Beam
Beam (Smartshare, Inc.) is a serverless GPU compute platform offering inference endpoints, task queues, and secure sandboxes through a Python SDK. It targets AI developers and teams needing sub-second cold starts, autoscaling from zero, and BYOC across nine cloud providers.
- Company typePrivate
- Founded2021
- HeadquartersNew York, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What Beam does
Beam (legal entity Smartshare, Inc.) is a serverless GPU compute platform purpose-built for AI workloads, founded in 2021 and headquartered in New York. The company offers three integrated products — Inference (serverless GPU model serving endpoints), Task Queues (asynchronous, GPU-heavy batch workloads with retries and callbacks), and Sandboxes (secure, stateful, isolated code execution environments for AI agents) — all exposed through a Python-native SDK using decorators such as @endpoint and @task_queue. Underlying the platform is the open-source beta9 container runtime (AGPL-3.0), which combines a gVisor user-space kernel, runc, memory snapshotting, GPU checkpoint restore, and a distributed storage layer to deliver sub-second cold starts even for large (10–100Gi) container images, scaling automatically from zero to thousands of GPU containers and supporting H100, H200, A100, B200, L40S, A10G, and RTX 4090/5090 hardware across 30+ regions and nine cloud providers via a bring-your-own-cloud model.
The company's go-to-market is product-led and community-led: developers self-onboard at beam.cloud, receive $30 in monthly free credits, deploy via a single `beam deploy` command, and engage through an active Slack community and technical blog. Revenue is generated through per-second GPU and CPU usage billing, a $89/month Team subscription tier (50 GPU containers, 1,000 CPU containers, 3 seats, live chat), and a custom-priced Growth enterprise tier with unlimited concurrency. A freemium bottom of the funnel funnels developers into higher tiers as workloads scale. Beam is SOC 2 Type II certified and runs workloads in non-root containers.
The company targets AI/ML developers and teams building inference endpoints, fine-tuning LLMs and diffusion models, processing batch audio/video with task queues, and executing untrusted or LLM-generated code through sandboxes. Named customers include Magellan AI (podcast ad intelligence), Geospy (geolocation CV), Hooktheory (AI music composition), Gepetto (virtual staging), Frase, Jamie, Ogilvy, Stratum AI, Invoke, EdgeImpulse, and Shippabo, spanning advertising, music technology, geospatial, SaaS, IoT, and logistics verticals. The company is venture-backed by Y Combinator, Tiger Global Management (lead), Charge Ventures, Hustle Fund, Soma Capital, and Alumni Ventures, with $3.5M disclosed and the company operating with 1–10 employees.
Beam firmographics
Firmographics- Name
- Beam
- Legal name
- Smartshare, Inc.
- Website
- https://beam.cloud
- Company type
- Private
- Founded year
- 2021
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- Beam (Smartshare, Inc.) is a serverless GPU compute platform offering inference endpoints, task queues, and secure sandboxes through a Python SDK. It targets AI developers and teams needing sub-second cold starts, autoscaling from zero, and BYOC across nine cloud providers.
- Ownership category
- akta.pro rank
Beam industry classification
Industry- Product category
- Serverless AI Cloud Infrastructure
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (51821), Computer Systems Design and Related Services (54151)
- SIC
- Services-Computer Programming Services (7371), Services-Computer Programming, Data Processing, Etc. (7370), Services-Computer Processing & Data Preparation (7374)
- akta.pro primary industry
- AI Compute Cloud & GPU-as-a-Service (HDAAAAAK)
- akta.pro secondary industries
- Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem) (HDAEANAC), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI), End-to-End Enterprise AI Platforms (MLOps & Model Lifecycle Management) (HDAEANAA), AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers) (HDAAAAAG)
Keywords
Where Beam is headquartered
LocationHeadquarters
- HQ city
- New York
- HQ country
- United States
- HQ region
- North America
Markets served
Beam business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Infrastructure, Personnel, Operations, Marketing or Sales
Revenue model
- Serverless GPU Compute: Pay-per-second billing for GPU compute (H100, H200, A100, B200, L40S, A10G, RTX 4090/5090). Billing stops the moment endpoint goes idle. No reserved capacity or minimums. Includes $30 monthly free credit.
- Serverless CPU Compute: Per-second billing for CPU and RAM. CPU at $0.0000528/core/second, RAM at $0.0000056/GB/second.
- Team Plan Subscription: $89/month subscription tier providing 50 GPU containers, 1000 CPU containers, 3 seats (additional at $25/seat), and live chat support.
