HPE (HPE.US) Q3 conference call: Changxie locks in production capacity to extend the boom cycle to 2027, and the growth rate of online orders has reached 3.5 times revenue and is accelerating transformation

The Zhitong Finance App learned that HPE (HPE.US) handed over record reports in the third quarter of fiscal year 2026: revenue, gross profit margin, and EPS all exceeded expectations. Management sent a strong signal of optimism. CEO Antonio Neri pointed out that AI demand is moving from pilot to enterprise production and deployment. Agentic AI and inference workloads have become the core driving force, and enterprise customers have reached an inflection point in demand; thanks to Juniper's integration exceeding expectations, the order growth rate is 3.5 times higher than revenue, and AI network orders have reached a cumulative total of 2.2 billion US dollars and the annual target has been raised. Despite continued supply constraints, the company locked in production capacity through multi-year agreements and raised the 2026 and 2027 fiscal year guidelines sharply based on strong backlogs and pipeline visibility. Revenue is expected to increase by 13%-17% in 2027, and EPS to reach $4.40-4.60. Management is convinced that the company is at the beginning of multi-engine structured growth.

During the FY2026 third quarter results call, HPE management sent a strong signal to the market that the company was at an inflection point in structural growth with an unusually optimistic tone. CEO Antonio Neri and Chief Financial Officer Mary Myers have jointly drawn a growth blueprint where AI demand is fully blossoming, network business is poised, and supply constraints are gradually improving. Based on strong order backlogs and pipeline visibility, the 2026 and 2027 fiscal year performance guidelines were drastically raised.

The inside clearly stated that AI has evolved from early proof-of-concept training deployments to a wider range of enterprise workflow transformation opportunities. Customers are investing heavily in Agentic AI applications and AI inference, which require accelerated computing infrastructure, secure data storage, and enterprise-grade cloud management. He emphasized that enterprise customers have become the main force in this round of growth — AI-related programs have received higher investment than traditional IT projects, and board-level participation has increased significantly. Practical experience with more than 1,200 AI use cases within the company (more than 300 of which have already been put into production) enables HPE to better serve customers. Neri said, “The company has reached an inflection point, and the corporate market will be even stronger in 2027.”

Inside, the network business is positioned as the core of the company's strategy. Juniper's acquisition integration is ahead of schedule, and cost synergy is progressing on track. Network orders increased 36% this quarter, 3.5 times the revenue growth rate. Cumulative AI network orders reached US$2.2 billion, exceeding the annual target, so the company raised its year-end target to US$25-3 billion. Neri placed special emphasis on the gigawatt-scale AI infrastructure cooperation agreement signed with Oracle and the upcoming AMD Helios platform opportunities — the latter has not yet been included in the 2027 guidelines, but is viewed as a “multi-billion dollar” potential increase. He confidently stated, “We are becoming a networking company, and the Helios opportunity will be an important growth engine for us in the future.”

Supply bottlenecks are still a major factor limiting revenue conversion, particularly in the production capacity of components and wafers such as DDR5 and NAND. However, Neri said that the company has locked in production capacity by signing a multi-year long-term supply agreement, and the allocation can be adjusted flexibly every quarter. Marie added that procurement commitments have reached an all-time high. Both believe supply constraints will continue until 2027, but HPE's pricing discipline and customer relationship management have enabled them to remain profitable in an environment of rising costs.

Third-quarter revenue of $12.2 billion, +34% year over year, gross profit margin of 40%, operating profit of $2 billion, and EPS of $1.11, all set record records. Based on strong Q3 results and several major deals signed after the end of the quarter (including a multi-billion dollar hyperscale inference server order), the company raised the FY2026 EPS guidelines to $3.75-3.85, with free cash flow of at least $3.75 billion; the fiscal year 2027 revenue growth guidance was raised to 13%-17%, EPS 4.40-4.60 US dollars, and free cash flow of at least $5 billion. Marie stressed that the drastic reduction in restructuring expenses in 2027, combined with the lower demand for working capital in the online business, will significantly increase the cash conversion rate.

