Penguin Solutions (NASDAQ:PENG) used its fiscal second-quarter earnings call to outline what CEO Kash Shaikh described as a market shift in artificial intelligence from model training toward production deployments increasingly focused on real-time inference. Shaikh, who led his first earnings call as CEO, said the transition is broadening demand beyond hyperscalers into enterprise, “Neocloud,” and sovereign AI markets, where the company is positioning itself as an “AI factory platform” provider spanning infrastructure, memory systems, software, and services.
For the second quarter of fiscal 2026, Penguin Solutions reported net sales of $343 million, non-GAAP gross margin of 31.2%, and non-GAAP diluted EPS of $0.52. CFO Nate Olmstead said total net sales were down 6% year-over-year, while gross margin improved by 0.4 percentage points to 31.2%. Non-GAAP operating margin was 13.2%, and adjusted EBITDA was $50 million.
Management frames AI inference as a driver of infrastructure and memory demand
Shaikh said Penguin is “building Penguin into an AI factory platform company,” describing six core elements of its platform:
- ICE ClusterWare, the company’s AI infrastructure management software
- Penguin Memory AI systems aimed at inference workloads
- Penguin Advanced Computing systems optimized for AI
- Penguin OriginAI reference architectures (“blueprints”) for AI factories
- End-to-end services (design, build, deploy, and managed services)
- A partner ecosystem including NVIDIA, SK Telecom, and partners such as Dell
Shaikh said the company plans to invest more in the AI factory platform, including product innovation and go-to-market initiatives, and announced the appointment of Ian Colle as senior vice president and chief product officer. Shaikh said Colle previously worked at Amazon Web Services and brings experience scaling AI infrastructure platforms and high-performance computing.
Quarterly segment results: Memory strength offsets Advanced Computing decline
Olmstead said Q2 product net sales were $279 million (down 8% year-over-year) and services net sales were $64 million (up 1%). Segment results were mixed, with strong growth in Integrated Memory partially offset by a sharp year-over-year decline in Advanced Computing.
Advanced Computing net sales were $116 million, or 34% of company total, down 42% year-over-year. Both Shaikh and Olmstead attributed the decline to the timing of large deployments, a transition away from hyperscaler concentration, and the “previously disclosed wind down” of Penguin Edge. Olmstead added that hyperscale hardware sales from Q2 last year did not recur in Q2 this year.
Despite quarterly volatility, management emphasized progress in non-hyperscale AI HPC. Shaikh said non-hyperscale AI HPC net sales grew 50% year-over-year for the first half of the fiscal year and represented more than 40% of first-half segment net sales, supported by bookings growth and new customer wins. He said the company added five new AI HPC logos in Q2, bringing first-half total new logos to seven, compared with three in the first half of last year. He cautioned that sales cycles can run 12–18 months, creating quarterly variability but supporting deeper, more durable customer relationships.
Integrated Memory net sales were $172 million, or 50% of total net sales, up 63% year-over-year. Shaikh cited AI-driven demand across networking, telecommunications, and computing, along with favorable pricing. Olmstead said the company is using the balance sheet to make strategic purchases as it competes for constrained materials.
Optimized LED net sales were $56 million, or 16% of total net sales, down 7% year-over-year. Shaikh said the company is maintaining a “disciplined approach to investment and capital allocation” amid mixed conditions.
Product initiatives: CXL, KV cache, and photonic memory appliance
At NVIDIA’s GTC event in March, Shaikh said Penguin announced two inference-centric initiatives aligned with its platform strategy. First, the company introduced the “Penguin Memory AI” server line built around Compute Express Link (CXL) memory expansion, including what he said was the immediate availability of a “Memory AI KV cache server,” which stores inference context to accelerate large language model responses. Second, Penguin expanded its OriginAI factory architecture portfolio with blueprints designed for larger workloads and low-latency inference requirements.
