim CEO, Executive Vice President and CFO of NVIDIA): Thank you, Harvey. Our growth momentum is real. With an industry-leading multi-year technology roadmap, supply agreements extending through the end of this decade, and sustainable demand, we are positioned to further accelerate NVIDIA's leadership. Our ambition will not change, our passion for innovation will not change, and our commitment to customers, partners and shareholders will not change.

I will now move to the financial results. We delivered another outstanding quarter, with revenue, operating income and earnings per share all reaching record highs. Total revenue was $116 billion, more than doubling year over year and up 20% sequentially.

Vera Rubin had an extraordinary start: in its first quarter of volume shipments, the Rubin platform contributed nearly $20 billion in revenue, making it the fastest-ramping product in NVIDIA's history. We are meeting strong compute demand from a diverse and broad set of end customers, including hyperscalers, frontier labs, AI-native companies, enterprise customers and governments.

Combined with additional support on the supply side, we have raised our fiscal 2028 outlook: we now expect revenue growth of approximately 75% year over year, up from 70% previously, and on top of a higher fiscal 2027 base. Our outlook takes into account data center readiness, the strength and durability of end demand, and our customers' financing capacity.

Third-quarter data center revenue grew 21% sequentially to $108 billion, with balanced growth across hyperscale cloud and AI cloud, industrial and enterprise segments. Within that, hyperscale cloud revenue was $60 billion, up 23% sequentially, driven by the Vera Rubin ramp and continued strong performance from Grace Blackwell. Cloud business revenue growth at our hyperscale cloud customers has now accelerated for six consecutive quarters. The cloud industry's backlog has now reached $2.5 trillion, up from $2 trillion three months ago, underscoring the enormous multi-year opportunity ahead for our partners and for NVIDIA.

OpenAI and Anthropic continue to drive frontier progress with their own models: OpenAI released its latest GPT-6 series, and Anthropic released its Claude 5.5 series. As frontier model capabilities improve, applications and revenue are scaling rapidly; by the end of October, Anthropic's annualized revenue had reached $81 billion. We expect the leading frontier labs to contribute about 15% of our data center revenue this year. Open-source models are accelerating AI adoption: weekly token volume on OpenRouter grew 35x year over year, and 3.5x in the past three months alone. Agentic workloads now account for about 80% of total tokens. A rapidly expanding ecosystem of inference providers is bringing these models to enterprises everywhere. At NVIDIA, we are committed to fostering a vibrant open-source model ecosystem so that every country, every company, and every industry can build its own intelligence.

In September, we signed a definitive agreement to acquire Hugging Face. Hugging Face is the home base of the open-source model community, serving more than 18 million developers and 200,000 enterprises. Together, the two sides will expand its platform to make intelligence more open, more powerful, and more accessible. Nemotron is currently the most widely used U.S. open-source model on OpenRouter and the starting point for developers building their own intelligence. This foundation is empowering the entire ecosystem. This quarter, leading enterprise software and security companies—including Microsoft, Salesforce, Palantir, CrowdStrike, Cadence, and Synopsys—announced products built on post-trained open-source models and customized for specific domain use cases. Training partners such as Scale, Mercor, Prime Intellect, and Applied Compute are helping enterprises turn open-source models into their own models based on their own data, tools, and workflows.

In vertical domains, startups such as Lila Sciences, Harvey, Therma, and Abridge are using post-trained open-source models to deliver results in science, law, and healthcare. All of this is forming a powerful flywheel.

Since May, token prices have fallen nearly 60%, accelerating token usage and compute demand. As a result, rental prices for every GPU generation are rising. Hopper GPUs are renewing under multi-year agreements at up to 2x their original price, and B300 prices have risen more than 60% since launch. Every hyperscaler and neocloud we serve is running NVIDIA compute at full utilization, which confirms the productivity, durability, and fungibility of our infrastructure.

As agents run longer and become more autonomous, enterprises need stronger control mechanisms to ensure safety. The "NVIDIA Open Agent Safety Platform," which we released together with more than 100 partners, is precisely intended to provide that control.