Neri concluded, “Our strategy is very clear—with the network as the lead, AI as the engine of growth, and the entire product portfolio serves customer needs and drives profit and cash.” He dismissed market concerns about the sustainability of demand, pointing out that customers “wait is not an option” and that AI investment is an incremental budget rather than a replacement. Marie stressed that the company has achieved its leverage target more than a year ahead of schedule and is entering the 2027 fiscal year in a strong position. Overall, the signal from management is that HPE is in a multi-engine-driven structural upward cycle, and supply constraints will only prolong the boom, rather than reverse the trend.

Below is a transcript of the HPE FY2026 third quarter results conference call:

Executive speeches

Shannon Cross

Senior Vice President, Chief Strategy Officer and Head of Investor Relations

Good afternoon, I'm Shannon Cross, Chief Strategy Officer at HPE. Welcome to the Q3 FY2026 results conference call, attended by President and CEO Antonio Neri and Chief Financial Officer Mary Myers.

Before handing it over to Antonio, allow me to remind you that this conference call is being webcast and can be replayed after the meeting. Press releases and presentations have been posted on the Investor Relations website. Some of the financial information mentioned in this conference is a forward-looking statement. Based on our best judgment on business and external factors, the company is under no obligation to update it. Please note that the financial data discussed are preliminary estimates and may differ from the final 10-Q quarterly report submitted. For ease of discussion, oral data has been compiled. Please refer to financial reports for detailed data. Non-GAAP financial indicators have been reconciled with GAAP measures. For details, please see the Investor Relations website.

Unless otherwise specified, all revenue growth rates are year-on-year, all financial indicators and growth rates are non-GAAP, and earnings per share refer to diluted non-GAAP net profit. Some of the data has been normalized to include Juniper Networks' results in HPE's consolidated caliber since the beginning of fiscal year 2025.

Antonio and Marie will make a statement by referring to the main points in the presentation. We also disclose records of some financial metrics; see endnotes to the paper. Now I have to ask Antonio.

Antonio Neri

Chief Executive Officer, President and Director

Thanks Shannon and good afternoon everyone. Our strategy was once again validated this quarter. We delivered another record financial performance, demonstrating the sustainability of profit growth and the strict execution of the entire company. We have fully exceeded the company's financial commitments and achieved record highs in revenue, gross margin, non-GAAP operating profit, and earnings per share. AI has been a growth driver for many years, with demand across HPE's entire product portfolio. Customer demand accelerated this quarter, and orders in both business segments grew faster than revenue. We recorded the highest number of orders in a single quarter in the company's history, and the backlog of orders also set a record. Supply bottlenecks are still limiting our ability to meet strong demand. We are working closely with our partners to secure more multi-year supply agreements, provide customers with alternative product configurations, and strengthen planning coordination to more accurately forecast supply.

In the third fiscal quarter, HPE achieved revenue of US$12.2 billion, a significant increase of 34% over the previous year. The revenue growth rate so far this fiscal year is about double that of the same period last year. Non-GAAP gross margin reached a record 40%. Non-GAAP operating profit was $2 billion, 2.5 times that of the same period last year. Non-GAAP earnings per share were $1.11, another record, and surpassed $1 for the first time in a single quarter. Excellent operating results directly translated into stronger cash generation capacity. Free cash flow in the third quarter reached US$958 million, the highest in the same period in history.

Last quarter, we raised our FY2026 guidance and presented the FY2027 growth framework for the first time. Based on record performance, orders and a backlog of orders, we have further raised our outlook for the 2026 and 2027 fiscal years, which Marie will explain in detail later.

Before handing it over to her, I'd like to discuss the market situation and business segment performance, and explain the important milestones in the Juniper Networks acquisition case. In August, the US Federal Court approved our settlement with the Department of Justice and determined that it was “in the public interest.” We are satisfied with this result, which strengthens our confidence in the long-term value of integrating our two major networking product portfolios. One year after the completion of the acquisition, the integration plan and cost synergy have progressed beyond expectations. Business performance continues to strengthen, and innovation and execution capabilities continue to improve. Our stronger competitiveness has translated into providing customers with more innovative network solutions and higher profitable growth for shareholders.