Shaikh also highlighted deployments and collaborations meant to demonstrate demand for production inference infrastructure. He cited a collaboration with Deepgram and Dell in which Penguin designed and deployed an inference environment using Dell PowerEdge servers and NVIDIA RTX PRO 6000 Blackwell GPUs to support speech-to-text, text-to-speech, and voice agent capabilities for healthcare and retail applications. He also referenced ongoing work with Georgia Tech’s AI makerspace, developed in partnership with NVIDIA.
In Integrated Memory, Shaikh said Penguin recently received a “substantial order for CXL cards” from a generative AI company building inference solutions and said it sold CXL-powered KV cache servers to a “tier one” financial institution for an on-premise AI factory. He also said Penguin is continuing development of a photonic memory appliance (PMA), formerly called OMA, intended to extend memory capacity and bandwidth for large-scale AI environments.
Shaikh noted Penguin was an early investor in Celestial AI, which he said was recently acquired by Marvell in a multi-billion-dollar transaction. Olmstead said Penguin received approximately $32 million of proceeds from the disposition of its investment in connection with that sale during the quarter.
Margins, balance sheet, and updated FY2026 outlook
Olmstead said Q2 non-GAAP gross margin improved sequentially and year-over-year, driven by product mix in Advanced Computing, favorable pricing in memory, and tariff recovery in LED. Looking ahead, he said gross margins are projected to be lower in the second half due to a higher mix of lower-margin AI hardware and memory sales, rising memory costs within AI factory solutions, and less tariff cost recovery in LED. He added that the company expects a modest sequential increase in operating expenses in the second half, reflecting seasonality and increased R&D investments, including for ClusterWare and Memory AI.
On the balance sheet, cash, cash equivalents, and short-term investments ended the quarter at $489 million, up $28 million sequentially. Penguin ended Q2 with $450 million of debt, down $20 million sequentially due to retirement of 2026 convertible notes. Olmstead said the company finished the quarter in a net cash position and has no scheduled debt payments due until 2029 based on the current maturity schedule. He also said Penguin spent $32 million to repurchase approximately 1.7 million shares during the quarter, with $64.5 million remaining under current repurchase authorizations as of Feb. 27, 2026.
For fiscal 2026, Olmstead said the company raised its full-year outlook, citing “solid half one performance” and an improved second-half outlook for memory. At the midpoint, Penguin now expects 12% net sales growth and non-GAAP diluted EPS of $2.15, up from prior expectations of 6% net sales growth and $2.00 of non-GAAP diluted EPS. He said the outlook assumes continued diversification of customer mix and excludes any Advanced Computing AI hardware sales to hyperscale customers, while also reflecting the wind down of Penguin Edge, which the company expects to “essentially cease” by the end of fiscal 2026.
Segment outlook ranges for the full year were updated as follows:
- Advanced Computing: -25% to -15% year-over-year net sales change
- Integrated Memory: +65% to +75% year-over-year net sales growth
- LED: -15% to -5% year-over-year net sales change
Olmstead said full-year non-GAAP gross margin is now expected to be about 28% ±0.5 percentage points, reduced by 1 percentage point to reflect a higher mix of lower-margin memory sales and higher memory costs in the AI hardware business. He reiterated expected full-year non-GAAP operating expenses of $250 million ±$5 million and noted the non-GAAP diluted share count outlook was lowered to approximately 53 million shares, primarily due to share repurchases.
In closing remarks, Shaikh said Penguin is “still in early shift” as AI transitions toward inference, but pointed to customer demand, product launches, and bookings momentum as supporting confidence in the company’s direction. He said management’s priorities include product innovation at the intersection of AI infrastructure and memory, execution speed, deeper customer engagement through its ecosystem, and continued customer diversification.
About Penguin Solutions (NASDAQ:PENG)
Penguin Solutions, Inc engages in the designing and development of enterprise solutions worldwide. It operates through three segments: Advanced Computing, Integrated Memory, and Optimized LED. It offers dynamic random access memory modules, solid-state and flash storage, and other advanced integrated memory solutions for networking and telecom, data analytics, artificial intelligence and machine learning applications; and supply chain services, including procurement, logistics, inventory management, temporary warehousing, programming, kitting, and packaging services.
