Starting up in 2026
This roadmap takes one idea from a thought to a registered, shipped company with paying customers, in the order the work actually happens. It is written for a first-time founder with any idea, technical or not, and it defaults to India: by 31 March 2026 the Department for Promotion of Industry and Internal Trade had recognised more than 2.23 lakh startups, a record 55,200 of them in that one financial year, and India's tech startups raised about 10.5 billion dollars in 2025, third in the world behind the United States and the United Kingdom according to Tracxn. Switch to the UAE or Global tab and the registration, tax, trademark, payment, grant, incubator and event nodes change to that place; everything else stays.
Why it matters: Most first attempts fail on sequence, not on talent: people build before they validate, register before they sell, and raise before they have evidence. Each stage here ends with something you can check off before the next.
Do this: Read this whole stage first. Pick your track at the top. Open the one-page idea brief from the box beside the title and keep it open while you work through The idea and Validate. Mark nodes done, learning or skipped as you go; progress is saved in this browser and shared across the three tracks.
Watch out: Treating the map as a reading list. Every node ends in an action; the roadmap only works if you do the action before moving on.
Done when: You have chosen a track, read the three nodes in this topic that apply to you, and can say in one sentence what you are building and for whom.
- TemplateOneRoadmap: the one-page idea brief (PDF)
- OfficialStartup India: the Government of India's startup portal
- ArticlePaul Graham: Do things that don't scale
- VideoY Combinator: how to get and evaluate startup ideas
What a startup is (and is not)
A startup is a new company searching for a way to make money that it can repeat and grow, under real uncertainty about whether anyone wants the thing. Steve Blank's definition, a temporary organisation searching for a repeatable and scalable business model, is the useful one because it says what the early work is: searching. A small business executes a known model from day one, a kirana store or a salon, and that is a fine thing to build, but it does not need this roadmap's validation stage. A startup does not have to be venture-backable, software, or in Bengaluru; a D2C brand, a clinic chain, a tutoring company or an agency can all be startups while they are still finding the model.
Why it matters: Calling everything a startup leads people to copy venture playbooks, raise money and hire before there is a business. Knowing which thing you are building decides whether you need a cap table or a shop licence.
Do this: Write down which of three things you are building: a product company that must find its model, a service company that will sell time and later productise, or a known business in a new place. Keep the answer; it changes the legal structure, the funding stage and how much of the build stage applies to you.
Watch out: Deciding you are a startup because the word sounds better. The test is uncertainty about the model, not the technology in the product.
Done when: You can explain to a friend, without the word startup, what you will sell, to whom, and why you are not yet sure it will work.
Why 2026 is the cheapest year to start
The cost of a first version has collapsed. A non-programmer can describe an application to Lovable, Bolt or Replit and have a working web app with a database and sign-in in an afternoon; a founder who can read code can direct Claude Code or Codex to build the real thing for about twenty dollars a month; hosting on Vercel, a database on Supabase and analytics on PostHog all start free. Startup programmes then pay for the next year: AWS Activate, Google for Startups and Microsoft for Startups offer cloud credits from a few thousand dollars for unfunded founders to six figures for funded ones, and Anthropic, GitHub, Notion and HubSpot have programmes of their own. The policy side moved too: India abolished the angel tax in the Finance Act 2024, extended the Section 80-IAC tax holiday to startups incorporated up to 1 April 2030, and approved a second 10,000 crore rupee Fund of Funds in February 2026; Dubai opened Dubai Founders HQ with Antler programmes in 2026.
Why it matters: The excuse that building needs a technical co-founder and a lakh of rupees is gone. What remains expensive is attention: yours, and your customers'.
Do this: Before you spend anything, list what you would have paid for a year ago - a developer, a designer, servers, software - and find the node on this map that replaces it with a free tier, a credit programme or an AI tool. Keep that list; it becomes your first budget.
Watch out: Cheap to build also means cheap for everyone else to build. Distribution and a real relationship with customers are the advantages left, which is why the first-users stage is longer than the build stage.
Done when: Your first-year tool budget is written down and is under what you spend on a phone plan, with the credits you will apply for marked.
Building in India: the landscape
India is the world's third-largest startup ecosystem by funding and one of the two largest by count. By 31 March 2026 DPIIT had recognised over 2.23 lakh startups, which together reported more than 23 lakh direct jobs; Tracxn counted around 131 unicorns by mid-2026 and 10.5 billion dollars of tech funding in 2025; the Nasscom-Zinnov report for 2025 found deep tech raising 2.3 billion dollars, up 37 percent, with AI startups taking 91 percent of that. The state is unusually present: recognition, a tax holiday, a seed fund through 300 incubators, fee rebates on IP, self-certification under labour laws, and a 10,000 crore rupee second Fund of Funds approved in February 2026. UPI, Aadhaar-based onboarding and GST invoicing make a digital-first business easier to start here than almost anywhere.
Why it matters: The opportunities and the frictions are both specific. Bharat, the next few hundred million users outside the metros, buys on WhatsApp and UPI and is price-sensitive; enterprise sales cycles are long; GST registration becomes unavoidable once you sell across states or through marketplaces; the talent pool is deep and the competition for it is now global and remote.
Do this: Decide whether your first customers are Indian consumers, Indian businesses, or foreign buyers paying from India (an export business, zero-rated under GST with a Letter of Undertaking). Each has a different payments node, a different pricing logic and a different first-users playbook on this map.
