Original | Odaily Planet Daily (@OdailyChina)
Author | Wenser (@wenser2010 )
When SemiAnalysis is mentioned, the recent market upheaval in the U.S. storage and semiconductor industries triggered by its research report is still fresh in memory.
As an independent research and investment firm on track to generate over $100 million in annual revenue, SemiAnalysis now operates as a multifaceted organization encompassing consulting, model service platforms, and technology labs. Yet, the person leading this rapidly growing firm—highly praised by NVIDIA founder Jensen Huang and AMD CEO Lisa Su—is not a technician or engineer deeply immersed in chip manufacturing, but rather a “beekeeper and old-timer from online forums” who once kept bees in Minnesota and anonymously discussed technical issues on various U.S. enthusiast forums.
This episode of Person Decoded, Odaily Planet Daily, features Dylan Patel, founder of SemiAnalysis.
Founder of SemiAnalysis: A self-taught "forum tech guru" who has maxed out their skill points
Compared to Citrini, who focuses more on macro trends and long-term themes, SemiAnalysis takes a more focused approach on the semiconductor industry, and its founder, Dylan Patel (hereinafter referred to as Dylan), is truly an industry legend.
Early life: Beekeeper from rural Georgia, American version of a "tieba elder brother"
According to Dylan, who shared this during a food-focused interview on Latent Space, he grew up in the countryside of Georgia, USA, and attended the University of Georgia. After graduating, he even worked as a beekeeper in Minnesota for about a year and a half.
At that time, he was once in a state of “being lost.” Today, summarizing his personal journey in his own words, he says: “I feel like I’ve just gone through many phases of life... there doesn’t seem to be any clear, direct path to follow.” (Odaily Planet Daily Note: This refers to his transitions from a chip enthusiast, semiconductor forum moderator, anonymous chip blogger, to founder of an investment research firm, hedge fund founder, and many other roles.)
When it comes to his entry into the industry, it was anything but straightforward.
Between the ages of 8 and 12, he was highly active on semiconductor forums as a "forum warrior" (Odaily Planet Daily note: similar to Chinese tech enthusiast communities like Tieba, commonly referred to as "Tieba veterans"), where he deeply studied chip documentation and exchanged ideas with community enthusiasts by repairing hardware such as Xbox consoles, teaching himself semiconductor knowledge.
That’s how he began, as an anonymous chip blogger, sharing chip knowledge and discussing semiconductor manufacturing technologies and supply chains on platforms like Reddit, WordPress, and Silicon Twitter.
In May 2020, Dylan officially launched his personal blog channel, SemiAnalysis, with the goal of providing accurate and independent technical analysis of the semiconductor industry. At that time, the “AI explosion driven by GPT” had not yet occurred, and the semiconductor industry remained a niche technical field with very little deep content available in the market.
Initially, SemiAnalysis was a very niche personal content channel built on WordPress. Following repeated suggestions from his good friend Doug, Dylan later migrated it to the Substack platform and switched from a free model to a paid subscription model. (Odaily Planet Daily note: According to Dylan himself, this friend joined Substack in the years that followed.)
Since then, Dylan has built his own "personal business system" around paid content channels, including technical analysis, business consulting, and research reports on semiconductor supply chains, AI infrastructure products, cloud ecosystems, machine learning models, and more high-tech industries.
SemiAnalysis: From a one-person company to a global team of over 60
In 2025, SemiAnalysis had gradually evolved from Dylan’s original “one-person company” (OPC) model into a global company with a professional research team of approximately 60 people, and had established a dedicated chip and semiconductor teardown laboratory, STEEL (SemiAnalysis Teardown Engineering & Evaluation Lab), in Oregon, USA.
Last year, the firm’s revenue reached $20 million; this year, according to The Information, SemiAnalysis’s revenue is projected to exceed $100 million, primarily driven by subscriptions, models, and consulting from hyperscalers, semiconductor giants, startups, and institutional clients. According to Dylan himself, he plans to establish a venture capital firm in the future. Previously, he has invested personally or through SPV structures in approximately 20 startups and helped raise $50 million for the compute giant Fluidstack via an SPV financing structure.
External influence: Highly recognized by Jensen Huang, AMD CEO, and others
After nearly six years of development, Dylan and SemiAnalysis have become the industry textbook and essential guide for the AI sector and semiconductor industry.
Previously, NVIDIA founder Jensen Huang mentioned SemiAnalysis multiple times during his GTC developer conference keynote, specifically highlighting the details of its research reports, particularly the benchmark evaluations such as NVIDIA InferenceX, amounting to a public endorsement.
Earlier, in December 2024, after approximately five months of in-depth testing and benchmark evaluation of AMD’s MI300X GPU, the SemiAnalysis team published a critical report titled “MI300X vs H100 vs H200 Benchmark Part 1: Training - CUDA Moat Still Alive,” which noted that although the AMD MI300X GPU demonstrated strong theoretical hardware competitiveness, its ROCm software stack suffered from significant gaps—such as numerous bugs, poor usability, and an immature ecosystem—resulting in a far inferior real-world experience compared to NVIDIA’s CUDA software architecture, rendering it ineffective as a robust solution for training workloads.
Hours after the report was released, AMD CEO Lisa Su personally reached out to Dylan and spoke with him the following day. Notably, the conversation was originally scheduled for only 30 minutes, but due to the high volume of information, numerous feedback points, and detailed technical discussions among engineers, it ultimately extended to 90 minutes. This was a rare in-depth dialogue between the CEO of a company with a $100 billion market cap and an independent third-party investment research firm. In the end, Su publicly thanked him for his “constructive feedback” (even when critical) (Odaily Planet Daily note: the original quote was “Feedback is a gift even when it's critical”).
High praise and positive response from AMD CEO Lisa Su
In April 2025, SemiAnalysis againreleased a follow-up report stating, “Over four months later, AMD has accelerated progress in ROCm, developer relations, CI/CD, and more, with the AMD MI450X poised to outperform NVIDIA,” and also acknowledged AMD’s subsequent improvements, becoming a landmark case where independent research directly influenced major corporate decisions.
In early June, Citrini analyst Jukan shared part of a SemiAnalysis report, which stated that “NVIDIA’s next-generation AI server cluster, Rubin NVL72, has undergone significant changes in memory configuration. To address supply chain constraints and ensure timely delivery of Rubin racks, the capacity per rack has been drastically reduced from the originally planned 55TB to 28TB—a 50% decrease—by adopting a reduced-spec 96GB SOCAMM memory module instead of the previous 192GB high-capacity module.” This news likely contributed to downward pressure on memory-related stocks, including Micron and SK Hynix, on the same day.
In response, Dylan said: "I like how people often quote us out of context. Actually, our original report never used such clickbait headlines." He later posted a picture showing the original report title as "Thanks for the Memories...".
In contrast, SemiAnalysis places greater emphasis on "technical implementation details," focusing on real bottlenecks in AI development—such as power shortages, supply chain dependencies, inference scalability, and NVIDIA ecosystem dynamics—and often integrates realistic constraints into optimistic demand forecasts, making it more suitable for guiding specific investment and industry decisions. For more information on SemiAnalysis, see 《From Community "Hardware Enthusiasts" to AI's "Hyde Park": How SemiAnalysis, Earning Nearly $100 Million Annually, Is Disrupting the Semiconductor Market?》
Recommended reading:
Latent Space Season 1: Scenario Cooking with SemiAnalysis Founder Dylan Patel