Penguin Solutions, Inc. earnings

Management highlights

### Market Environment Shift - AI adoption is transitioning from early prompt-and-response experimentation to production-scale inference and agentic AI, where persistent, context-rich, task-oriented AI agents act as operators rather than just advisors. - This shift expands infrastructure demand across the full AI data center stack: demand grows not just for GPUs and GPU-attached high bandwidth memory, but also for general purpose compute, additional memory, storage, and networking, as memory has become a primary bottleneck for large-context AI inference performance. ### Strategic Positioning: AI Factory Platform - Penguin Solutions is positioned as a leading full-stack AI factory platform company at the intersection of memory and AI infrastructure, combining five core elements: Clusterware AI operating software, memory AI and integrated memory solutions, Compute AI advanced systems, AI factory reference architectures, and end-to-end design/build/deploy/managed services. - The integrated platform compresses customer time to production and time to revenue, delivering superior ROI and token economics for inference workloads, which addresses key customer pain points for production-scale AI. ### Customer and Partner Momentum - The company follows a land-and-expand go-to-market strategy: in the last four trailing quarters, 13 new AI infrastructure logos were added, 7 of which have already expanded their business; 16 new memory logos were added, 5 of which have expanded engagements. - Key Q3 wins and expansions include an expanded engagement with Deepgram, an expanded engagement with a tier-1 financial institution, a new win with a leading quantitative trading firm, and the completion of the Spectra sovereign supercomputer deployment in partnership with Sandia National Laboratories and Next Silicon. - Penguin received industry recognition as an NVIDIA AI Factory Specialized Partner and 2026 Dell Technologies Global Alliances Americas AI Partner of the Year. ### Product Innovation and Operational Execution - Memory AI product line: CXL-based memory AI KV cache servers deliver up to 2x higher inference performance, up to 8x lower time-to-first-token latency, and 4-5x more cost-effective memory expansion than GPU-HBM, gaining strong traction with enterprise customers. The photonic memory appliance (PMA) is in development via partnership with Celestial AI (now part of Marvell). - Clusterware AI: a hardware-agnostic unified control plane for heterogeneous AI infrastructure, with a newly introduced AI factory operations agent that adds a natural language conversational interface to automate operations and improve administrator productivity, opening new software revenue opportunities. - Origin AI: validated reference architectures that reduce deployment risk and time to revenue, with newly launched Origin AI inference solutions leveraging the company's 4 billion hours of GPU runtime experience. - Leadership transition: CFO Nate Olmstead will step down July 8; Aaron Johnson, VP of Finance and Accounting, will serve as interim CFO starting July 9, with a formal search for a permanent CFO underway. No changes to operating priorities, financial discipline, or execution focus resulting from the transition. - Financial discipline: the company balances growth investments with working capital and supply chain management, and is applying agentic AI internally to accelerate product development, improve time-to-market, and increase supply chain scalability.

Guidance

- For full-year fiscal 2026, management significantly raised guidance from the prior outlook: net sales are now projected to grow 22% year-over-year (up from 12% prior), and non-GAAP diluted EPS is now expected to hit $2.60 (up from $2.15 prior). - Segment-level FY26 outlook updates: Advanced Computing net sales are now expected to decline 15% to 20% year-over-year (improved from prior guidance); Integrated Memory net sales are expected to grow 90% to 95% year-over-year (up from prior guidance); Optimized LED net sales are now expected to decline approximately 5% year-over-year (improved from prior guidance). FY26 guidance assumes no advanced computing AI hardware sales to hyperscale customers and the full wind-down of the Penguin Edge business by the end of the fiscal year. - Full-year FY26 non-GAAP gross margin is guided to 28.5% ±0.5 percentage points (up 0.5pp from prior guidance), with expected downward gross margin pressure in Q4 relative to Q3's favorable pricing environment. Total non-GAAP operating expenses are now guided to $260 million ±$5 million, and non-GAAP diluted EPS is expected to be $2.60 ±5 cents. - Preliminary planning guidance for fiscal 2027 calls for approximately 30% year-over-year growth in both total company net sales and non-GAAP diluted EPS, measured from the FY26 guidance midpoint. A full formal FY27 outlook will be provided on the next quarterly earnings call. For Advanced Computing, mid-teens year-over-year growth is a preliminary starting planning assumption, with the segment's wind-down impacts largely lapsing in 2027.