- Sandbox Environments: Isolated secure environments billed per second for active runtime. GPU options include RTX 4090 and A10G at same rates as serverless.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Pay-as-you-go | Developer Plan: Free tier with usage-based GPU and CPU compute |
| Subscription | Monthly | Team Plan: $89/month with higher concurrency limits and support |
| Subscription | Annual | Growth Plan: Custom enterprise tier with unlimited concurrency |
| Usage-based | Pay-as-you-go | On-Demand GPU: B200 SXM6 starting at $3.93/hr |
| Usage-based | Pay-as-you-go | Serverless GPU: Per-second billing for RTX 4090 and A10G |
Go-to-market motion2 records
Distribution channels4 records
Marketing channels7 records
Beam product offering
Product offeringCore offering
Beam is a serverless AI cloud platform that provides on-demand GPU compute for running AI inference, async task queues, and secure sandboxed code execution. Developers deploy AI models and workloads through a Python-native SDK that abstracts away infrastructure management, with sub-second cold starts, autoscaling from zero, and per-second billing across multi-cloud environments.
Product overview
Beam is a serverless AI cloud platform offering three integrated products: Inference for deploying high-performance model serving endpoints, Task Queues for asynchronous GPU-heavy workloads, and Sandboxes for secure code execution. All three products share a unified Python-native SDK, enabling developers to go from Python code to production deployment without managing infrastructure, Dockerfiles, or YAML. The platform provides sub-second cold starts, autoscaling, and supports H100, H200, A100, B200, L40S, A10G, and RTX 4090/5090 GPUs.
Differentiator
Problem solved
Functional benefit
Products and services
- Inference Serverless GPU inference endpoints for deploying any AI model (LLMs, diffusion, custom PyTorch/TensorFlow) in production. Supports sub-second cold starts, autoscaling from zero to thousands of GPUs, H100/H200/A100/B200/L40S/A10G/RTX 4090/5090 hardware, and per-second billing. For AI/ML developers and teams building model-serving products.
- Task Queues Async task queues for running long-running, GPU-heavy functions across the cloud. Features automatic retries (3x by default), event-based callbacks, scheduled jobs, fan-out parallelism, and QueueDepthAutoscaler configuration. Built on the @task_queue Python decorator. For teams running batch processing, audio transcription, image pipelines, and training jobs.
- Sandboxes Secure, isolated code execution environments for AI agents and untrusted code. Stateful and persistent runtimes with memory snapshots, sub-second boot times, GPU support (RTX 4090, A10G), dynamic port exposure, file system snapshotting to create reusable templates, branch/restore into thousands of concurrent isolated runs, and Docker-in-Docker capability. For teams building AI agents and running LLM-generated code at scale.
- On-Demand GPU Compute On-demand GPU compute available across hardware types including B200 SXM6 ($3.93/hr), H200 SXM5 ($1.99/hr), H100 PCIe ($1.74/hr), A100 80GB SXM4 ($1.30/hr), L40S PCIe ($0.72/hr), RTX PRO 6000 PCIe ($1.04/hr), A6000 PCIe ($0.51/hr), RTX 5090 PCIe ($0.68/hr), and RTX 4090 PCIe ($0.42/hr). Per-hour billing for reserved workloads.
- Beam JavaScript/TypeScript SDK A JavaScript/TypeScript SDK for managing Beam applications from client-side code. Beta release extending the Python SDK's functionality to web and Node.js developers.
Quantifiable outcome
- 50% increase in throughput for Magellan AI
- +9 more outcomes
Companies that use Beam
Customer profileNamed customers12 records
Segments4 records
Ideal customer profiles3 records
Beam technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability9 records
Feature8 records
Beam partnerships and signals
Strategic signalPartnerships
Four partnerships are on record, tiered core.
- Amazon Web Services (AWS)coreAWS is a supported cloud provider for Beam's bring-your-own-cloud model. Users can connect their AWS accounts and Beam orchestrates GPU workloads on AWS infrastructure.
- Google Cloud Platform (GCP)coreGCP is a supported cloud provider for Beam's bring-your-own-cloud model. Users can connect their GCP accounts and Beam orchestrates GPU workloads on Google Cloud infrastructure.
- Microsoft AzurecoreAzure is a supported cloud provider for Beam's bring-your-own-cloud model. Users can connect their Azure accounts and Beam orchestrates GPU workloads on Azure infrastructure.