Despite supply restrictions affecting the transformation of demand to revenue, orders and revenue for network products and services have reached record levels. The park and branch sector achieved record revenue as customers upgraded old edge infrastructure and deployed AI-driven network operation and maintenance. Orders are ahead of revenue, reflecting the differentiated advantages of our autonomous driving network and the flexibility of our multi-cloud deployment model. Demand for routing and data center switching accelerated this quarter, with orders leading the way in revenue, and the backlog of orders reached a record high. The backlog growth indicates that hyperscale customers and new cloud service providers are in strong demand for our routers, switches, and AI operation and maintenance software, and they continue to increase their investment in AI cloud capital expenses.

The growth in the backlog indicates that customer demand has exceeded current supply. We expect to turn more orders into revenue in the fourth quarter, which further strengthens our confidence in the continued growth of the network business in FY2027. Today, we announced the expansion of cooperation with Oracle to jointly accelerate the construction of gigawatt-scale AI infrastructure. Oracle will deploy HPE Juniper network routers and switches for one of its largest AI cloud infrastructure deployments. With the industry's most comprehensive AI networking product portfolio, HPE has unique support capabilities.

SASE and security businesses also contributed to growth. The strong performance of firewalls, SD-Branch, and branch SRX reflects customer demand for integrated network and security solutions. Welcome to the Internet Business Investor Day later this month, where we will share more confidence in online strategy, innovation, and business momentum.

The network business continued to strengthen through careful integration, while the cloud and AI business performed particularly well in an environment where supply was severely limited, achieving record revenue, operating profit, and operating profit margins. AI is shifting from early proof-of-concept training deployments to broader enterprise workflow transformation opportunities. Customers are increasingly investing in new Agentic AI applications and AI inference, which require accelerated computing infrastructure, secure data storage access, and enterprise-grade cloud management. Our comprehensive cloud and AI product portfolio fits perfectly at this inflection point in the market. Demand for traditional servers, AI systems, storage, private cloud solutions, and GreenLake cloud services continues to be strong.

The server product category is outstanding. Traditional servers and AI systems are in strong demand, and demand from large enterprises, new cloud service providers and sovereign customers is strong.

We expect demand to remain extremely high as the pipeline size far exceeds the backlog of orders. The server business is undergoing a fundamental transformation — the way customers value IT infrastructure is changing. They are not only concerned about whether servers can run AI workloads, but also how they can enable new business workflows through new AI applications. We have seen that AI-related enterprise plans to receive higher investment than traditional IT projects, and that board of directors and other senior management participate more in investment decisions. They are promoting AI technology to unleash greater potential for business transformation.

The storage business performed well this quarter, with record revenue, benefiting from our focus on our own IP products and providing a modern multi-protocol data platform for the AI era. Customers are beginning to evaluate where AI workloads run most efficiently — balancing the best security management with the lowest cost per token. While still modernizing data storage environments, organizations are preparing for the next generation of AI-driven workloads and applications, hoping to keep data close to AI infrastructure and control sensitive information. These trends are driving strong demand for our Alletra MP storage solutions. We believe a comprehensive data storage value proposition will continue to accelerate the momentum of the storage business.

Our private cloud AI platform allows customers to manage enterprise AI applications and AI agents while maintaining control over data, governance, security, and operations.

Enterprise customers are in strong demand for PCAI, and they want to optimize the token economy for large-scale local AI deployments, which has rapidly expanded the order and customer base. GreenLake remains one of the biggest differentiators, as customers can manage infrastructure and software through a secure hybrid cloud operating model, no matter where traditional or new AI workloads are located. Customers are expanding the use of GreenLake cloud services, expanding existing usage by consuming new software and intelligent cloud services, and boosting net retention rates.

