Confidential Computing: Companies, Startups & Research

Intel Corporation, Advanced Micro Devices, Google, Microsoft, Nvidia and Huawei are prominent companies in this field. Intel and AMD develop confidential virtual-machine technologies, while Google works on enclave-based private AI compute. Nvidia is associated with confidential accelerators, Microsoft with secure VM migration, and Huawei with confidential-VM patents and protected I/O methods.

Flashbots, CYSEC SA, Rebellions Inc and Phala Network work on distinct confidential-computing applications. Flashbots explores TEE-backed blockchain protocols, CYSEC develops verified launch and remote attestation for VMs, and Rebellions has patented trusted-execution methods. Phala Network builds a confidential smart-contract platform. These are different areas of activity, not a ranking.

ETH Zurich, Southern University of Science and Technology and the University of Birmingham are active research organizations in the field. ETH Zurich studies weaknesses in confidential VMs, SUSTech examines VM architectures including RISC-V and Arm CCA, and the University of Birmingham investigates practical attacks on TEEs, including memory-aliasing attacks.

FACTS records $146,605,949 in grants across 62 awards since 2020. The grants indicate sustained public or research-related support, but do not describe all private investment or commercial spending in the sector. Funding-round records are also present, though they can include broader company financing that is not specific to confidential computing.

Huawei leads the listed patent holders with 67 patents, followed by Intel Corporation with 36. Huawei Technologies Co., Ltd. has 30, and BEIJING VOLCANO ENGINE TECHNOLOGY CO., LTD has 23. Other notable holders in the data include Microsoft and Alipay Information Technology Co., with 18 patents each.

Patent activity fell 38.8% in the latest 12 months, while publications grew 10.9%. FACTS records 624 patents in the 24-month window. These measures point in different directions: research output is rising even as patent activity declines, so patent counts alone do not capture the pace of technical development.

The main approaches are process enclaves, confidential virtual machines, Arm realms, confidential containers and confidential accelerators. Process enclaves isolate selected applications, while confidential VMs protect a full VM from its host. Containers adapt isolation to cloud-native workloads, and accelerators extend it to GPU computing. Each approach makes different trade-offs in software changes, integration and assurance.

The United States leads with 5,369 documents, followed by China with 2,027 and the United Kingdom with 1,065. Germany has 1,026 documents. These figures describe the geographic distribution of activity in FACTS; they do not show that work is limited to those countries or that all activity is concentrated in a single organization.

AI and machine learning account for 498 documents, while financial services account for 408 and cloud infrastructure for 213. Other recorded applications include collaborative data analytics and healthcare. The use cases center on processing sensitive information—such as private datasets or financial records—without exposing it to the surrounding infrastructure.

Recent examples include NANOZONE in October 2026 and PriPTune in September 2026. NANOZONE examines memory protection on Arm CCA, while PriPTune proposes TEE-GPU collaboration for privacy-preserving LLM fine-tuning. Other current research addresses device attestation across TEE and TPM platforms, runtime integrity, and extending applications into trusted execution.