ABAB Pedia

SemiAnalysis

SemiAnalysis is a semiconductor and AI research firm founded by Dylan PatelDylan PatelDylan Patel is the founder and CEO of SemiAnalysis, researching semiconductor supply chains, datacenters and AI economics. He started the business in 2020 and expanded individual industry writing into data models, consulting and technical testing, while discussing inference costs and hardware–software co-design.Open the full entry in 2020. It produces reporting, institutional data models, consulting and hardware testing, covering chip manufacturing, memory, datacenters, inference costs and AI economics.

Contents11 sections
Key facts

origins as an individual research project

Dylan Patel founded SemiAnalysis in 2020, initially writing semiconductor analysis alone before building a multidisciplinary team. Its official account describes research connecting technical detail, supply chains and business decisions.⁠[1][2]

litigation involving a former employee

On March 27, 2026, SemiAnalysis sued former employee Wei Zhou over contract and trade-secret claims. Zhou subsequently alleged improper confidential-data use and dismissal; the company denied his account in its complaint. These are contested allegations, not judicial findings.⁠[3][4]

AI subscription-value study

Topics: research coverage

Coverage spans semiconductor equipment and materials, fabrication, packaging, memory, networking and datacenters, alongside AI training, inference and software. Its public newsletter is separate from institutional research and consulting products.⁠[6]

Topics: models and research methods

Its models include Accelerator & HBM, Datacenter Industry and AI Cloud TCO. Research relates chip shipments and manufacturing capacity to datacenter demand, using industry interviews, technical conferences and bottom-up estimates. Model outputs are research estimates rather than corporate financial disclosures.⁠[7][2]

Topics: the Tokenomics model

Here, Tokenomics concerns the supply, demand and economics of AI tokens. The model estimates inference throughput from installed accelerators, architectures and workloads, then connects it to application demand, revenue and future compute purchases. Institutional access combines Excel models, dashboards and analyst discussions.⁠[8]

Topics: InferenceX and reproducible testing

InferenceX, formerly InferenceMAX, measures inference throughput, latency, cost and energy on real accelerators, with public configurations, execution records and results. Fixed-sequence tests and AgentX multi-turn coding represent different workloads. Tests rerun after configuration changes and retain historical data.⁠[9]

Topics: cloud evaluation and chip teardown

Its tools include ClusterMAX for GPU cloud services, a fab map and a die-yield calculator. It also operates STEEL, a teardown laboratory in Oregon, extending its work into chip and process analysis.⁠[7][4][10]

Topics: interpreting subscription tests

The study isolated input, output, cache-write and cache-read usage, then valued a specified workload mix. Plan limits and model choices change the result; API-equivalent amounts are not cash subsidies. The associated institutional dashboard is separate from an ordinary newsletter subscription.⁠[5]

Official website and public accounts

Sources

  1. Dylan Patel — Founder and CEO (opens in a new window)
  2. Our story and research approach (opens in a new window)
  3. SemiAnalysis LLC v. Wei Zhou — complaint, CGC-26-635328 (opens in a new window)
  4. Dylan Patel’s Information Machine (opens in a new window)
  5. Anthropic Subscriptions Offer 5x+ More Value Than OpenAI (opens in a new window)
  6. About SemiAnalysis (opens in a new window)
  7. SemiAnalysis — research, models and tools (opens in a new window)
  8. Tokenomics Model (opens in a new window)
  9. InferenceX — methodology and reproducibility (opens in a new window)
  10. Dylan Patel — STEEL launch announcement (opens in a new window)
  11. SemiAnalysis | x (opens in a new window)
  12. SemiAnalysis | li (opens in a new window)
  13. SemiAnalysis | yt (opens in a new window)