In the third quarter, the number of GreenLake customers reached 52,000, an increase of 18% over the previous year, compared to 44,000 in the same period last year. HPE Financial Services continues to deepen customer relationships and provide significant competitive advantages, particularly as more customers seek financing options to support AI investments. This quarter's financing amount, residual value, and return on equity all hit third-quarter records.

In summary, HPE once again delivered outstanding quarterly results, which fully exceeded the company's promises and demonstrated continued differentiation in strategy and execution. Looking ahead, the same growth drivers give us confidence that we will continue to achieve higher profitable growth, stronger cash flow generation, and higher return on shareholder capital. Thank you to all of our employees for their dedication and excellent execution.

Now I have to ask Marie.

Mary Myers

Executive Vice President and Chief Financial Officer

Thanks Antonio and good afternoon everyone. We achieved another strong quarter, reflecting the acceleration of AI demand, the steady momentum of the network business, and strict execution throughout the company. The demand environment remains strong, and orders continue to grow faster than revenue. AI infrastructure investment is rapidly increasing, and enterprise spending is focused on Agentic AI workloads and AI inference.

Importantly, this opportunity is expanding across use cases, customer vertices, and geographies, further strengthening the value of HPE's expanded product portfolio and our ability to provide customers with integrated solutions across enterprise technology stacks. We continue to focus on executing in a dynamic supply environment, managing product portfolios and investment costs while increasing operating leverage. This discipline was reflected in financial performance, supporting sustained profit growth in the 2026 and 2027 fiscal years.

The financial results are described below. Revenue was US$12.2 billion, up 34% year over year, exceeding the guideline limit, and orders increased by 42% (normalized caliber), driven by traditional server, AI systems, and network demand. Gross margin surpassed 40%, thanks to traditional server pricing discipline and improved networking product portfolio. Looking ahead, we expect gross margin to return to a more historic level, partly offset by an increase in the share of network products due to the growth of AI systems and the normalization of traditional servers.

Operating expenses increased 17% month-on-month, reflecting higher variable compensation brought about by record performance. Operating expenses are expected to decline in FY2027 as variable compensation is normalized, and Catalyst's transformation efficiency and Juniper's synergy continue to show. Operating profit was US$2 billion, up nearly 40% month-on-month. The operating margin was 16.2%, an increase of 290 basis points over the previous month, benefiting from gross margin expansion and operating leverage. Earnings per share of $1.11 were well above the guideline limit. GAAP earnings per share were $1.06. Free cash flow for the third quarter of $958 million was supported by strong operating profit and accounts receivable recovery.

By sector, the network business revenue was 2.9 billion US dollars, and normalization increased 10% year over year, in line with expectations. Orders increased 36%, about 3.5 times the revenue growth rate. Orders are growing widely, covering AI infrastructure investments for data center switching and routing, as well as strong demand for campus and branch autonomous driving networks. Demand for AI networks accelerated this quarter, with orders reaching a new high of US$700 million, a year-on-year increase of three digits. Our product portfolio and competitive position in vertical expansion, horizontal expansion, and cross-domain expansion are enhanced by the recent launch of Tomahawk 6-based direct liquid cooled switches and differentiated PTX/MX routing product lines. The pipeline continues to grow and is expected to accelerate further as the Helios platform goes live. We expect AI networks to be an important growth engine for the company. Cumulative AI network orders reached 2.2 billion US dollars, exceeding the 2026 fiscal year target, so the year-end target was raised to 2.5 billion to 3 billion US dollars. To match the increase in orders, we have more than doubled our online procurement commitments month-on-month.

The normalization of park and branch revenue increased 8%. Routing revenue accelerated to 23%, benefiting from increased demand for AI network entry and exit infrastructure. Security increased by 12%. Data center network revenue declined 6% due to supply restrictions impacting delivery times.

Order momentum is stronger across product categories, with high double-digit growth in data center switching and routing, and low double-digit growth in campuses and branches. We continue to focus on improving order conversion to drive faster revenue growth and scale effects. In the customer vertical, corporate revenue grew 12% and service providers grew 5% (normalized). Enterprise growth is being driven by large global customers who prioritize network modernization between campus and branch and data center exchanges.

