1. What is Zoho Cloud?
Zoho Cloud is the infrastructure Zoho built for itself and now rents to developers. Zoho runs its 55+ business apps on hardware it owns, in its own data centres, on software it wrote on top of Linux and Postgres. Co-founder Sridhar Vembu has said Zoho does not host its products on AWS or Azure (Free Press Journal, Oct 2026).
That stack now serves more than one million paying organisations and 150 million users (Wikipedia). In June 2026 Zoho went one layer lower and unveiled Nathu La, its own server platform built with Intel on Xeon 6 (The Hans India). Developers reach all of this through Catalyst, which Zoho has just made free for students.
Four ways to build on Zoho
Zoho does not force one development model. You pick the layer that matches your team.
| Approach | Tool | Who it suits | Typical output |
|---|---|---|---|
| No-code / low-code | Zoho Creator + Deluge script | Business analysts, citizen developers | Forms, workflows, internal apps |
| Integration and automation | Zoho Flow, Deluge custom functions | Ops teams | Glue between Zoho and third-party SaaS |
| Extend Zoho apps | Developer Console, widgets, custom functions in CRM/Desk/Books | Zoho partners | CRM extensions, marketplace add-ons |
| Pro-code, full stack | Zoho Catalyst (Serverless, AppSail, CloudScale, QuickML) | Software engineers | Web apps, APIs, microservices, AI apps |
A fifth path is emerging. Catalyst 3.0 ships MCP connectors, agent skills and plugins so AI coding agents (Claude, Codex, Cursor, Copilot) can provision and deploy Catalyst services directly. Zoho also previewed AppOS at Zoho Day 2026, a shared data and process foundation for generated apps (Constellation Research).
The rest of this post focuses on the pro-code path, Catalyst, because that is where Zoho competes head-on with AWS.
2. Zoho vs AWS: focus vs breadth
AWS gives you every possible building block; Zoho gives you the ones most business apps actually use. That is the whole argument, and it cuts both ways.
A made-in-India cloud
Zoho is bootstrapped, profitable and headquartered in Chennai. It runs data centres in Mumbai, Delhi and Chennai among 18+ worldwide, and hosts data by country or region (Free Press Journal). For Indian enterprises and public-sector buyers who care about data residency under the DPDP Act, an Indian-owned stack end to end is a real differentiator, not just a slogan. Zoho is now pitching the same sovereign-cloud story in the UAE, Saudi Arabia and South Africa (Open Source For You).
The newcomer’s problem with AWS
AWS lists over 200 services, and many overlap. Want to run a container? You can choose Lambda, ECS Express Mode, ECS on Fargate, ECS on EC2, EKS, Elastic Beanstalk or Lightsail. Each has its own pricing model, IAM permissions and networking rules. A newcomer spends the first week choosing, not building.
To be fair, AWS has started pruning. It launched a Product Lifecycle page in 2025 and moved 19 services or capabilities into maintenance in October 2025 (usage.ai summary; AWS News Blog). But “maintenance” means existing customers keep using them, so the console, docs and Stack Overflow answers still carry years of legacy options.
What Zoho does instead
Catalyst has about 30 services on one console with one bill and $0 egress (Catalyst 3.0). There is usually one obvious service per job. When Zoho replaces a component, it retires the old one: File Store, Event Listeners and Cron reached end of life on 30 April 2026, replaced by Stratus, Signals and Job Scheduling (Catalyst release notes).
| Dimension | AWS | Zoho Catalyst |
|---|---|---|
| Service count | 200+ | ~30 |
| Ways to run a web container | 6+ | 1 (AppSail) |
| Egress fees | Charged per GB | $0 |
| Identity and permissions | IAM policies, roles, SCPs | Project roles + scoped collaborators |
| Compute model | VMs, containers and functions | Containers and functions, by design |
| Global regions | 30+ | Zoho data centres (US, EU, IN, CA and more) |
| Ecosystem tie-in | Huge third-party ecosystem | Native hooks into Zoho CRM, Creator, Books, Cliq |
| Best for | Any workload at any scale | Business apps, APIs, AI features, Zoho-centric shops |
The trade-off is honest: Zoho is simpler because it does less. If you need GPUs, Kafka, a custom VPC or 30 regions, AWS still wins.