OpenShell constrains agent behavior by limiting agents' access to files and tools; NVIDIA Sentry, running on BlueField-4, monitors behavior at the chip level and stops out-of-bounds agent operations within milliseconds. OpenShell has received broad ecosystem support from partners in AI models, enterprise software, cybersecurity, cloud infrastructure, robotics, and other fields. Companies working with NVIDIA include Anthropic, AWS, CoreWeave, CrowdStrike, Dell, Google Cloud, HPE, IREN, JPMorgan Chase, Lenovo, Microsoft, Nebius, OpenAI, and SpaceX AI. Advances in agent harnesses are accelerating the adoption of agentic AI. After Meta's Muse agent launched on NVIDIA infrastructure, installations in the U.S. and Canada exceeded 10 million within six weeks. To support this massive scale and future growth, Meta plans to expand deployment of Vera Rubin hardware, with the goal of maximizing performance and efficiency for its next-generation AI agents.

Agentic AI is driving extraordinary demand for CPUs, far exceeding supply. NVIDIA Vera CPU is precisely suited to meet this growing demand, with leading performance in data center processing and scientific simulation workloads. CoreWeave, Meta, Nscale, and OpenAI have now adopted Vera, joining leading partners such as AWS, OCI, and SpaceX AI that we announced last quarter. We continue to see strong demand for Vera and expect total CPU revenue to more than double in fiscal 2028. The AI Cloud, Industrial and Enterprise (ACIE) business had revenue of $47 billion, up 18% sequentially and nearly tripling year over year. Sequential growth came mainly from delivering Vera Rubin to leading neocloud customers and from frontier labs beginning to build their own data centers. Neoclouds are playing an increasingly important role in the global AI infrastructure ecosystem, providing more customers with access to NVIDIA compute and expanding the reach of our platform.

We have always supported the growth of this ecosystem and will continue to be disciplined and selective going forward. NVIDIA's full-stack AI technology platform gives us a unique advantage in capturing opportunities in sovereign AI, enterprise, and physical AI. Sovereign AI revenue more than tripled year over year in the third quarter. In September, we announced partnerships with several NVIDIA cloud partners in Australia to build 2 gigawatts of land, power, and shell capacity by the end of 2027, with the opportunity to host multiple generations of NVIDIA DSX AI factories in the future. Within enterprise, financial services saw the strongest growth, with on-premises revenue approaching $5 billion on a trailing 12-month basis, more than doubling year over year. As more customers use NVIDIA infrastructure to accelerate quantitative trading, we expect enterprise revenue to more than double in fiscal 2028.

Physical AI revenue nearly tripled year over year in the third quarter and is on track to exceed $25 billion for the full year. At GTC Berlin, we expanded the NVIDIA Isaac robotics platform and introduced new humanoid robot reference designs based on Isaac GR00T and Jetson Thor. In automotive, NVIDIA Hyperion and Alpamayo are helping Mercedes-Benz and Bolt (in partnership with Lucid) build robotaxi fleets, helping Waabi develop L4 autonomous trucks, and helping Volvo advance safe autonomous driving. Edge computing revenue was $8.2 billion in the third quarter, up 14% sequentially and 42% year over year. Sequential growth came from stronger enterprise demand for Blackwell workstations and from edge compute demand for physical AI applications. Revenue from licensed Hopper data center products sold to China was $2.4 billion in the third quarter. Our guidance does not include any licensed Hopper data center products.

Next, supply chain, financial commitments, and guarantees. Expanding the global supply chain to support customer growth remains our top priority. Because demand continues to exceed supply, we are working closely with partners to accelerate capacity expansion in key areas such as wafers, substrates, and advanced packaging. Today, we announced new agreements with Kinsus and Samsung Electro-Mechanics in substrates and with SPIL in advanced packaging, further strengthening the capacity and resilience of the global supply chain. These agreements build on the partnerships announced last month with Amkor, KYEC, and Unimicron. In the third quarter, our purchase commitments increased to $390 billion, reflecting our confidence in the long-term sustainability of demand for NVIDIA infrastructure. We are securing the data center and cloud capacity needed to support the development of hardware, software, and an increasing number of open-source models.

In total, our data center lease and cloud service commitments increased to slightly more than $100 billion, driven primarily by long-term leases. Commitments related to new cloud agreements totaled $32 billion at the end of the third quarter. Under these agreements, we earn revenue from front-end sales of infrastructure and share in the upside from lease revenue above the minimum commitment. Total guarantees were $118 billion at the end of the third quarter, including the $105 billion in land, power, and shell credit support for the PORTS-Pike campus that we discussed last quarter. The campus will be dedicated to OpenAI's use of NVIDIA infrastructure under a 20-year lease. The SB Energy project is proceeding as planned, with capacity expected to come online in phases starting in 2028. We continue to work closely with infrastructure capital partners, who play a key role in financing AI infrastructure buildout.