The profit margin for network operations was 22%, in line with the guidelines, reflecting strict execution and the initial realization of Juniper's synergy effects, partially offset by higher variable compensation.

Switch to the cloud and AI sector. Revenue for the third fiscal quarter was US$9 billion, up 25% year over year, exceeding expectations, reflecting the strength of traditional servers. The increase in ASP drove server revenue to a record high. Pricing discipline and scaling up drove operating profits of over $1.5 billion. The growth rate of operating profit accelerated, increasing 61% month-on-month, and three-digit year-on-year. The operating profit margin was 17%, an increase of 460 basis points over the previous month, demonstrating the ability to scale up profitability.

Server revenue grew 35%, and accelerated month-on-month. Strong growth in traditional server ASP offset supply-limited sales. Orders grew strongly by double digits year over year, from large enterprises, sovereign customers, and cloud service providers. Our supplier agreements have been extended for many years, partly ensuring the capacity allocation we needed to shorten delivery cycles, improve backlog conversion, and drive new order growth, and procurement commitments are at an all-time high.

We're seeing companies moving from AI pilots to production deployments, using traditional servers to host Agentic AI and inference workloads. For example, a global financial services institution uses AI for market analysis and transaction insight, while another large retail client deployed local Agentic AI workloads to reduce public cloud AI token costs. As evidence of strong growth, we are pleased to report that after the end of the quarter, HPE received a multi-billion dollar server order from a hyperscale customer dedicated to inference, which supports our judgment that inference and agentic AI demand is accumulating.

Orders for AI systems reached 2.4 billion US dollars, an increase of more than 30% over the previous month, and customers were widely distributed. Business demand has more than doubled as AI initiatives have become a board priority. The backlog of AI systems increased 14% month-on-month to a new high, and the pipeline size was several times larger than the backlog. AI systems generated nearly $1.6 billion in revenue for the quarter. AI system revenue is expected to improve month-on-month in the fourth quarter, depending on when the backlog is converted.

Storage revenue increased 10%, driven by strong orders, increased ASP, and a combination optimization leaning towards high-value private IP and private clouds. PCAI orders grew by three digits this quarter, and customers used our AI factory platform to support Agentic AI and inference plans. Alletra MP orders and revenue grew strongly in double digits year over year. We're seeing huge growth potential for the X10000 object and file system, expanding the AI solution portfolio to address the rapidly expanding unstructured data market.

Revenue from the financial services business remained roughly flat year over year, and the return on equity continued to exceed 20%.

About integration and transformation plans. We are making good progress in building a more efficient company, and several projects to reduce sales costs and operating expenses have progressed beyond expectations. Juniper's synergy is progressing according to plan. The goal is to achieve an annualized operating rate savings of 600 million US dollars by the end of fiscal year 2028, and the integration cost is better than expected. We highlighted the contribution of AI process simplification within Catalyst last quarter, and further expanded AI and operational simplification efforts this quarter. HPE is now deploying an in-house Agentic AI platform that uses intelligent routing to send each workload request to the most cost-effective AI model based on its own private cloud AI, open source, and open weighting models. According to internal analysis, our PCAI solution can reduce token costs by up to 60% compared to the public cloud. Routine tasks remain local, and cutting-edge models are used only for the most complex jobs.

In terms of cash flow, operating cash flow was US$1.6 billion, and free cash flow was US$958 million in the third quarter. As a result, we raised our FY2026 free cash flow target to at least $3.75 billion. The cash conversion cycle improved by 1 day, mainly due to a decrease in the number of days due (benefiting from more favorable quarterly bill linearity and stronger recycling), offset by an increase in inventory days (increased procurement for expected delivery). Inventory at the end of the quarter was US$11.8 billion, up year over year, reflecting higher product costs and targeted procurement to support increased orders and backlog.