3. Example: an expense-claims app on Catalyst
A complete expense app (web UI, login, receipt upload, OCR, database, weekly report) needs nine Catalyst components and one CLI. The same app on AWS touches about twelve services plus IAM wiring between them.
The app: employees log in, upload a receipt photo, Zia OCR reads the amount, the claim is saved, and managers get a weekly summary.
Step 1: create the project
npm install -g zcatalyst-cli
catalyst login
catalyst init # pick: Functions (Advanced I/O, Node.js), Client (Slate)
In the console, add a Data Store table Expenses (columns employee, amount, vendor, receipt_key, status) and a Stratus bucket receipts. Turn on Authentication with the hosted login page.
Step 2: the API function (Advanced I/O, Express)
// functions/expenses/index.js (simplified)
const express = require("express");
const multer = require("multer");
const catalyst = require("zcatalyst-sdk-node");
const app = express();
const upload = multer({ storage: multer.memoryStorage() });
app.post("/claims", upload.single("receipt"), async (req, res) => {
const cat = catalyst.initialize(req); // inherits the logged-in user
const user = await cat.userManagement().getCurrentUser();
const key = `${user.user_id}/${Date.now()}.jpg`;
await cat.stratus().bucket("receipts").putObject(key, req.file.buffer);
const row = await cat.datastore().table("Expenses").insertRow({
employee: user.email_id,
receipt_key: key,
status: "PENDING_OCR",
});
res.status(201).json(row);
});
app.get("/claims", async (req, res) => {
const cat = catalyst.initialize(req);
const rows = await cat
.zcql()
.executeZCQLQuery("SELECT * FROM Expenses ORDER BY CREATEDTIME DESC");
res.json(rows);
});
module.exports = app;
Step 3: OCR on upload (Signals + event function)
Add a Signals rule on the receipts bucket for Object Upload (Stratus Signals docs). It invokes an event function that calls Zia OCR, parses the total and updates the row to SUBMITTED. No queue, no IAM role, no permissions policy to write.
Step 4: weekly report and deploy
Create a cron in Job Scheduling for Monday 09:00 IST that runs a Job function to email managers via Catalyst Mail. Then:
catalyst deploy # functions + Slate frontend + config, one command
Catalyst gives you a development and a production environment out of the box, logs in Catalyst Logs, and CI/CD through Catalyst Pipelines.
Same app, service by service
| Job | Zoho Catalyst | AWS equivalent |
|---|---|---|
| Host the React UI | Slate | S3 + CloudFront, or Amplify Hosting |
| User sign-up and login | Authentication | Cognito user pool + hosted UI |
| REST API | Advanced I/O function | API Gateway + Lambda |
| Relational data | Data Store (ZCQL) | RDS / Aurora, or DynamoDB |
| Receipt storage | Stratus | S3 |
| React to uploads | Signals | S3 Event Notifications / EventBridge |
| OCR | Zia OCR | Textract |
| Scheduled job | Job Scheduling | EventBridge Scheduler + Lambda |
| Catalyst Mail | SES | |
| Cache | Cache | ElastiCache |
| Logs and monitoring | Catalyst Logs | CloudWatch Logs + X-Ray |
| CI/CD | Pipelines | CodePipeline / CodeBuild or GitHub Actions |
| Permissions between services | Implicit within the project | IAM roles and policies per Lambda |
| Infrastructure as code | catalyst.json + CLI |
CloudFormation, SAM or CDK |
The last two rows are where newcomers lose days on AWS. On Catalyst, every component in a project can already talk to every other one, scoped by the end user’s role.
4. Example: a Linux workload, the container way
Hand-built Linux servers are the legacy way to run software. The modern way is to package your app and its Linux userland as a container image, and let the platform run it. Zoho skips the VM era entirely: AppSail, Catalyst’s managed PaaS, runs any OCI-compliant Docker image (AppSail docs). No SSH, no patching, no snowflake servers. That is the point, not a gap.
AppSail’s custom runtime lets you run Go, Rust, PHP, Ruby, Deno, Bun or any version of Java, Python or Node.js, and pull images from Docker Hub, Amazon ECR or Google Artifact Registry (Catalyst blog).