As we move forward, we will be prudent in how we participate and disciplined in our capital commitments. We are proud of the role we have played in supporting the ecosystem to date. As the market matures and more sources of capital emerge, we are rigorously evaluating future investments, commitments, and guarantees. NVIDIA's extreme co-design capabilities across compute, networking, systems, algorithms, and software are unmatched, driving leapfrog improvements in performance per watt and performance per dollar on an annual cadence. On the MLPerf Endpoints benchmark, Vera Rubin delivers 10x the tokens per watt of GB300. Vera Rubin is still early in its lifecycle, and we expect continued software optimization to deliver higher performance over time. The transition from Grace Blackwell to Vera Rubin has been exceptionally smooth.

Just three months after mass production and delivery began in August, mass-produced racks have already been deployed in the data centers of every major hyperscaler and leading neocloud. Our hyperscale cloud partners brought Vera Rubin to mass production in about one month, half the time Grace Blackwell required. Networking continues to expand, with third-quarter revenue exceeding $21 billion. Our co-packaged optics (CPO) roadmap is on track, and CPO scale-out has entered mass production. At SC26, Lambda announced the first mass deployment of NVIDIA Photonics CPO switches, connecting more than 10,000 GB300 GPUs. Compared with pluggable optical modules, NVIDIA Photonics switches deliver 5x the energy efficiency, 15x the reliability, and 10% lower latency. We remain on track to launch Vera Rubin Ultra next year and Feynman in 2028. The next great leap is already on the horizon. At GTC Berlin, we introduced DSX POD, a full-stack modular AI factory reference architecture.

Just as Grace Blackwell NVL72 moved rack assembly from the data center into the supply chain, DSX POD extends that approach to the entire cluster. The POD arrives ready to connect, cutting deployment time from months to weeks and accelerating time to first token. At an annualized revenue of nearly $50 billion for a 1-gigawatt AI factory, producing the first token one quarter earlier can bring operators more than $10 billion in additional revenue. We are working with more than a dozen partners, including Dell, Foxconn, Schneider Electric, and Vertiv. We plan to deliver the first PODs to customers in the United States and central Europe early next year. Now to the rest of the income statement. GAAP and non-GAAP gross margins were 74.8% and 74.9%, respectively, above guidance on a richer product mix.

GAAP and non-GAAP operating expenses rose 9% and 7% sequentially, respectively, mainly on higher compute and infrastructure spending. Operating expenses were slightly below the outlook, mainly due to the timing of NRE expense recognition. The non-GAAP effective tax rate was 17%, up from a year ago, primarily due to higher revenue. In the third quarter, we returned a record $28 billion to shareholders, including $22 billion in share repurchases and $6 billion in cash dividends. Year to date, we have returned 65% of free cash flow. Looking ahead, we plan to return all excess free cash flow, after strategic uses, to shareholders this quarter and beyond.

This quarter, we approved an additional $150 billion increase to our share repurchase program, the largest in history; remaining authorization at the end of the third quarter was $227 billion. We still plan to complete the repurchase authorization by the end of fiscal 2028.

Now, I'll turn to the fourth-quarter outlook. Please note that the fourth quarter will have an extra week. Total revenue is expected to be $130 billion, plus or minus 3%. We expect sequential growth to be driven by strong demand for AI infrastructure and continued ramping of Vera Rubin.

The U.S. Department of Commerce released new export control measures this week, but because demand far exceeds supply, we expect results to remain strong. As noted, we now expect fiscal 2028 revenue to grow about 75% year over year on a higher base. GAAP and non-GAAP gross margins are expected to be between 71% and 72%, consistent with the outlook provided last quarter.

We expect the fourth quarter to be the trough in gross margin, after which it will recover to the 72% to 73% range in fiscal 2028. GAAP and non-GAAP operating expenses are expected to be approximately $11.0 billion and $10.7 billion, respectively, affected by the extra week and transition-related costs. We will continue to prioritize investments that support market growth and strengthen our market position. For full-year fiscal 2027, we still expect a GAAP and non-GAAP tax rate of 16% to 18%, excluding any discrete items and major changes in the tax environment. Finally, we plan to file our Form 10-Q in the coming weeks; in the meantime, a document with supplemental financial information is available on our investor website.