Watch out: Building for India at Silicon Valley prices, or for Silicon Valley from India without ever speaking to a buyer there. Choose the market before the product.
Done when: You have named your first market in one of those three categories and read the India nodes in Make it legal and Ship.
Building in the UAE: the landscape
The UAE sells founders three things: a tax regime that is zero on the first 375,000 dirhams of profit and nine percent above it, with Small Business Relief for companies under three million dirhams of revenue; a licence system of more than forty free zones that lets a foreigner own one hundred percent of a company and get a residence visa with it; and a position between India, Africa and Europe with a buyer base that pays in a hard currency. Abu Dhabi's Hub71 passed four hundred active startups in 2026 with more than four billion dollars raised between them; Dubai's in5 incubator reports its startups have raised over 2.45 billion dollars; Expand North Star, the startup half of GITEX, moves to Expo City Dubai on 8 to 10 December 2026 and is the largest startup gathering in the region. Dubai Founders HQ opened in 2026 and runs its programmes with Antler.
Why it matters: The costs are specific too. A free zone licence runs from roughly 12,500 dirhams a year in the cheaper zones to over 35,000 at DMCC in year one, a business bank account takes weeks and documents, VAT registration becomes mandatory at 375,000 dirhams of taxable turnover, and the domestic market is small, so most UAE startups sell to the GCC or further from day one.
Do this: Decide whether the UAE is your market, your base, or both. A base with customers elsewhere wants the cheapest compliant licence and a bank that handles multi-currency; a company selling to UAE firms and government wants a mainland or DIFC presence and a local network.
Watch out: Choosing a zone by its advert. Visa quota, activity list, office requirement and bank acceptance differ by zone and matter more than the headline fee.
Done when: You know which of base, market or both you are, and have read the UAE nodes in Make it legal, Credits and Incubators.
Building for the world from day one
A global startup picks its customers by problem rather than by passport: the product is in English, priced in dollars, sold over the internet, and the company is usually incorporated where investors and payment providers are comfortable, which in 2026 still means a Delaware C-corporation, a UK limited company or a Singapore private limited company. Stripe Atlas will form a Delaware company for a one-time 500 dollars; Y Combinator's standard deal is 500,000 dollars for seven percent plus an uncapped SAFE; Entrepreneur First and Antler back people before they have a company. The upside is a market of billions and a funding market that understands SAFEs and remote teams; the cost is that you compete with everyone, and a founder in Pune or Dubai pitching to buyers in Austin has to work their time zone and their references.
Why it matters: Global by default is a strategy with real trade-offs: dual compliance, a cross-border tax question the moment you have a company in one country and a founder in another, and an audience you cannot meet for coffee. Chosen deliberately it is powerful; chosen because it sounds bigger it is expensive.
Do this: Pick one wedge market - one country, one segment, one language - for the first hundred customers, even if the company is designed to be global. Decide where to incorporate only after reading Where to incorporate in Make it legal, and after you know where your first buyers and your first investors are.
Watch out: Incorporating abroad before you have revenue there. A US company owned from India creates filings in both places and needs a professional on each side.
Done when: You have named the wedge market, the currency you will price in, and where your first ten buyers live.
Founder-market fit
Founder-market fit is the reason you, specifically, should build this, specifically. It is domain knowledge from a job you had, a problem you suffered, access to the buyers, a distribution channel you already own, or a skill the problem needs that most people lack. Investors use the phrase because it predicts survival: founders with it find the first customers faster, make fewer expensive wrong guesses, and keep going through the year when nothing works. It is not a credential, and it can be built in a few months of deliberate work inside the market.
Why it matters: The idea stage will generate several plausible ideas. Fit is the filter: among equally good problems, the one you have an unfair way into wins.
Do this: For your idea, write one line under each of five headings: what I know about this that outsiders do not; who I can reach who has the problem; what I can build or do that most cannot; why I will still care in year three; what I would need to learn first. If three of the five are blank, either change the idea or spend the next month acquiring them by working inside the problem.
Watch out: Mistaking enthusiasm for fit. Loving an industry from outside is a reason to interview people in it, not yet a reason to build for it.
Done when: The five lines are written, at least three are substantive, and the blank ones have a plan.
Time, money and the day job
Most Indian startups begin as evenings and weekends, and that is a reasonable way to run the idea and validation stages: ten to fifteen focused hours a week is enough to interview twenty people, put up a page and try to pre-sell. What it is not enough for is the first-users stage, which needs you available when customers are. The honest plan is a trigger, not a date: leave the job when a signal appears, such as a paying pilot, a waitlist that converts, or a grant or programme that covers living costs. Until then, protect the runway you have - six months of expenses in the bank is the usual floor - and keep the company's spending near zero, which the credits stage makes possible.
Why it matters: Two things end side projects: a contract and a calendar. Many Indian employment agreements contain IP-assignment and exclusivity clauses that can claim work done on company time or equipment, and notice periods of sixty to ninety days are common, so the leave date is never the day you decide.
Do this: Read your employment contract for IP, moonlighting and non-compete language; do all startup work on your own devices, your own accounts and your own time; write down your runway in months and the signal that would make you go full time; tell one person the date you will review it.
Watch out: Quitting on enthusiasm the week after a good interview. One conversation is not a signal; ten that end in a pre-order are.
Done when: You know your runway in months, your trigger to go full time is written, and your contract has been read for the three clauses.