- HetznercoreHetzner is a supported cloud provider for Beam's bring-your-own-cloud model. Users can connect their Hetzner accounts for cost-effective GPU orchestration.
Scale indicators15 records
Recent moves6 records
Expansion highlights6 records
Beam competitors and assessment
Company assessmentDirect peers
- Modal: Modal is a serverless compute platform purpose-built for AI workloads, with Python SDK, autoscaling from zero, and GPU support. Direct competitor to Beam's Inference and Task Queues offerings with nearly identical product positioning.
- CoreWeave: CoreWeave is a large-scale GPU cloud provider built on top of NVIDIA hardware, serving AI training and inference customers. Direct competitor for GPU-as-a-service workloads at higher capacity tiers.
- Replicate: Replicate runs machine learning models in the cloud via API, with a focus on serving open-source models at scale. Direct competitor referenced in Beam's customer case studies (Gepetto migrated from Replicate to Beam).
- Lambda Labs: Lambda provides GPU cloud infrastructure and deep learning workstations, serving AI researchers and production teams. Direct competitor for on-demand GPU compute, particularly at the high-end GPU tier.
- Together AI: Together AI offers an AI compute cloud for training, fine-tuning, and running open-source models, with serverless inference endpoints. Direct competitor to Beam's inference offering, especially for LLM serving.
- RunPod: RunPod is a GPU cloud platform offering on-demand and serverless GPU compute for AI/ML workloads. Competes head-to-head with Beam on GPU pricing, instance types, and developer experience.
Broad incumbents
- AWS SageMaker: AWS's managed machine learning platform covering training, deployment, and inference at scale. Hyperscaler incumbent that competes broadly with Beam on managed AI infrastructure, though with a more enterprise-oriented and less developer-friendly experience.
- Google Vertex AI: Google Cloud's unified AI platform for building, deploying, and scaling ML models. Broad incumbent competing across inference, training, and MLOps use cases where Beam also plays.
- Paperspace (DigitalOcean): Paperspace, now part of DigitalOcean, offers GPU cloud instances and notebooks for AI/ML workloads. Competes in the same GPU-as-a-service category as Beam, with broader cloud portfolio backing.
Emerging players
- E2B: E2B provides secure code execution sandboxes specifically designed for AI agents and LLM-generated code. Directly comparable to Beam's Sandboxes product and a notable emerging competitor in the agent infrastructure category.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
Beam social profiles
Digital presenceBeam compliance and trust
Trust signalCompliance1 record
Beam financial estimates
Financial estimateRevenue estimate
Valuation estimate
Beam leadership team
Management profileNumber of profiles
Profiles2 records
Beam funding detail
Funding detailFunding overview
Funding rounds1 record
Investors13 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Beam M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Beam
What does Beam do?
Beam is a serverless AI cloud platform that provides on-demand GPU compute for running AI inference, async task queues, and secure sandboxed code execution. Developers deploy AI models and workloads through a Python-native SDK that abstracts away infrastructure management, with sub-second cold starts, autoscaling from zero, and per-second billing across multi-cloud environments.
Is Beam a public or private company?
Beam is a private company. It is classified as venture growth investor backed and is currently operating.
When was Beam founded?
Beam was founded in 2021. It employs 1 to 10 people.
Where is Beam based?
Beam is headquartered in New York, United States, in the North America region.
How does Beam make money?
Four revenue lines are on record. Serverless GPU Compute is the primary driver. The others are serverless CPU Compute, team Plan Subscription and sandbox Environments.
Who are Beam's main competitors?
Direct peers on record are Modal, CoreWeave, Replicate, Lambda Labs, Together AI and RunPod. Broad incumbents are AWS SageMaker, Google Vertex AI and Paperspace (DigitalOcean). E2B is listed as an emerging player.
Does Beam have an API?
Yes. Beam provides a Python-native SDK for deploying AI inference endpoints, task queues, and sandboxes. Developers use the @endpoint decorator and @task_queue decorator to define workloads in Python and deploy with 'beam deploy'. The SDK handles autoscaling, authentication, and telemetry automatically. A TypeScript/JavaScript SDK is also available (in beta). Developer documentation is at docs.beam.cloud.
What industry is Beam in?
Beam's product category is Serverless AI Cloud Infrastructure. Its primary akta.pro industry code is HDAAAAAK, AI Compute Cloud & GPU-as-a-Service, with a secondary code of HDAEANAC, Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem). Its NAICS code is 51821 and its SIC code is 7371.