$PENG This earnings report is truly impressive.

Not only did it beat expectations, but it also raised the full-year guidance again.

Q3 revenue hit $479 million, up 48% year-over-year, significantly exceeding the market's expectation of $401 million; Non-GAAP EPS reached $0.84, up 79% year-over-year, also clearly beating the market's estimate of $0.54. Adjusted EBITDA reached $67.6 million, up 51% year-over-year, with all core metrics hitting all-time highs for the company.

More importantly, the company raised its full-year outlook again.

The full-year revenue growth expectation was directly raised from around 12% to around 22%, and the Non-GAAP EPS guidance was raised from $2.15 to $2.60. Not only is this above market expectations, but management also stated that both revenue and EPS for this year are expected to exceed the upper end of the previous guidance range. This indicates that the AI demand they are seeing shows no signs of slowing down.

I think there are three key highlights in this earnings report.

First, the Memory business is even stronger than expected.

Integrated Memory revenue reached $275 million, up 111% year-over-year, not only continuing its growth but also doubling. Over the past four quarters, the company added 16 new Integrated Memory customers, with more and more customers expanding their procurement scale.

Management mentioned on the call that with the rapid development of Inference and Agentic AI, memory is gradually becoming one of the most important performance bottlenecks in AI systems. As model contexts get longer and agents need to store more states, demand for high-performance memory will continue to rise.

This is highly consistent with the message from Micron's last earnings report — the AI-driven Memory super cycle is being validated in more and more companies' results.

Second, AI Infrastructure is entering a harvest period.

Over the past four quarters, the company added 13 new AI Infrastructure customers, and added another 4 this quarter. At the same time, it officially became an NVIDIA AI Factory Specialized Partner and launched the AI Factory Operations Agent, continuing to improve the ClusterWareAI platform.

Management emphasized the AI Factory Platform throughout the entire call, rather than just the Memory business. I think many people still view PENG as a memory company, but it is gradually transforming into an AI Factory Platform company, moving from selling hardware to providing a complete set of solutions for AI cluster deployment, management, and operations.

Third, profitability is more stable than many feared.

Although DRAM costs have continued to rise recently, the company's Non-GAAP gross margin remains above 28%, without significant deterioration due to memory cost pressure. This indicates that the product mix and execution capabilities are both continuously improving.

Also, there is one more point worth noting. As of the earnings release, PENG's short float was close to 21%. This earnings report not only exceeded expectations across revenue, EPS, and EBITDA, but also raised full-year guidance again. If trading volume continues to increase tomorrow, it is possible that some short sellers will choose to cover, which could lead to significant short-term stock price volatility.

Overall, I think this earnings report further validates Penguin's transformation direction. On one hand, it benefits from the AI-driven Memory super cycle; on the other hand, its AI Infrastructure and AI Factory Platform businesses continue to accelerate. This is a company that benefits from both the Memory and AI infrastructure tracks. If the proportion of AI Factory business continues to increase in the future, the market's valuation logic for it may also continue to evolve.