In the third quarter, we returned $324 million to our common shareholders, including $189 million in dividends and $135 million in share repurchases. After completing the H3C transaction, gross proceeds of approximately $1.4 billion were received, and term loans were repaid in cash. The net leverage ratio at the end of the quarter was 1.8 times, below the target of 2 times. The sale of the telecommunications solutions business was completed last month, and it is planned to repay $1.25 billion in maturing notes this month. We plan to return at least 75% of free cash flow to shareholders in the fourth quarter.

About the guidelines. Based on strong results in the third quarter and confidence in the sustainability of demand, we raised our outlook. Fourth-quarter revenue is expected to be between US$13.9 billion and US$14.8 billion, reflecting continued strong demand from the two major sectors. Network revenue is expected to grow by 11%-13%. Thanks to strong orders and improved supply chain transformation, operating profit margins have increased moderately from month to month, benefiting from revenue growth and Juniper's synergy effects.

In the cloud and AI sector, revenue is expected to grow by 60%-72%, reflecting continued demand, rising traditional server ASP, and higher AI revenue conversion. Operating profit margins are expected to fall to the middle double digits month-on-month. Total operating expenses fell by a single digit month-on-month in the fourth quarter due to reduced variable compensation and increased synergy between Catalyst's transformation and Juniper. The overall operating profit margin declined month-on-month, mainly due to the increase in the share of AI systems in the cloud and AI sectors and pricing factors. Earnings per share are expected to be $1.20 to $1.30, and GAAP earnings per share are $1.12 to $1.22. Based on our third-quarter results and fourth-quarter outlook, we raised our FY2026 earnings guidance range to $3.75 to $3.85 and GAAP EPS to $2.93 to $3.03. Free cash flow for the 2026 fiscal year is at least $3.75 billion.

Given the intensity of demand and considerable backlog at the end of the third quarter, as well as some large orders signed after the end of the quarter, we updated our FY2027 framework: expected consolidated revenue growth of 13%-17%; network revenue growth of 14%-17%; cloud and AI revenue growth of 14%-18%; company operating profit growth of 14%-18%; and operating profit margin of 14%-15%, benefiting from a slight decrease in operating expenses. The profit margin for network operations is high in the 20% range; the profit margin for cloud and AI operations is about 13%. Earnings per share were $4.40 to $4.60, up 16%-20% from the midpoint of the 2026 EPS guidance; free cash flow of at least $5 billion. Importantly, this framework is based on higher guidance for FY2026 and points to significant improvements in FY2027.

In summary, the third quarter was an excellent one for HPE. We achieved strong financial results, raised our 2026 and 2027 fiscal year commitments, and reached our leverage target more than a year ahead of schedule. Demand is still faster than revenue, a record backlog of orders, and Juniper integration and Catalyst plans are progressing beyond expectations. Entering the final quarter of fiscal year 2026 and looking ahead to 2027, we are executing in a strong position, with enduring demand, strong profits, and operating discipline to maintain both.

Q&A session

Katherine Murphy (Goldman Sachs):

We're excited to see full product portfolio momentum, record orders, and revenue growth guidance of 13%-17% for FY2027. First, what is the basis for current demand representing an ongoing infrastructure cycle rather than a customer's upfront spending? Second, in the revised FY2027 guidance, how much comes from new hyperscale reasoning and Oracle transactions, and how much from improvements in remaining business?

ANTONIO NERI:

Thanks Katherine. The guidelines are based on what the market sees, and the market tells us that demand is still unusually strong. AI clouds continue to be built on a large scale, and we are participating in a rigorous manner. On the network side, demand for routing and data center switches is significant. AI network orders reached a record of US$700 million this quarter, and are expected to reach US$25-3 billion for the whole year. The 2027 guidance does not include the AMD Helios opportunity, which will gradually be launched from the end of this year to 2027, so demand is extremely strong. Confidence in cloud and AI comes from enterprise acceleration — companies have clearly reached an inflection point, driven by the deployment of Agentic AI and AI inference, with a huge number of use cases. We have more than 1,200 use cases within our company, more than 300 have been put into production, and we continue to learn faster. We've seen a wide range of enterprise market trends across multiple vertical industries. This benefits our traditional server, storage, and private cloud stacks because customers don't need a large number of GPUs or CPUs, but rather a compact capital infrastructure to optimize token costs. The number of local tokens is growing very fast. The pipeline and transformation prospects underpin our confidence in the 2027 guidance.