The Zoho way: three files, two commands
# Dockerfile
FROM ubuntu:24.04
RUN apt-get update && apt-get install -y nginx && rm -rf /var/lib/apt/lists/*
COPY site/ /var/www/html/
COPY nginx.conf /etc/nginx/sites-available/default # listens on 8080
EXPOSE 8080
CMD ["nginx", "-g", "daemon off;"]
docker build -t linux-web:latest .
catalyst appsail:add # choose "Docker Image" -> docker://linux-web:latest, port 8080
catalyst deploy appsail # done: you get an HTTPS AppSail URL
AppSail handles TLS, the public URL, scaling, logs and the host OS. Add a custom domain from the console when you’re ready.
The legacy way: a hand-built EC2 server
- Pick a region, an AMI (Ubuntu 24.04) and an instance type from hundreds of options.
- Create or choose a VPC, a public subnet, an internet gateway and a route table.
- Create a security group and open ports 22, 80 and 443.
- Create a key pair and store the
.pemfile safely. - Size the EBS root volume and choose its type.
- Attach an IAM instance profile if the server needs S3 or other services.
- Launch, allocate an Elastic IP, and associate it.
- SSH in,
apt install nginx, deploy your files, configure systemd. - Get TLS: Certbot on the box, or an ALB + ACM certificate.
- Install the CloudWatch agent, schedule EBS snapshots, plan OS patching.
The result is a “pet” server: configured by hand, hard to reproduce, and yours to patch forever. Every SSH session is a chance for drift between what you think is running and what is. This is the pattern containers were invented to end.
Side by side
| Zoho AppSail (container) | AWS EC2 (bare server) | |
|---|---|---|
| Concepts to learn first | Project, AppSail service, Docker image | Region, AMI, VPC, subnet, IGW, SG, key pair, EBS, IAM, EIP |
| Steps to a live HTTPS URL | 3 commands: build, add, deploy | 10+ steps, or a CloudFormation template |
| OS patching | Rebuild the image; the host is Zoho’s job | Yours, on every server |
| Reproducibility | Same image in dev and prod | Hand-configured server drifts over time |
| Rollback | Redeploy the previous image tag | Restore a snapshot and hope |
| Portability | Same image runs on Docker, Kubernetes or any cloud | Tied to the instance and its AMI |
| TLS certificate | Automatic | Certbot or ALB + ACM |
| SSH into the box | Not needed: logs and redeploys replace it | Required for most changes |
Even on AWS, the right answer today is a container, not a bare server. But AWS makes you pick the container path yourself, and the “simple” option keeps moving. Its previous run-my-container service, App Runner, closed to new customers on 30 April 2026, and AWS now points users to ECS Express Mode (AWS docs). Zoho makes the container the only path. For a newcomer, that is exactly the right default: build an image, deploy it, and never think about a server again.
5. Example: an AI policy assistant on Zoho
A retrieval-augmented (RAG) chatbot over your HR policies takes one QuickML endpoint and one function on Zoho. On AWS it takes Bedrock, a knowledge base, a vector store, S3, Lambda, API Gateway and IAM.
What Zoho gives you for AI
- QuickML LLM Serving hosts open models; the current recommended one is GLM-4.7-Flash, a 30B-parameter mixture-of-experts model (QuickML docs).
- QuickML RAG adds your documents as context, with an Agentic RAG mode, tool calling and conversation mode (Catalyst blog).
- GenAI endpoints publish any LLM or RAG configuration as a REST API with OAuth and SDK snippets.
- Zia Services cover classic AI: OCR, object detection, face analytics, image moderation, barcode scanning and text analytics.
- QuickML pipelines train no-code prediction models on data pulled straight from Zoho CRM or Creator.
- ConvoKraft builds the chatbot front end.
Build it
- In the Catalyst console, open QuickML → RAG, upload the HR policy PDFs, pick GLM-4.7-Flash.
- Test questions in the playground, tune temperature and system prompt.
- Click Publish as GenAI endpoint. Catalyst returns the endpoint URL and auth details.
- Call it from a Catalyst function and surface it in your web app, Zoho Cliq or ConvoKraft.
// functions/askHR/index.js (illustrative; copy the exact payload from your endpoint's API page)
module.exports = async (req, res) => {
const { question } = req.body;
const r = await fetch(process.env.HR_RAG_ENDPOINT, {
method: "POST",
headers: {
Authorization: `Zoho-oauthtoken ${process.env.ZOHO_TOKEN}`,
"Content-Type": "application/json",
},
body: JSON.stringify({ query: question }),
});
res.json(await r.json());
};
The AWS equivalent
- Upload PDFs to an S3 bucket.
