Blyss

What Blyss does

Blyss operates a confidential AI inference platform that runs large language models inside hardware-based secure enclaves, making data observation mathematically impossible even by Blyss itself. The core product, branded Confidential AI, is exposed through an OpenAI-compatible API at enclave.blyss.dev and supports use cases including code copilots for proprietary codebases, embedding-based semantic search over private documents and images, and personal AI assistants trusted with emails and calendars. The security stack combines AMD SEV-SNP confidential VMs (encrypted VM memory with CPU-signed attestation), NVIDIA H100 Confidential Computing mode (encrypted PCIe traffic between CPU and GPU with signed VBIOS attestation), dm-verity filesystem measurement, and Let's Encrypt TLS certificates whose private keys are generated inside each enclave, with every attestation committed to a public Certificate Transparency log for continuous third-party verifiability.

The company was founded in 2023 by Samir Menon and Neil Movva, who bring prior experience at Apple, Yubico, and Nvidia spanning cryptography, AI research, and high-performance computing. Blyss is backed by Y Combinator (seed round of $500,000 closed April 2023) and is based in San Francisco. Its go-to-market is hybrid: an API-first, product-led motion targeting individual developers through self-serve signup and open-source tooling (Homomorphic Encryption SDK, blyss-verifier, sev-attest-tool, gpu-enclave-img), paired with a sales-led motion targeting enterprises via booked calls for confidential AI deployments. Revenue mechanics are usage-based API pricing, though specific tiers and price points are not publicly disclosed, and the service is presently positioned as a private beta.

Name

Blyss

Legal name

Blyss

Website

https://blyss.dev

Company type

Private

Founded year

2023

Operating status

Operating

Headcount range

1–10 employees

Short description

Blyss is a San Francisco-based startup, founded in 2023 and backed by Y Combinator, that provides a verifiably confidential LLM inference API by running AI models inside AMD SEV-SNP and NVIDIA H100 hardware enclaves with public Certificate Transparency attestation. It serves developers, enterprises with private data, and privacy-conscious individuals.

Ownership category

akta.pro rank

Product category

Confidential AI Infrastructure

NAICS

Computer Systems Design and Related Services (54151), Computer Systems Design and Related Services (5415)

SIC

Services-Computer Programming Services (7371), Services-Computer Programming, Data Processing, Etc. (7370)

akta.pro primary industry

Confidential Computing & Hardware-backed Protection (TEE/HSM) (HDADAFAJ)

akta.pro secondary industry

Privacy-Preserving On-Device ML (secure enclaves, TEEs, differential privacy) (HDAAAJAH)

Keywords

• Confidential AI inference
• Secure enclaves
• LLM API services
• Homomorphic encryption
• Privacy-preserving computing

Headquarters

HQ city

San Francisco

HQ country

United States

HQ region

North America

Markets served

GTM type

B2B and B2C

Offering type

Software

Cost components

Technology or R&D, Personnel, Operations

Revenue model

1. Confidential AI API Access: API access to confidential LLM service. OpenAI-compatible endpoint allows easy integration. Specific pricing not publicly disclosed.

Pricing tiers

Go-to-market motion3 records

Distribution channels3 records

Marketing channels4 records

Core offering

Blyss offers "Confidential AI," a verifiably confidential LLM API service that runs inference inside hardware-based secure enclaves (AMD SEV-SNP confidential VMs combined with NVIDIA H100 Confidential Computing GPUs). Customers access it through an OpenAI-compatible endpoint at enclave.blyss.dev, with every connection backed by a signed attestation recorded in a public Certificate Transparency log. The offering is complemented by open-source developer tools, including a Homomorphic Encryption SDK and a blyss-verifier Python package for independent verification.

Product overview

Blyss offers Confidential AI as its primary core product — a verifiably confidential LLM API that runs inside secure GPU enclaves using AMD SEV-SNP and NVIDIA H100 Confidential Compute technology. The OpenAI-compatible API enables text generation for applications like code copilots, semantic search, and personal AI assistants. Supporting developer tools include the Homomorphic Encryption SDK (open-source) and blyss-verifier Python package for verifying confidentiality claims. All VM and tooling code is open source.

Differentiator

Problem solved

Functional benefit

Brands

• Confidential AI: Verifiably confidential LLM API service using secure hardware enclaves (AMD SEV-SNP and NVIDIA H100 Confidential Compute) that keeps data encrypted even while in use.

