Penguin Solutions, Inc. earnings recap
Q4 FY2026, reported October 6, 2026 · Beat 7 of last 7 quarters
Penguin Solutions, Inc. reported Q4 FY2026 revenue of $567M, a beat of 8.8% against consensus, and EPS of $1.00, a beat of 29.5%.
The Buildout is a website built by one person using AI agents, and could have factual errors.
Penguin's quarter shows the AI buildout broadening beyond GPU compute into the memory and operations layers: 78% of Q4 net sales came from AI-driven businesses growing 141% year-over-year, with Integrated Memory at a record $341 million. Record company backlog, memory backlog extending at least four quarters, and neocloud engagements — including designing a GB300 NVL72-based platform, a 36,000 GPU AI factory in Norway and multiyear operations contracts — suggest a growing services and managed-operations component to AI factory deployments. Strengthening demand and pipeline for CXL memory expansion cards points to memory capacity, not just compute, as an increasingly binding constraint on inference economics.
| Estimate | Actual | vs est | ||
|---|---|---|---|---|
| Revenue | $521M | $567M | +8.8% | beat |
| EPS | $0.77 | $1.00 | +29.5% | beat |
Penguin Solutions reported Q4 net sales of $567 million, up 68% year-over-year and 18% sequentially, with record gross profit dollars, operating income, operating margin, net income and adjusted EBITDA. Integrated Memory set a record at $341 million, or 60% of company net sales, up 158% year-over-year, while Advanced Computing was $154 million, up 11% year-over-year, with non-hyperscale AI infrastructure up 99% and representing 66% of that segment; Optimized LED was $72 million, up 7%. AI-driven businesses (non-hyperscale AI infrastructure and Integrated Memory) were 78% of Q4 net sales and grew 141% year-over-year, and full-year net sales were $1.73 billion, up 26%, with EPS of $2.87, up 51%. In the quarter the company added six new AI infrastructure customers, including four neoclouds, a large quantitative trading firm and an enterprise customer, plus two new memory customer logos; for the full year it added 17 new AI infrastructure customers and 15 new memory customers. Balance sheet and cash items included $647 million of cash and equivalents, inventory of $749 million, accounts receivable of $796 million, and cash used in operating activities of $163 million in the quarter, reflecting working capital investment to support growth.
Management formally raised fiscal 2027 guidance well above the preliminary view shared last quarter: net sales growth of approximately 40% at the midpoint, plus or minus 10 percentage points, to roughly $2.43 billion, which on the prior fiscal 2026 outlook midpoint represents approximately 45% growth. By segment, they guided Advanced Computing to approximately 40% growth (up from the mid-teens preliminary view), Integrated Memory to approximately 50% growth, and Optimized LED to be relatively flat. Gross margin is guided to approximately 28%, plus or minus 2 points, with operating expenses of approximately $275 million, plus or minus $10 million, growing considerably slower than sales; diluted EPS is guided to approximately $4.45, plus or minus $0.70, or about 55% growth at the midpoint, outpacing net sales growth through expected operating margin expansion. Kash Shaikh pointed to book-and-ship strength, a strong pipeline, a record company backlog, and accelerated demand from neocloud customers, and said memory backlog extends at least four quarters and continues to strengthen. Management flagged variability in customer deployment schedules, extended lead times for certain components, and evolving memory supply and cost assumptions as swing factors, and noted new supply arrangements with a leading memory supplier and an additional AI hardware supplier to support growth. The company also closed a $750 million 0% convertible note offering due 2031 and added Stephen Cumming as Chief Financial Officer, with Aaron Johnson returning to Vice President of Finance and Accounting after serving as Interim CFO.
“Our AI-driven businesses continue to scale. In Q4, non-hyperscale AI infrastructure and integrated memory represented 78% of company net sales and grew 141% year-over-year.”
on AI business mix and growth
“In AI infrastructure and Integrated Memory, bookings grew faster than net sales in the fourth quarter, and we ended fiscal 2026 with record company backlog and strong momentum heading into fiscal 2027.”
on Backlog and bookings
“For fiscal year 2027, we expect net sales to grow approximately 40% at the midpoint, plus or minus 10 percentage points or an increase of approximately $700 million to approximately $2.43 billion at the midpoint.”
on Fiscal 2027 guidance
What improved in the last 90 days to take Advanced Computing guidance from mid-teens to 40% growth, and how do neocloud customer unit economics compare to enterprise and quant customers — is there more hardware and what is the margin impact?
Kash Shaikh attributed the raise primarily to AI infrastructure: stronger-than-expected bookings since the last earnings announcement, a strong pipeline, continued memory strength and further accelerating AI infrastructure demand. On neocloud unit economics, he said they are quite similar to enterprise on-premises AI factory deployments, but noted a unique product-market fit because Penguin can design, build, procure and operate the factories for up to three to five years, with ClusterWare helping meet the SLAs those customers owe their offtakers.
What mechanics in the marketplace are driving increased neocloud capture and commercial enterprise activity, and how much are the NVIDIA and Dell relationships contributing to these dynamics at an increased pace?
Kash Shaikh said Penguin engages consultatively early in the design cycle with customers — whether large enterprises or neoclouds — rather than simply shipping hardware, drawing on more than two decades of cluster work with NVIDIA, and uses ClusterWare to tie the elements together. He said the NVIDIA AI Factory Specialized Partner designation adds credibility, and that in memory Penguin is focused on data center demand driven by agentic AI, which he views as more durable than cyclical consumer demand, with CXL memory expansion cards for inference workloads seeing growing pipeline and bookings.
How much of the fiscal 2027 Advanced Computing revenue outlook is already in backlog, and with the low year-over-year operating expense increase, are you investing enough in head count and bandwidth to pursue more projects?
Kash Shaikh declined to disclose what portion of the outlook is covered by backlog but said backlog is strong across both memory and AI infrastructure. On operating expense, he described an efficient model: more than 95% of engineering is in India, memory engineering is in Taiwan, and manufacturing is in Malaysia, so costs do not scale linearly with revenue; he added that AI code generation is used by 100% of the company's software engineers and AI is applied across functions including finance and legal, while noting Penguin will continue to invest in innovation such as ClusterWareAI and CXL.