Hewlett Packard Enterprise (HPE) recently released a mixed bag of earnings, beating non-GAAP earnings and free cash flow expectations, while falling short on revenue. While the surface-level results might appear unremarkable, CEO Antonio Neri's comments during the earnings call hint at a potential paradigm shift in how enterprises approach AI deployment. Could this be the first sign of a major wave of repatriation, with businesses bringing their AI workloads back in-house?

Neri's confidence stems from the explosive demand for AI solutions, particularly accelerated compute infrastructure. HPE's cumulative orders for APU-equipped servers (including GPUs) have reached a staggering $4 billion, nearly 25% of total server orders since the start of 2023. Although much of this demand currently originates from hyperscalers and model builders, there's a growing realization among enterprises that they require a private, secure, and sustainable environment for fine-tuning AI models and running inference workloads.

The Repatriation Hypothesis: A Closer Look

Neri's use of the word "repatriation" suggests that enterprises might be reconsidering their reliance on public cloud for AI, driven by concerns over data control, security, and escalating cloud costs. HPE's observation of a 100% increase in non-Tier 1 customer orders for GPU-equipped servers within its traditional compute segment supports this hypothesis. This surge in demand for in-house AI capabilities indicates a potential shift in enterprise strategy.

HPE's Strategic Positioning for AI-Driven Transformation

HPE's portfolio, encompassing edge computing, hybrid cloud solutions, and its pending acquisition of Juniper Networks, positions it strategically for this AI-driven transformation. Neri emphasized the crucial role of networking in supporting AI workloads, stating that "AI will drive the need for more ports...more networking bandwidth." Juniper's expertise in networking, combined with HPE's edge and hybrid cloud capabilities, could create a powerful ecosystem for on-premise AI deployment.

Growth in AI Inference: A Key Indicator

"HPE anticipates that AI inferencing, the process of using trained AI models to make predictions, will "accelerate over time." This shift, propelled by the need for real-time data processing and business transformation, could be the catalyst for a wave of repatriation."

Cumulative APU Orders and Revenue

*Estimated based on transcript data

The Challenge: Revitalizing the Core Server Business

However, there's a challenge. HPE's core server business, which relies heavily on traditional CPU-driven systems, has been grappling with customer digestion and pricing pressure. The question is whether the surge in AI demand and potential repatriation will simply offset the ongoing weakness in traditional servers or spark a broader revitalization of the compute segment.

Sequential Improvement in Traditional Server Demand

HPE has reported sequential improvements in unit demand for traditional servers over the past two quarters. While it's still early to declare a full recovery, it suggests that demand is stabilizing, potentially setting the stage for a future uptick.

HPE's Differentiators in the Enterprise AI Market

HPE's success in capturing the emerging enterprise AI market hinges on leveraging its unique strengths. The company's expertise in supercomputing and its commitment to providing a hybrid cloud experience through its GreenLake platform could prove decisive. Neri highlighted HPE's ability to offer "supercomputing as a public cloud instance" and the increasing demand for tuning AI models in a "private, secure, and sustainable way." These differentiators, combined with the company's focus on specialized sales resources and new AI-optimized solutions, could unlock significant growth opportunities in the coming quarters.

Conclusion

HPE's earnings call may have revealed more than just quarterly results. It could be a glimpse of a seismic shift in enterprise IT, driven by the growing realization that AI is not just for cloud giants, but for everyone. If the repatriation hypothesis proves true, HPE, with its diverse portfolio and strategic positioning, could be at the epicenter of this exciting new wave.

"Fun Fact: HPE is one of the few companies to have built and operated a supercomputer that can perform one quadrillion calculations per second (a petaflop). This incredible computing power, housed in a massive data center in Wyoming, is used to solve some of the world's most challenging scientific and engineering problems."