Penguin Solutions, Inc. (PENG) | The Buildout — AI Infrastructure
The Verdict
Penguin Solutions designs, integrates, and manages high-performance AI computing infrastructure and specialty memory products. It operates at the intersection of compute and memory, offering full-stack AI factory platforms that combine GPU clusters, orchestration software, and advanced memory systems. The company also manufactures specialty DRAM and flash modules, including emerging CXL memory products for inference workloads, and provides related services for enterprise, sovereign, and research AI deployments.
| Market Cap | — |
| Revenue (TTM) | $1.5B |
| Revenue Growth | +12.0% |
| EBITDA Margin (TTM) | 10.9% |
| Net Debt | $62M |
| Earnings Beats | 7 of 7 |
| P/E (TTM) | — |
| EV/EBITDA (TTM) | — |
What We Like
- Memory segment grew 63% in Q2 FY2026 and full-year guidance was raised to 65–75% growth, driven by AI demand and tight supply.
- Non-hyperscale AI/HPC revenue grew 50% in the first half of FY2026 and now represents more than 40% of Advanced Computing sales, with 7 new logos won.
- CXL memory products moved to commercial orders, including a substantial order from a generative AI company and a KV Cache server deployment at a Tier 1 financial institution.
- Balance sheet carries $489 million in cash with no debt maturities until 2029, enabling strategic memory inventory pre-buying.
- Partnerships with NVIDIA (AI Factory Specialized Partner) and Dell (2026 AI Partner of the Year) deepen, expanding customer reach.
What We’re Watching
- Advanced Computing guidance lowered to -25% to -15% for FY2026; bookings conversion expected to drive FY2027 recovery, but lumpiness risk persists.
- Gross margin guided down 100bp to 28% due to memory mix shift; further compression possible if memory input costs keep rising.
- CFO departure effective July 8, 2026 creates leadership gap during strategic pivot; successor not yet named.
- Memory growth depends heavily on securing materials; high end of guide (75%) is at risk if supply remains constrained.
The thesis is strengthening as memory revenue accelerates and the non-hyperscale AI compute business gains traction, but the gross margin compression from the memory mix shift tempers the benefit. The bookings pipeline and CXL traction provide early evidence that the AI factory platform strategy is working, while the deliberate exit from hyperscale and Edge businesses creates a cleaner base for FY2027. The key open question is whether the memory pricing cycle can sustain through the fiscal year without eroding margins more than guided.
Earnings Beat
Penguin reported Q3 FY2026 revenue of $478.7 million, up 40% sequentially, with gross margin of 27.8% and EBITDA of $63.3 million (13.2% margin). Management disclosed that AI-driven businesses — Integrated Memory and non-hyperscale AI infrastructure — accounted for 74% of revenue, reflecting surging memory demand and early conversion of the Advanced Computing pipeline.
| Metric | Q3 FY2026 | Q2 FY2026 | Q3 FY2025 | YoY |
|---|---|---|---|---|
| Revenue | $479M | $343M | $324M | +47.6% |
| Gross margin | 27.8% | 27.3% | 29.3% | -150bps |
| EBITDA | $63M | $38M | $24M | +166.0% |
| EPS | $0.84 | $0.70 | $0.05 | +1571.1% |
Management tone: No earnings call on record for the latest period.
Management Guidance
On the Q2 FY2026 call, management guided full-year total sales growth of ~12%, non-GAAP EPS of $2.15 ± $0.15, and non-GAAP gross margin of 28% ± 0.5 pp. Advanced Computing is expected to decline 25–15%, while Integrated Memory is guided to grow 65–75%. The outlook assumes no hyperscale AI hardware sales and that the Penguin Edge business essentially ceases by year-end.
Trajectory
Revenue accelerated sharply in Q3 FY2026 to $478.7 million, a 40% sequential increase, driven primarily by surging memory demand and the beginnings of non-hyperscale AI infrastructure bookings conversion. Gross margin remained under pressure at 27.8%, reflecting the higher memory mix, but EBITDA margin expanded to 13.2% on operating leverage. The prior-year quarter had revenue of $324.3 million, indicating the recent step-up in scale.
The Model
The model projects FY+1 revenue of $2,201 million and EBITDA of $361 million (16.4% margin). FY+2 revenue rises to $2,620 million with EBITDA of $422 million (16.1% margin). Near-term projections are anchored by memory segment growth in the 65–75% range and the gradual conversion of the Advanced Computing bookings pipeline. The FY+2 year assumes continued enterprise AI adoption and the lapping of Edge and hyperscale headwinds, though the model's EBITDA margin implies ongoing pressure from the memory mix shift.
| Metric | FY2025 | Next FY (E) | Following FY (E) |
|---|---|---|---|
| Revenue | $1.4B | $2.2B | $2.6B |
| YoY Growth | — | +60.8% | +19.0% |
| EBITDA | $114M | $361M | $422M |
| EBITDA Margin | 8.4% | 16.4% | 16.1% |
Projections are the median of 5 independent model runs. The model’s revenue sits 6.6% above analyst consensus.
On the Q2 FY2026 call, management guided full-year total sales growth of ~12%, non-GAAP EPS of $2.15 ± $0.15, and non-GAAP gross margin of 28% ± 0.5 pp. Advanced Computing is expected to decline 25–15%, while Integrated Memory is guided to grow 65–75%. The outlook assumes no hyperscale AI hardware sales and that the Penguin Edge business essentially ceases by year-end.