Toshiya Hari: Thank you, Colette. We will now move to the Q&A session. Ian Buck, Vice President of Hyperscale and High-Performance Computing, will join Colette to take questions. Operator, please begin collecting questions.

Operator: Thank you. Please limit yourself to one question each, thank you. Our first question comes from Joseph Moore of Morgan Stanley. Please go ahead.

Joseph Moore (Morgan Stanley Analyst): Okay, thanks. I wanted to ask about the 75% growth outlook for next year. I know you've previously been constrained by supply of wafers and other things. What gives you the confidence to raise that number? Do you still think demand will be stronger than that number? Previously you mentioned thinking about 100% under the new BIS rules; do you still think 100% is the right way to think about demand?

Colette Kress: Yes, thanks Joe. That's right. As we look at the next fiscal year, we had originally indicated about 70% year-over-year growth, and we are now raising that to 75%. We spent a lot of time evaluating both the continued growth in demand and the additional supply we can provide to different customers. Together, these continue to bring us more growth opportunities. In addition, we also evaluated the impact of current export controls and will spend some time studying their various provisions. We are confident in the demand in front of us and in the (incremental) supply we can provide, so we can achieve 75%.

Operator: Our next question comes from Vivek Arya of Bank of America Securities. Please go ahead.

Vivek Arya (Bank of America Securities Analyst): Thank you for taking my question. Colette, I understand some decisions are still evolving, but I'd like to understand: regarding the company's structural moat and talent retention, what do you want to convey to Nvidia investors? What is sustainable? What might change during this leadership transition? How do you and the board view the company's competitive positioning and financial strategy during the transition?

Colette Kress: Let me start, and then I'll ask Harvey to add a few words. The most important thing we are considering right now is our employees. Employees are absolutely essential and indispensable to everything at the company. We always keep our employees in mind. As with any group, there will be differing opinions. But we believe our employees are engaged and willing to devote their careers to Nvidia, and that has been true since the company's founding. During the transition we will take care of our employees, ensure they are safe, and ensure that work can be completed in the current environment. At the same time, this is also a moment to think about leadership and how important leadership is to the entire organization.

Over the past week, our leadership has stood up very firmly, both in supporting us in completing the financial reporting work here and in supporting all employees around the world. Supporting them in going about their work as usual every day is something I firmly believe they will do. But frankly, they have performed even better than that, seamlessly completing what we needed to accomplish. Please remember that we have the world's best technical experts and a world-class multi-year roadmap, and that is important.

From a business perspective, it is business as usual. We have communicated with customers, suppliers, and partners, and they support us 100%. Customers have expressed support for the business outlook and are helping us through this period. They are also working closely with our suppliers to close the supply-demand gap. This week, TSMC's leadership was in our offices and is fully assisting us.

We have locked in memory and advanced packaging. I think the most important thing is the situation inside the company and our strength to keep moving forward. I'll hand it to Harvey to add.

Harvey Jones: Colette covered that very well. From the board's perspective, we are taking this search very seriously. We will consider both internal and external candidates. The board understands that we are in a world where advanced silicon technology and the national security landscape are profoundly changing the market, and the new CEO must navigate these waters very carefully. The search will be very thorough, and we want to appoint a permanent CEO as soon as possible, but we will not rush before we find the truly right person.

Operator: Our next question comes from Timothy Arcuri of UBS. Please go ahead.

Timothy Arcuri (UBS Analyst): Thanks a lot. Colette, I wanted to ask about the new commentary on gross margin—you mentioned a range of 72% to 73% in calendar 2027. I'd like to understand your ability to sustain pricing power and margins over the long term. I ask because frontier model developers are shifting more aggressively to ASICs and using compute efficiency as a new KPI. Given that, how much gross margin compression would you accept to protect your competitive position?

Colette Kress: Thanks Tim. We discussed gross margin on the last earnings call, covering not only the last two quarters of fiscal 2027 but also giving an outlook for fiscal 2028. If you recall, at that time there were significant changes in supply costs within the quarter, and also arrangements related to fiscal 2028, so it was important to recalculate. Yes, we considered pricing and related factors, and we have absorbed many costs and included them in those numbers. What we gave in the fourth quarter is the low point.

As you know, that is the low point as we fulfill many customer purchase orders we have already received, and we want to remain solid as we switch to new pricing in the new year. Regarding the changes you mentioned, the arrangements related to fiscal 2028 are consistent with what was communicated last quarter and there is no change at present. We are confident we can achieve these numbers. As for the long-term and future pricing approach, please remember that, with respect to our full roadmap and systems, one of the most important parts is that they are very, very world-class, and everything we achieve in these systems is extremely important.