MARY MYERS:

Regarding the guidelines, the company's total revenue increased by 13%-17%. Oracle transactions and core business in the network sector are included in 14%-17% network growth. Hyperscale transactions are part of the 14%-18% growth in cloud and AI. AI revenue will increase in cloud and AI in 2027.

Amit Darianani (Evercore):

Congratulations on the excellent results. Antonio, I want to ask about the online business. The 10% organic growth seems to be slightly lower than peers and expectations, but the 36% increase in orders is very strong. Please talk about the reason for the gap? When will revenue catch up with orders? Also, please explain more about the Oracle announcement, what products do you offer?

ANTONIO NERI:

The order momentum is extremely strong. The order growth rate is 3.5 times that of revenue. The limiting factor is supply. Supply is expected to gradually match orders, with revenue growth rising from 10% to 11%-13% in the fourth quarter to 14%-17% in 2027. We have redoubled our procurement commitments and are working closely with our suppliers. Orders will continue to lead revenue, but the gap is expected to close. Core parks and branches account for about 50%, low double-digit growth in orders, record revenue, and are expected to continue to improve. On the Oracle side, this is an expansion of Juniper's original business, but the scale is gigawatt-scale. The global deployment uses our QFX switching products (based on Broadcom Tomahawk 6) and software and AI Ops, which is a horizontal expansion; the cross-domain extension uses the PTX routing platform (based on self-developed Express 5 silicon), which is an important differentiation in scale. Taken together, this is a multi-year multi-gigawatt deal.

Aaron Rex (Wells Fargo):

Congratulations on the strong results. In terms of traditional servers, growth is mainly driven by pricing. How is the unit growth? When AI enters the enterprise environment, what are the installation basics and upgrade opportunities? Also, are hyperscale transactions changing strategies, and are you more active in seeking more hyperscale opportunities?

ANTONIO NERI:

Regarding the latter, this hyperscale customer is actually used as an enterprise for internal reasoning, rather than selling traditional Tier 1 cloud infrastructure. The strategy has not changed. Regarding units, units will grow as supply improves, and we expect units to strengthen in the fourth quarter. Traditional servers and storage are accelerated, and private clouds are also benefiting. Localizing AI deployments will drive unit growth, but supply will still limit 2027. The guidelines reflect the balance between supply and demand.

Joseph Cardoso (J.P. Morgan Chase):

Antonio, why did you mention Helios opportunities are not included in the 2027 guidelines? If included in the future, how much room will there be for network revenue and profit margins to increase?

ANTONIO NERI:

We are working closely with AMD, and the infrastructure is expected to be orderable later this year. The pipeline is large, mainly for large-scale AI training and inference, but the customer is concentrated. We would like to see more confirmed schedules before incorporating them. The opportunities are huge, and the market could reach tens of billions of dollars. This time, the HPE Juniper vertical expansion switch is not only embedded in the HPE Helios rack, but can also be sold separately. Currently not included in 14%-17% network growth.

MARY MYERS:

We will provide more comprehensive guidance for 2027 at the end of the fourth quarter.

Wamsey Mohan (Bank of America):

Can you share more details about the $3.5 billion speculative deal? Including content and economics? Is the strategy for hyperscale opportunities changing?

ANTONIO NERI:

The customer is inconvenient to disclose, but it is a hyperscale customer, but it is used as an enterprise, for internal reasoning, and is not a traditional Tier 1 infrastructure.

Asia Magant (Citi):

Can you talk about where supply bottlenecks are most severe, and expectations for supply mitigation and long-term agreements?