Segment performance

1. Advanced Computing: Q3 FY26 net sales of $138 million, representing 29% of total company net sales. This is 4% year-over-year growth and 19% sequential growth. Within the segment, the non-hyperscale AI infrastructure sub-segment grew 81% year-over-year and represented 58% of total Advanced Computing net sales (up from 33% in the year-ago quarter). 2. Integrated Memory: Q3 FY26 net sales of $275 million, representing 57% of total company net sales. This is 111% year-over-year growth and 60% sequential growth, driven by AI-driven demand and favorable pricing. 3. Optimized LED: Q3 FY26 net sales of $66 million, representing 14% of total company net sales. This is 7% year-over-year growth. Overall, combined AI-driven businesses (Integrated Memory and non-hyperscale AI infrastructure) represented 74% of total company net sales and grew 104% year-over-year. Total company net sales hit a record $479 million, up 48% year-over-year and 40% sequentially. Product net sales were $414 million (87% of total, up 60% year-over-year) and services net sales were $65 million (13% of total, down 1% year-over-year).

Risks & headwinds

- Forward-looking performance is subject to risks from uncertain macroeconomic conditions, shifting customer AI demand trends, and ongoing supply chain constraints, including extended component lead times that can slow order fulfillment and project ramp-up. - Industry-wide elevated memory costs may slow customer demand and compress gross margins for both the Advanced Computing and Integrated Memory segments. - The global and U.S. tax environment is rapidly evolving, which could lead to changes in the company's long-term non-GAAP effective tax rate. - Longer sales cycles for new AI infrastructure customer engagements could create variability in revenue recognition timing.

Analyst Q&A

• Q: The full-year memory guidance was raised to 90-95% growth after a strong Q3. How much of this increase comes from higher pricing vs new products like CXL cards, and what contribution do new hyperscale memory customers make to the 30% FY27 growth outlook?

  A: The raised guidance reflects both higher volumes and higher pricing for the overall memory business. CXL is a growing new product line, but most of the memory business remains focused on specialized data center memory modules for OEM customers. Hyperscale memory customer shipments are a small portion of the overall portfolio and do not represent a meaningful contribution to the preliminary 30% FY27 growth outlook, consistent with the existing FY26 run rate.

• Q: What is a preliminary growth guardrail for Advanced Computing in FY27, and how could component supply scarcity impact FY27 results?

  A: Mid-teeds year-over-year growth is the current preliminary starting assumption for the segment, as the impacts from the Penguin Edge wind-down and hyperscale customer transition largely lap in 2027, allowing core non-hyperscale AI infrastructure growth to drive results. The 3-6 month booking-to-revenue lag for non-hyperscale AI infrastructure gives the company good visibility into first half FY27 results, and the current outlook already accounts for existing component lead time constraints.

• Q: How is services demand evolving for AI infrastructure, and what strategic changes are being made to the growing memory business?

  A: Full end-to-end services (system integration, design, build, deployment, and long-term management of full AI factories) is a core strategic advantage for Penguin, as customers prefer a turnkey solution over stitching hardware together themselves, increasing customer ROI and deepening relationships. For memory, the key strategic focus is doubling down on data center-focused products, where AI-driven demand for memory is growing most rapidly.

• Q: Is there potential for additional monetization of Clusterware AI as agentic capabilities are added, and how does this interact with services adoption?

  A: The new AI factory operations agent enhances services differentiation by adding automation for infrastructure monitoring and management, increasing the value of Penguin's end-to-end offerings. It also creates new standalone software revenue opportunities: Clusterware AI can be sold to customers with existing AI infrastructure to help them improve operational efficiency, expanding Penguin's addressable market for software.