Again, it drives an extremely important part of the work completed, and I think our pricing has been well thought out, but that is a matter for the future, and we will continue to think about where pricing goes.

Operator: Our next question comes from CJ Muse of Cantor Fitzgerald. Please go ahead.

CJ Muse (Cantor Fitzgerald analyst): Good afternoon, thanks for taking my question. ACIE (AI cloud providers, industrial, enterprise) is a very important growth engine for NVIDIA. I'd like you to talk about your willingness to do vendor financing going forward, and if there is a change, what that means for growth in this important sub-segment, especially for NCP customers?

Colette Kress: Thanks, CJ. We look at the ACIE customer group in parts. One of the most important areas has always included the neoclouds. Neoclouds play a key role in serving enterprises and many countries around the world (especially from a sovereignty standpoint), and we will continue to work with them. Keep in mind there are also other types of customer groups in this category. We invest a lot in delivering for enterprise customers; building their own AI factories is important to them, and neoclouds essentially let them have a private cloud to do that.

We also see many customers in areas such as physical AI advancing related work. Edge computing sometimes also includes data center adjacencies. The types of customers in this category are very diverse. As for what kind of financing may be needed in the future, the part we are most likely to focus on and must evaluate in detail to make sure we get it right is land, power, and shell. The availability of land, power, and shell is an extremely critical part of the whole architecture of building these data centers, and they may need help.

This is a high-quality asset for anyone to hold, so we will list it as one of our top priorities. As stated in today's announcement, we have been talking with a number of financing funds that can play a big role in the work with neoclouds and even enterprise customers. Many of the companies we work with are not only in enterprise or large-scale businesses; keep in mind they are also investment-grade (IG) customers. IG customers in particular can help mobilize the capital needed.

We can consider assisting them with prepayments, which helps their data center buildouts. For now, we will look very carefully at which types of companies we are willing to help with financing, because a lot has changed over the past quarter, and we know this is our priority. We already have some very strong existing customers and will continue to work with them.

Operator: Our next question comes from Stacy Rasgon of Bernstein Research. Please go ahead.

Stacy Rasgon (Bernstein analyst): Hi everyone, thanks for taking my question. Colette, I want to follow up on this. In the CFO commentary, I saw a new sentence, a new paragraph, about neoclouds, saying "your commitments only become effective once the neocloud obtains financing and purchases the deployed data center infrastructure products and related compute capacity, with a maximum commitment of..." and saying "certain commitments may be reduced or canceled" (if they can't). My question is, this is new language in the CFO disclosure. Is this a new change in how you do business with neoclouds, or has it always been this way and you're only telling us now? Because it sounds like you are not the ones actually providing the financing. It reads like they have to obtain financing themselves before they can get your backstop and commitments.

Colette Kress: Thanks, Stacy. Let me clarify for everyone what that sentence means. When we look at various guarantees, a lot of land, power, and shell needs to be built and integrated, including construction, raising capital through equity and/or debt to complete the buildings and bring them to a usable state, and also procuring compute, installing compute, and getting it ready. What we are saying in our statement is: our guarantee only begins at that point. If, unfortunately, there are delays after the above work starts or difficulties in construction, our guarantee no longer exists.

Most of our guarantees are effectively tied to the point when "they are ready," especially in scenarios like the arrangements we made for projects such as Portsmouth. It is important to understand this, because it shows that we have indeed thought it through deeply and chosen companies we believe can sustainably raise the needed debt and equity to complete the data center buildout, so that we have full confidence in providing that guarantee. Is this a new approach? No. Have we changed anything? No.

That sentence is simply saying that we want to be clear—it has actually already been communicated in other materials, but because we received some questions about exactly how it works, we wanted to state it very clearly.

Operator: Our next question comes from Ben Reitzes of Melius Research. Please go ahead.

Ben Reitzes (Melius Research analyst): Hi, thank you very much, great presentation tonight, Colette and team. I want to ask, what do you think about the view that there will be a significant slowdown in 2028 due to a change in government and possible policies? You surely know there is huge concern in the market about a slowdown after a strong calendar 2027. Are you seeing things that give you more confidence? After the midterm "blue wave," have you spoken with people in the administration (not the administration, the political world), and is there anything you can share that gives you more confidence beyond 2027?