Products and services

• Confidential AI Verifiably confidential LLM API service that runs AI models inside secure hardware enclaves (AMD SEV-SNP and NVIDIA H100 Confidential Compute), ensuring all user interactions remain encrypted and unobservable even by Blyss. Exposed as an OpenAI-compatible endpoint at enclave.blyss.dev/v1 for chat completions, code assistance, and semantic search.
• Homomorphic Encryption SDK Open-source SDK for homomorphic encryption available on GitHub (github.com/blyssprivacy/sdk), enabling developers to implement privacy-preserving computations.
• blyss-verifier Python package (pip install blyss-verifier) that lets users verify the confidentiality claims of the live Confidential AI service at enclave.blyss.dev in a single command against public Certificate Transparency logs.

Quantifiable outcome

• 100% confidential AI inference - impossible for anyone including Blyss to observe user interactions
• +1 more outcomes

Segments3 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

AI capability4 records

Feature5 records

Partnerships

Three partnerships are on record, tiered core.

• Let's EncryptcoreTechnology or Integration · 12 January 2024Let's Encrypt provides TLS certificates for confidential VMs. Each secure VM requests a fresh TLS certificate using a private key generated from within the VM, enabling end-to-end encrypted connections.
• AMDcoreTechnology or Integration · 1 January 2023AMD SEV-SNP technology provides hardware-based secure enclaves with encrypted VM memory. AMD CPUs with Zen 3 or later support the SEV-SNP feature used by Blyss for confidential computing.
• NVIDIAcoreTechnology or Integration · 1 January 2023NVIDIA H100 GPU Confidential Computing mode enables fast, confidential LLM execution. In CC mode, PCIe traffic is encrypted and GPU produces signed attestation. Currently exclusive to H100.

Scale indicators2 records

Recent moves6 records

Expansion highlights4 records

Direct peers

• Opaque Systems: Opaque Systems provides a confidential AI and analytics platform built on hardware secure enclaves (AMD SEV-SNP, NVIDIA H100) with similar attestation-based guarantees. It is the most directly comparable competitor in confidential AI infrastructure.
• Anjuna Security: Anjuna offers a confidential computing platform that runs workloads inside hardware enclaves (SEV-SNP, Intel TDX, Nvidia H100 CC) with attestation. Overlaps with Blyss on the confidential-VM / attestation layer powering enterprise AI and data workloads.
• Edgeless Systems: Edgeless Systems builds Constellation and Contrast for confidential AI and Kubernetes on confidential VMs with GPU support (including H100). Directly comparable stack and target customers (privacy-sensitive AI deployments).

Emerging players

• Mithril Security: Mithril Security focuses on confidential AI inference using trusted execution environments and open-source tooling. Smaller and earlier-stage than Blyss but addresses the same confidential-LLM-inference problem.
• TripleBlind: TripleBlind provides privacy-preserving AI/data collaboration using a different cryptography stack (not TEE-based). Comparable enterprise pitch of 'AI without exposing data' though the underlying mechanism differs from Blyss's enclave approach.
• Enveil: Enveil offers homomorphic encryption-based secure search and AI. Comparable value proposition of running AI/search over encrypted data, but uses FHE rather than hardware enclaves.
• Fireworks AI: Fireworks AI provides an OpenAI-compatible LLM inference API with a strong focus on performance and developer experience. Same distribution model (drop-in API), serving the same developer audience, though without the confidential-computing layer.
• Together AI: Together AI offers an OpenAI-compatible inference API for open-source LLMs on GPU infrastructure. Same product surface and target customer, but runs standard GPUs without confidential-compute attestation.

Broad incumbents

• Microsoft Azure Confidential Computing: Azure Confidential VMs (with AMD SEV-SNP and confidential GPU SKUs) provide the underlying infrastructure Blyss builds on. Azure is the largest incumbent offering confidential AI workloads natively, posing significant competitive risk if it ships a comparable managed confidential LLM service.
• NVIDIA Confidential Computing: NVIDIA offers the H100 Confidential Compute SDK and reference integrations that Blyss relies on. As a chip+platform incumbent, NVIDIA could expand into the application layer (managed confidential inference) and squeeze out independent software vendors like Blyss.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat3 records

Key risks5 records

Key highlights5 records

Customer concentration

• GitHubgithub.com/blyssprivacy
• Xx.com/BlyssDev
• LinkedInlinkedin.com/company/blyssprivacy

Revenue estimate

Valuation estimate

Number of profiles

Profiles2 records

Funding overview

Funding rounds1 record

Investors1 record

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Blyss