Anyscale — research brief

# Anyscale — research brief

**Slug:** anyscale · **Date:** 2026-09-21 · **Theme:** ai-infra (Distributed Computing / Managed Ray)

## Labels
- **FACT** — primary company page, pricing, docs, or official nav/CTA copy
- **SECONDARY** — press / third-party paraphrase
- **INFERENCE** — model teaching, explicitly labeled

## Company snapshot
- **Legal / brand:** Anyscale, Inc. (FACT — site footer © Anyscale, Inc. 2026)
- **HQ:** 600 Harrison Street, 4th Floor, San Francisco, CA 94107 (FACT — about)
- **Also:** Bangalore office (FACT — about)
- **Founded:** 2019; Ray developed ~2016–2017 at UC Berkeley RISELab (FACT — about timeline)
- **Founders (historical launch):** Robert Nishihara, Philipp Moritz, Ion Stoica (+ Michael I. Jordan on launch announcement) (FACT — 2019 Series A blog)
- **Positioning:** Unified AI platform / fully managed Ray from the creators of Ray (FACT — homepage, docs, product)

## Teaching arc (revenue ladder)
**Open-source Ray (free, self-managed) → Get Started with $100 Hosted credit → Hosted PAYG (Anyscale Credits / hr rate card) → BYOC production (your VPC / K8s + marketplace billing) → Committed contracts + Enterprise SLAs (+ Startup Program up to $20K credits).**

Land is open-source Ray + $100 trial credit on Hosted; expand is usage-based Hosted compute; production default for many teams is BYOC inside AWS/Azure/GCP/K8s; commit unlocks volume discounts and 24×7 support.

## Pricing (FACT — anyscale.com/pricing)
### Commercial axes
1. **Pay as you go** — only pay for compute used; no monthly fixed fees claimed on pricing page
2. **Committed contracts** — volume discounts; can use existing GPU reservations

### Deployment axes
| | **Hosted** | **BYOC** |
|---|---|---|
| Deployment | Limited regions; VMs | Any cloud/region or on-prem; VMs or K8s |
| Data residency | Anyscale-managed infra | Your VPC |
| Compute | Anyscale-hosted | Use existing GPU reservations |
| Billing | Monthly credit-card invoices | Invoice via Anyscale or AWS/Azure/GCP marketplace |
| Support | Business hours; 5 case submissions | Enterprise SLAs; 24×7; unlimited cases |

### Hosted usage rates (Anyscale Credits / hour — FACT pricing table)
| Instance | AC / hr |
|---|---:|
| CPU Only | 0.0135 |
| NVIDIA T4 | 0.5682 |
| NVIDIA L4 | 0.9542 |
| NVIDIA A10G | 1.3635 |
| NVIDIA A100 | 4.9591 |
| NVIDIA H / B / GB families | Contact (not listed as open rate) |

### Credits / programs (FACT — site CTAs)
- Nav CTA: **“Get Started with $100 Credit”** → authkit.anyscale.com (FACT — site nav)
- **AI Startup Program:** up to **$20K** in Anyscale credits; dedicated Field Engineers; run on your own cloud; stack with cloud-provider credits (FACT — anyscale.com/startup)

## Products (FACT — docs + product)
- **Anyscale Runtime** — optimized Ray-compatible runtime
- **Workspaces** — hosted VS Code / Jupyter / terminal; IDE connect (incl. Cursor)
- **Jobs** — batch inference, training, data processing; queues & schedules
- **Services** — managed Ray Serve extension; zero-downtime upgrades / HA
- **Clouds** — deploy Anyscale cloud against AWS, Azure, GCP, or Kubernetes
- **Observability / governance** — dashboards, metrics/logs, SSO/SAML/SCIM, RBAC, SOC 2 Type 2 (docs Trust Center claim)
- Open-source **Ray** ecosystem: Ray Data, Ray Train, Ray Serve; integrations (vLLM, etc. — homepage)

## Customers / logos (FACT — company site / LinkedIn company page claims)
Named on company surfaces: e.g. Runway (case study), Canva, TripAdvisor, Physical Intelligence, Coinbase (LinkedIn company about / homepage logo wall — treat as disclosed references, not revenue mix).

## What we do **not** claim
- No public ARR, valuation, headcount, or GPU MW figures used here (private company; not invented).

## Sources
1. https://www.anyscale.com/pricing
2. https://www.anyscale.com/
3. https://www.anyscale.com/about
4. https://www.anyscale.com/startup
5. https://docs.anyscale.com/get-started/what-is-anyscale
6. https://www.anyscale.com/resources/case-study/runway
7. https://www.anyscale.com/blog/founders-of-open-source-project-ray-launch-anyscale-with-20-6m-in-funding-to-democratize-distributed-programming