Colette Kress: Thanks, Ben. Where we are right now, demand is very large, and the largest companies are all pushing forward with a lot of agentic solutions and new models rolling out, which is exactly what we are focused on supporting. You could say that just last week they were still constantly asking how they could get more supply to meet the demand in front of them. Everyone is preparing for this.

I know data centers in the U.S. are also a big focus: are they fully prepared to take care of the surrounding communities? I think it is very important for all companies to work together so communities understand the jobs that will be added, and so they feel confident about how they will be viewed and how they will operate in these areas. That is something to watch, but I think there are already a large number of data centers in use or ready, and that may well be the case. We will always stay in communication with the government to make sure we are aligned in our view of the future. But I think that, whether from 2027 or extending into 2028, demand driven by agents will continue, and we will carry forward the amount of demand in front of us.

Host: The next question comes from Joshua Buchalter of TD Cowen. Please go ahead.

Joshua Buchalter (TD Cowen analyst): Hi everyone, thanks for taking my question. First, I wish you and the employees all the best through this transition. In the prepared remarks, you made clear that demand still significantly exceeds supply, but should we expect any change in how Nvidia allocates capacity among key end-demand segments, such as hyperscalers, AI labs, and neoclouds? You mentioned AI labs may account for about 15% of revenue this year, which I think is roughly in line with what you disclosed in the summer. But, for example, how large could this become over time?

Colette Kress: Thanks for the question. On allocation, we do consider some important factors that must be clarified to make sure that, in the quarter in which we need to deliver, they can use it and use it very efficiently in the initial phase. That means whether land, power, and buildings are already structured, ready, and completed—this is one of the most important things. The second point is to consider their funding needs; the funding must be in place and ready before we allocate on that basis.

Many of our customers outside IG will want to provide and plan their offtake, because we need to understand whether they have a sufficiently solid offtake volume to support operations. Our allocation principle is "we serve everyone," especially when we already have purchase orders in hand and believe the relevant funding can be in place. We will continue to allocate as quickly as possible based on the supply we expect to be available. I believe this process will continue to expand as more supply comes online, and we will continue to make more allocations. But for now, that is probably the best description of our entire process.

This has nothing to do with any pricing differences; everyone gets the same pricing, which is extremely important to us. We just need to confirm that they are ready on their side to receive compute.

Host: The next question comes from Jim Schneider of Goldman Sachs. Please go ahead.

Jim Schneider (Goldman Sachs analyst): Good afternoon, thanks for taking my question. I'd like you to talk about capital allocation and the overall capital priorities. It sounds like there is no change to your existing commitments to partners, supply chain partners, or customers, and you reiterated previous comments on buybacks. Could you clarify whether the minimum return threshold or criteria for customer financing has changed, or remains unchanged? And could you also talk about the timing or relative priority of buybacks versus customer financing or other partner financing?

Colette Kress: Okay, let me be clear with you about our position on returning capital to shareholders, which is one of our highest priorities—let me make that clear first. We have a $150 billion increase in authorization; it will be used and is expected to be fully used through fiscal 2028. That is what we want to do, and we believe we have the ability to do it. Likewise, there are also some important uses of capital to keep in mind. The first is our suppliers. Suppliers are doing everything they can to expand supply for us, and part of that supply needs capital assistance, including various prepayments, to help them expand capacity.

We want to make sure we are by their side and help them move forward as quickly as possible. I think this is a very good use of our capital—supporting that part of the chain. In addition, there are some expected committed investments that we need to complete; they are not large in proportion, but they must be done. Then, considering that a group of very strong customers may need assistance with some form of guaranteed financing, we will do that kind of thing as well. We will deeply evaluate the company itself, the strength of its plans, and everything its data center requires.

There will be a portion that we do, but it also requires a lot of work to determine whether these are the right investments we should make. Overall, I think we absolutely have the capacity to continue advancing all stock buyback plans and complete them before the end of fiscal 2028.

Host: There are no more questions at this time. Colette Kress, I will hand the call back to you.

Colette Kress: Thank you. Thank you all for joining today. Our momentum is real, demand is accelerating, supply is growing, our customers and partners are with us, and our employees are fully committed. We can get it done. Thank you very much.

Host: That concludes today's conference call. You may now disconnect.

Eastmoney Infographic · Key Takeaways

(Source: Cailian Press) Mid-article body Embedded advertisement

Source

East Money