Penguin Solutions Smashes Through Its Own Guidance as Neocloud Bookings Surge and Memory Backlog Stretches Four Quarters Out
Q4 FY2026 earnings call, October 6, 2026 — fiscal fourth quarter and full-year results, FY2027 outlook
Penguin Solutions used its fiscal fourth-quarter call to deliver one of the more dramatic guidance resets investors will see this earnings season, lifting its FY2027 outlook well above the preliminary framework management laid out just 90 days earlier. CEO Kash Shaikh told analysts the company now expects net sales to grow approximately 40% at the midpoint in FY2027, to roughly $2.43 billion, with diluted EPS growing about 55% to $4.45. That compares with the "preliminary" view shared last quarter of 30% growth off the fiscal 2026 midpoint — a view Shaikh characterized as "much more directional" than the formal outlook now on the table. Using the same FY2026 base, the new guidance actually implies closer to 45% growth, he noted.
The quarter itself validated the upgrade: net sales of $567 million were up 68% year-over-year and 18% sequentially, beating the high end of the company's own outlook, while diluted EPS of $1.00 rose 133% year-over-year. Full-year revenue reached $1.73 billion, up 26%, with EPS of $2.87 coming in $0.22 above the top of July's guidance range.
Advanced Computing Guidance Nearly Triples on Neocloud Demand
The most striking revision came in the Advanced Computing segment, where management had guided to "mid-teens" growth for FY2027 just last quarter and is now calling for approximately 40% growth, plus or minus 10 points. Asked directly what changed in 90 days, Shaikh pointed to bookings that simply outran expectations. "The booking strength, we have booked more than we expected, and we have a very strong pipeline," he said, adding that "our memory business strength continues and our AI infrastructure is further accelerating."
The quarter's results already showed the shift underway: non-hyperscale AI infrastructure sales within Advanced Computing grew 99% year-over-year and now represent 66% of segment net sales, up from 60% a year ago, more than offsetting continued runoff in legacy and edge businesses. Penguin added six new AI infrastructure customers in the quarter, including four neoclouds, a large quantitative trading firm, and an enterprise account, bringing the full-year total to 17 new AI infrastructure logos plus 12 existing customers that expanded their relationships.
Management disclosed a string of named neocloud engagements that illustrate the size of the opportunity. Penguin was selected by a neocloud backed by a major South Korean technology company to build an NVIDIA GB300 NVL72-based platform ahead of a broader GPU-as-a-Service rollout; it signed a multiyear deployment and 24/7 operations contract with a publicly traded neocloud carrying more than $3 billion in signed compute contracts; it partnered with Lektra to deploy carbon-free micro data centers; and, most notably, it was selected to deploy and operate a 36,000-GPU AI factory in Norway for a neocloud sitting on $10 billion of contracted compute from a leading AI lab. Shaikh framed the land-and-expand model — initial deployments followed by multiyear managed-services contracts spanning three to five years — as central to why these wins are converting into durable revenue rather than one-time hardware sales.
Memory Backlog Now Stretches Four Quarters, Not Just a Quarter of Supply Tightness
Integrated Memory delivered a record $341 million in the quarter, up 158% year-over-year and 24% sequentially, and now accounts for 60% of total company sales. What stood out to analysts was the backlog commentary: Citizens' Rustam Kanga pressed Shaikh on whether supply-constrained backlog — which last quarter was running roughly equal to recognized revenue — had started converting. Shaikh's answer suggested the opposite is happening. "We are seeing the same trend continue," he said. "Our backlog, if any, is much higher than the revenue that we recognized in this quarter," and that backlog "now extend[s] to at least four quarters" and "continues to strengthen."
Management was explicit that it views AI-driven data center memory demand as structurally different from the historically cyclical commodity memory market. "While memory demand is high everywhere... we believe that demand may still be just demand and supply dynamics and maybe cyclical," Shaikh said, "however, the demand in the data center for memory is driven by AI, especially with the agentic AI workloads... that is going to sustain much longer." He floated a provocative framing for the inference opportunity: if enterprise adoption of agentic AI workflows is currently around 10%, "for every 10% increase, that demand can double." New CXL-based memory expansion cards aimed at inference workloads are showing a "very strong pipeline" and accelerating bookings, positioning the segment beyond its traditional base.
Operating Leverage Is Real, Not Just a Slide
The FY2027 guide calls for EPS growth (55%) to outpace revenue growth (40%) for a second consecutive year, following a fiscal 2026 in which EPS grew 51% against 26% revenue growth. Operating expenses are guided to just $275 million, plus or minus $10 million, versus $267 million in FY2026 — implying opex growth of roughly 3% against 40% revenue growth. Shaikh attributed this to a deliberately low-cost operating footprint: engineering is more than 95% based in India, memory engineering sits in Taiwan, and manufacturing is concentrated in Malaysia, all locations he described as giving the company capacity that scales without a linear cost increase. He also credited internal AI adoption, noting that "100% of our software engineers, wherever they are, are using code generation" and that AI tools are now embedded across finance and legal functions as well.
Needham's Matthew Calitri pushed back on whether such restrained opex growth risked underinvesting in headcount and project bandwidth given the revenue trajectory. Shaikh maintained the company is still investing in innovation — pointing to ClusterWareAI and the CXL memory roadmap as proof — while insisting the efficient cost base allows profitable scaling without linear expense growth.
Balance Sheet Repositioned With a $750 Million Convertible Offering
Penguin used the quarter to refinance its capital structure, completing a significantly oversubscribed $750 million zero-coupon convertible note offering due 2031 at an initial conversion price of approximately $116.70 per share, a roughly 50% premium to the stock price at pricing. The company layered on capped call transactions with an initial cap price near $175.05 per share to limit dilution, and used part of the proceeds to retire $296 million of its 2029 and 2030 convertible notes and fully repay the $100 million outstanding on its credit facility. Cash ended the quarter at $647 million, up $207 million sequentially, even as cash used in operations for the year totaled $152 million, reflecting the working capital drag — accounts receivable and inventory both roughly tripled year-over-year — that comes with scaling a now much larger, more capital-intensive memory and AI infrastructure business.
A CFO Transition Lands in the Middle of the Growth Story
The call opened with the announcement that Stephen Cumming has joined as permanent Chief Financial Officer, with Aaron Johnson — who had served as interim CFO since July — handling the Q4 results and outlook presentation before reverting to his prior role as Vice President of Finance and Accounting. Management offered no detail on Cumming's background beyond describing him as having "a strong track record in financial leadership, operational scaling and investor relations," leaving investors to get acquainted with him "in the coming months." The timing of a CFO handoff during a period of such significant guidance revision and balance sheet activity is worth watching, though it did not appear to unsettle the quarter's financial execution.
Supply Diversification Signals Hint at NVIDIA Concentration Risk Being Managed
Buried in the operational update was Penguin's disclosure that it "established a relationship with another AI hardware supplier to help expand supply availability and support growing demand," alongside a separate new supply arrangement with "a leading memory supplier." Management did not name either partner, but the dual announcements suggest a deliberate effort to broaden sourcing beyond its deep, decade-long NVIDIA relationship — even as the company simultaneously touted its status as an NVIDIA AI Factory Specialized Partner, an "invitation-only designation" it says reflects the depth of that partnership. Investors should note the tension between leaning harder into NVIDIA-based reference architectures for credibility and sales velocity, while quietly diversifying the supply base against component and lead-time risk that management flagged could still affect quarterly shipment cadence.