- Create a Bedrock Knowledge Base; choose an embedding model and a vector store (OpenSearch Serverless, Aurora pgvector or S3 Vectors).
- Request model access for the generation model.
- Write a Lambda that calls
RetrieveAndGenerate, give it an IAM role with Bedrock and S3 permissions. - Front it with API Gateway and Cognito.
import {
BedrockAgentRuntimeClient,
RetrieveAndGenerateCommand,
} from "@aws-sdk/client-bedrock-agent-runtime";
const client = new BedrockAgentRuntimeClient({ region: "ap-south-1" });
export const handler = async (event) => {
const { question } = JSON.parse(event.body);
const out = await client.send(
new RetrieveAndGenerateCommand({
input: { text: question },
retrieveAndGenerateConfiguration: {
type: "KNOWLEDGE_BASE",
knowledgeBaseConfiguration: {
knowledgeBaseId: process.env.KB_ID,
modelArn: process.env.MODEL_ARN,
},
},
}),
);
return { statusCode: 200, body: JSON.stringify(out.output) };
};
Side by side
| Zoho QuickML | AWS Bedrock | |
|---|---|---|
| Services to wire | QuickML + one function | S3, Bedrock KB, vector DB, Lambda, API Gateway, IAM, Cognito |
| Vector store | Managed for you | You choose and pay for one |
| Model choice | A few open models (GLM-4.7-Flash; Qwen being retired) | Dozens: Anthropic Claude, Meta Llama, Amazon Nova, Mistral and more |
| Token price | \(0.20 / M input, \)0.40 / M output (EveryDev) | Varies by model; frontier models cost far more |
| Data residency | Zoho-owned DCs, incl. India | Your chosen AWS region, incl. Mumbai and Hyderabad |
| Business data hooks | Native import from Zoho CRM, Creator | Connectors you build or buy |
| Custom training / GPUs | No-code pipelines only | SageMaker, full control |
For an internal assistant over company documents, Zoho is cheaper and faster to stand up, and the data never leaves an Indian-owned stack. If you need frontier-model reasoning, fine-tuning on GPUs or a specific vendor’s model, Bedrock and SageMaker are in a different league.
6. Verdict: when to choose which
Choose Zoho Catalyst when you want to ship a business app quickly and never think about infrastructure. Choose AWS when you need infrastructure itself as the product.
Pick Zoho Catalyst if:
- You are building web apps, APIs, internal tools or AI assistants, not running infrastructure.
- Your business already runs on Zoho CRM, Books, Creator or Cliq.
- You want a predictable bill, $0 egress and an Indian-owned, sovereign stack.
- Your team is small or new to cloud and can’t afford a week of IAM and VPC design.
Stay on AWS if:
- You need GPUs, custom networking or 30+ global regions.
- You need frontier AI models or custom model training.
- You depend on specialised services: streaming, data warehousing, IoT, HPC.
As an architect, I see Zoho Catalyst as proof that a cloud does not need 200 services to be useful. It gives a newcomer exactly what most applications need, built and run in India, and it keeps the catalogue small by retiring what it replaces. That is a design choice AWS, by its scale and history, cannot easily make.
Sources
- Introduction to Catalyst by Zoho
- Catalyst 3.0: The Agent-Ready, Full-Stack Cloud
- Catalyst SDK release notes (File Store, Event Listeners, Cron deprecation)
- AppSail benefits and custom runtimes
- Catalyst AppSail now supports Docker images
- QuickML GLM-4.7-Flash
- What’s new in Catalyst QuickML
- Stratus Signals triggers
- Zoho Catalyst on EveryDev (pricing)
- Zoho built its own cloud (Free Press Journal)
- Zoho Nathu La server (The Hans India)
- Zoho previews AppOS (Constellation Research)
- Zoho sovereign cloud in MEA (Open Source For You)
- Zoho Corporation (Wikipedia)
- AWS Product Lifecycle page announcement
- AWS October 2025 maintenance list (usage.ai)
- AWS App Runner availability change
Working on something similar?
We help teams take Generative AI, IoT and AWS systems to production. Talk to the engineers who write these articles.