ANTONIO NERI:

Supply bottlenecks are similar to before, such as DDR5, NAND, etc., and have continued for three quarters. Other components are limited by wafer production capacity. Some improvements are expected in 2027 due to partner investment in cleanroom production capacity, but structural solutions require wafer production capacity. In terms of memory, traditional DRAM is shifting to HBM, and demand for HBM is extremely high. Traditional servers use DRAM instead of HBM, so we're focusing on that. We have signed LTA for many years to lock in production capacity and can be used flexibly. Supply constraints will continue for a long time, affecting costs and pricing, but our management is effective.

Eric Woodlin (Morgan Stanley):

Marie, you mentioned the normalization of gross margins of traditional servers. Why is there normalization when demand is strong and units improve? Is it being consumed due to low cost inventory?

MARY MYERS:

Third-quarter gross margin was a combination of factors: revenue scale, pricing discipline, Catalyst efficiency, and current transaction mix. Looking ahead to the fourth quarter and 2027, the AI transaction portfolio will increase, and more AI revenue will be shipped (as can be seen from the increase in inventory). The traditional internal server transaction mix will also be normalized. We're happy with the guidelines, but the third quarter was a combination of factors.

Tim Long (Barclays):

Regarding AI networks, can Oracle's victory be a reference case to drive other hyperscale or new cloud customers? Second, are self-developed silicon wafers important to winning, and can they drive more cross-domain expansion?

ANTONIO NERI:

Yes. Oracle proved the scalability and AI capabilities of our products, and pioneered the launch of 1.6T (air cooling and direct liquid cooling). The self-developed silicon chip is an important differentiator, and only a few players can do it. For example, the latest PTX product (about three-quarters of the rack) can carry 60 million people from New York to London to watch Netflix at the same time, which is astonishing. As hundreds of gigawatts of infrastructure are deployed, routing capability is critical.

David Vogt (UBS):

Antonio, the 2027 guide was raised by about $4 billion. How is the supply side improving compared to 90 days ago? Marie, free cash flow increased by about 500 million US dollars. What is the reason for the improvement in conversion rates?

ANTONIO NERI:

On the supply side, production capacity is locked in through multi-year LTA, and the use of production capacity can be adjusted every quarter to support the matching of demand and supply.

MARY MYERS:

Free cash flow was at least $5 billion, up 33% year over year. The main reason for the improvement was that restructuring expenses were drastically reduced in 2027. Most of the restructuring planned by Catalyst and Juniper was completed in 2026, with a small amount remaining in 2027.

ANTONIO NERI:

In addition, the accelerated growth of the network business (14%-17%) requires less working capital due to faster inventory turnover.

Mark Newman (Bernstein):

Server revenue increased by 35%. Excluding AI servers, traditional servers seem to be growing faster? Orders increased 75%, how much of this was from ASP and configuration upgrades, and how much from unit growth? How long will the supply restrictions last?

ANTONIO NERI:

Units are expected to strengthen in the fourth quarter. Units are limited by supply in 2026, and the increase mainly comes from ASP. The future supply is still tight and the backlog is high, but the guidelines have taken supply and demand into account. Cloud and AI grew by 14%-18%, driven by the transformation of traditional servers, AI systems, and storage. Storage is growing twice as fast as revenue.

MARY MYERS:

The 2027 cloud and AI revenue guidelines have been raised to 14%-18%, reflecting strong demand.

Matthew Nickenan (Truist):

Antonio, are customers hesitating due to pricing actions that increase memory costs? Where are incremental IT budgets coming from?

ANTONIO NERI:

I didn't see any hesitation. Early customers were shocked by the rising costs, but understood that AI must accelerate. They are getting smarter, optimizing local token economics, and our internal analysis can save 60% of costs. Budgets are generally rising, and AI investments are incremental rather than replacement. For example, a major financial client decided to build an AI factory locally to improve agility and meet compliance. Businesses will be stronger in 2027, as confidence grows and success stories drive more success.

ANTONIO NERI:

Thank you all for your time. Reminder to Online Investor Day and welcome to attend. We are very excited about the Juniper acquisition, the integration was successful, and the future is huge, including new customers and Helios opportunities. We are being led by the Internet to become a network company. The rest of the portfolio serves customer needs and drives profit and cash. Thank you.