India Should Stop Copying the Biggest Models

India does not need to win an expensive race to build the largest artificial intelligence model. It needs systems that solve Indian problems, protect Indian data and work at a price that schools, hospitals, farms and small businesses can afford. That is the central argument in a Hindustan Times opinion by technology writer Vivek Wadhwa. He says the next phase will not belong only to giant language models. Smaller models, specialised tools, reasoning systems, edge computing and software agents may create more practical value. This is an analysis, not a settled forecast. Large models will continue to matter. But India should ask a harder question before spending public money. What useful outcome will the model deliver? 

Bigger Is Not Always Better

The first phase of generative AI rewarded scale. Companies trained larger systems on more data and used more computing power. The results were impressive. Models could write, translate, summarise and produce computer code. The weaknesses are also clear. These systems can invent facts, miss context and give confident answers without reliable evidence. Making a model larger may improve some tests, but it also raises the cost of chips, data centres, electricity and skilled staff. Satya Nadella has argued that customers should have a choice of models instead of depending on a few suppliers. The opinion article points to Microsoft’s interest in cheaper Chinese systems as evidence that businesses may increasingly select different models for different tasks. DeepSeek strengthened that argument by showing how lower-cost systems could challenge assumptions about the money and hardware needed for capable AI. It did not end the race for large models. It changed the economics of the debate. The lesson for India is not to copy China or the United States. It is to avoid spending years chasing a target that may become cheaper and easier to access before an Indian rival is ready.

Build for Indian Needs

AI in India will be useful only when it understands the country beyond standard English prompts. It must handle Hindi, Tamil, Bengali, Marathi and dozens of other languages. It must also recognise regional accents, mixed-language speech and local administrative terms. Data matters here. A foreign system may translate a sentence correctly but still miss the legal, medical or social meaning behind it. A tool used in a district hospital must understand Indian records and referral practices. A farming assistant must account for local crops, weather and mandi prices. The IndiaAI Mission has called for indigenous foundation models trained on Indian datasets. It has also made more than 18,000 graphics processing units available through cloud providers, according to official programme material. That public investment should support a range of systems. India may need some large national models for research, security and strategic independence. It also needs many smaller models built for specific sectors. A school does not need the world’s most expensive chatbot. It needs a reliable tutor that follows the curriculum, works on a basic device and explains a lesson in the student’s language. A primary health centre needs a tool that helps staff organise records and flag risk. It does not need a system that writes poems in fifty styles. Simple can be powerful.

Own the Data and the Delivery Layer

The opinion article makes another important point. The company that builds the base model may not capture most of the long-term value. The larger opportunity may sit in the layer around the model. That includes trusted data, security, workflow software, model selection, payments, audits and customer support. India already has experience building digital systems at scale. Aadhaar, UPI and DigiLocker show how common infrastructure can support public and private services. AI could follow a similar path if the rules protect rights and allow different providers to compete. OpenAI and other global companies will remain useful partners. India should not block foreign tools simply because they were built elsewhere. It should decide where they can operate, what data they may process and how users can challenge harmful errors. Sensitive government, health and court data should not move into an overseas service without clear safeguards. A cheaper tool is not cheaper if India loses control of the information feeding it. This is where a secure national cloud and model marketplace could help. A routine translation task may use a low-cost model. A complex scientific query may need a stronger one. Sensitive work may require an Indian model running on local infrastructure. The user should not have to understand the technical routing. The platform should choose safely and record what it did.

Indian IT Firms Face a Real Choice

India’s software services industry built its success by supplying skilled workers and managing large business systems. Generative AI threatens part of that model because companies may need fewer hours for coding, testing and basic support. The response cannot be denial. Indian firms can become implementation partners that help banks, manufacturers, hospitals and governments redesign work around AI. They understand old software, compliance and daily business processes. That knowledge is valuable. I have seen companies buy an expensive tool before deciding what problem it should solve. Months later, employees return to spreadsheets because the new system does not fit their work. Indian technology firms can prevent that waste. They can connect models to real data, train staff, test results and build checks for mistakes. The work will be less about selling headcount and more about delivering outcomes. That shift will be difficult. It may also create a stronger industry.

For The United Indian

Why This Matters

India should not measure progress by the size of one model. It should measure whether AI lowers costs, improves public services and creates products that work for Indian users.

The Bigger Picture

The next opportunity may belong to smaller systems, trusted data and useful applications. India has the engineers and digital infrastructure to compete there.

The country does not need to arrive late at another technology race. It needs to choose the race worth running.