What Could Go Right — and Wrong
- Memory demand proves structural and outlasts the pricing cycle, sustaining 65–75% growth through FY2026 and into FY2027.
- Advanced Computing pipeline converts faster than the modeled lag, driving revenue recovery and closing the segment's negative growth gap.
- CXL and MemoryAI products scale to represent more than 10% of total sales, cementing Penguin's differentiation at the compute-memory intersection.
- Photonic Memory Appliance reaches commercial deployment, providing a proprietary hardware advantage in memory pooling for AI inference.
- Gross margins stabilize at 28% or above as memory input costs moderate and higher-value integration services grow.
- Memory pricing reverses as supply catches up, causing revenue to fall short of guidance and compressing margins further.
- Advanced Computing fails to convert bookings; deployment delays continue and the segment shrinks beyond current guidance.
- A new CFO reorients capital allocation, reducing investment in MemoryAI or PMA, delaying product roadmap.
- NVIDIA's enterprise push renders Penguin's AI factory integration less valuable, or a major competitor replicates the bundled memory-plus-compute model.
Looking Ahead
The next 12 months center on converting a robust bookings pipeline into revenue, especially as the Edge and hyperscale headwinds fade in FY2027. Memory supply and pricing dynamics will remain critical, while CXL and MemoryAI product launches could broaden the revenue base. Management must also navigate the CFO transition and sustain the pace of new logo wins across enterprise, sovereign, and neocloud segments.
- Expected July 7, 2026Q3 FY2026 earnings call — Tests early bookings conversion, memory pricing trends, and FY guidance update.
- By end FY2026 (Aug 29, 2026)Memory full-year outcome — Whether the memory segment hits the 65–75% growth target and supply remains adequate.
- H2 FY2026Adv. Computing H2 revenue — Signposts whether non-hyperscale AI/HPC is gaining enough scale to offset legacy declines.
- By end FY2026Penguin Edge wind-down complete — Completion would remove a structural drag and confirm a clean base for FY2027.
- UnknownNew CFO appointment — Could signal shifts in investment priorities or capital allocation for the AI platform build-out.
- Next few quartersSovereign AI contract awards — Large-scale wins would validate the platform beyond enterprise and research deployments.
Financials
Annual Summary
| Metric | FY2025 | TTM |
|---|---|---|
| Revenue | $1.4B | $1.5B |
| Gross Margin | 28.8% | 27.9% |
| EBITDA | $114M | $277M |
| EBITDA Margin | 8.4% | 10.9% |
| Net Income | $25M | $97M |
| Free Cash Flow | $106M | $90M |
| Net Cash | — | — |
Key Ratios (Trailing)
- P/E TTM—
- EV/EBITDA TTM—
- EV/Revenue TTM—
- Price/FCF TTM—
- Gross Margin (TTM)27.9%
- EBITDA Margin (TTM)10.9%
- Net Margin (TTM)6.4%
- ROIC17.8%
- FCF Conversion-40.5%
- SBC / Revenue1.5%
The Company
Penguin Solutions designs and integrates high-performance AI computing infrastructure and specialty memory products. Its Advanced Computing segment delivers GPU-accelerated clusters, orchestration software (ClusterWare), and AI factory reference designs (OriginAI) to enterprise, sovereign, and research customers. The Integrated Memory segment supplies DRAM modules, flash storage, and emerging CXL memory products for AI workloads, including the MemoryAI KV Cache server. These two segments position the company at the intersection of compute and memory for AI deployments.
Manufacturing and testing are conducted at facilities in Newark and Fremont, California; Penang, Malaysia; and Huizhou, China for LEDs. The company uses contract manufacturers including Celestica and discloses dependency on sole or limited-source suppliers for critical components such as GPUs (NVIDIA), processors (Intel, AMD), and memory chips. The Tempe, Arizona facility is being wound down with the Penguin Edge business.
Business Segments
Competitive Landscape
Penguin competes against compute and storage systems providers like HPE and Supermicro, specialty memory manufacturers, and cloud-based GPU infrastructure providers such as CoreWeave. In the 10-K, the company also lists enterprise IT server vendors and LED product manufacturers as competitors. Management argues that its combination of AI systems integration and specialty memory design differentiates it from competitors who lack one of those capabilities.
- Named in 10-K as a compute and storage systems provider; competes with Penguin's Advanced Computing AI factory offerings.
- Super Micro Computer (SMCI)Named in 10-K as a supplier and competitor; offers GPU baseboards and AI systems integration that overlap with Penguin's AI factory build-outs.
- CoreWeaveCloud-based GPU infrastructure provider; competes for on-premise AI factory spend by offering alternative cloud access to accelerated compute.
- Lambda LabsCloud-based GPU infrastructure provider; an alternative to on-premise AI deployments that Penguin sells.
- Cloud-based GPU infrastructure provider; competes for enterprise AI workloads that might otherwise go to Penguin's on-premise solutions.
Supply Chain
Penguin sits between upstream component makers (GPUs, DRAM, processors) and downstream enterprise, sovereign, and research customers deploying AI infrastructure. The company relies on sole or limited-source suppliers for critical components, including NVIDIA for GPUs and major DRAM manufacturers for memory chips.