India’s proposal to develop an indigenous graphics processor within roughly two years is exactly the sort of ambition I want the government to have—and exactly the sort I refuse to applaud without a definition. A GPU can mean an architecture, a licensable design, a fabricated chip, an accelerator board or a commercially deployed ecosystem with compilers and developer support. Those are not interchangeable achievements. Public reporting on the November 4 remarks by the electronics ministry’s senior leadership described an effort under the next phase of the India Semiconductor Mission, alongside plans to commercialize the home-grown SHAKTI and VEGA processor families. Tom’s Hardware summarized the proposal and its still-unanswered implementation questions. I give the government credit for aiming beyond phone assembly. Now it must resist declaring victory at the design slide.
Moving up the value chain is the real reform
“Make in India” electronics has delivered visible scale. Phones assembled here, expanding exports and supplier investments are not imaginary accomplishments, whatever the opposition’s reflexive sneer. A government that creates incentives, negotiates with global manufacturers and builds industrial clusters deserves recognition when factories arrive. But assembly is the beginning of an electronics strategy, not its graduation ceremony. The high-value layers—components, tooling, intellectual property, design software, advanced packaging and manufacturing knowledge—determine strategic depth and durable margins.
The October policy review from PRS Legislative Research records new guidelines for the mobile-phone manufacturing scheme, including turnover thresholds and brand-wise incentive calculations. Policy is becoming more detailed because the easy headline phase is over. India now has to decide what behavior it is subsidizing. If incentives reward only finished-unit volume, firms will rationally import critical value and perform the minimum local transformation. If policy rewards components, engineering, supplier development and exports that survive after subsidies, the ecosystem becomes harder to uproot.
A GPU is not just a chip
The global AI boom has made accelerators seem like magical rectangles. In reality, the hardware’s usefulness depends on software. Developers need compilers, libraries, debuggers, documentation, frameworks and stable performance across real workloads. Nvidia’s advantage is not explained merely by transistor counts; its software ecosystem and long developer relationships create enormous switching costs. An Indian GPU without an accessible software stack may be a valuable research prototype but not a competitive platform.
This is why the two-year target requires careful language. A credible first milestone might be a validated architecture or prototype tailored to government, edge or academic workloads. It need not defeat the most advanced global data-center accelerator. Trying to imitate the frontier immediately would waste money and invite a theatrical failure. India should choose workloads where sovereignty, cost and customization matter: public-sector inference, language computing, industrial vision, defence applications or university research. Narrow success can create engineers and tools that support broader ambition.
Chinese brands still teach the uncomfortable lesson
Indian consumers continue to buy Chinese smartphone brands because those companies combine aggressive pricing, supply-chain speed, distribution, financing and features tuned to the market. National slogans do not erase a five-thousand-rupee difference or produce a better service network. This does not mean policy has failed. It means industrial capability is cumulative. Chinese firms benefit from component clusters and manufacturing ecosystems built over decades. India cannot wish that advantage away; it has to reproduce selected parts of it while developing its own strengths.
The lesson for an Indian GPU is similar. A patriotic chip with weak tools, uncertain supply and a high total cost will remain a procurement symbol. Users adopt platforms that solve problems. Government can act as an early buyer, but procurement should include performance and software milestones rather than become a sanctuary from comparison. The goal is not to protect a chip from the market forever. It is to help a capability survive long enough to meet one.
Government deserves credit, and scrutiny
I reject the fashionable view that every semiconductor announcement is automatically vanity. Advanced economies are spending public money to secure chip supply chains. India would be foolish to remain a passive market while compute becomes critical infrastructure. The Modi government has correctly identified electronics and semiconductors as strategic manufacturing sectors, used incentives to attract investment and made states compete for projects. That is a meaningful shift from the Congress-era comfort with services growth and imported hardware.
Credit, however, should purchase scrutiny rather than silence it. The government should publish a milestone ladder for the GPU program: architecture selection, team formation, toolchain release, tape-out, fabrication node, packaging, testing, pilot deployment and independent benchmarking. It should identify which parts are Indian-owned, locally designed, locally manufactured or imported. These categories can coexist, but they should not be blurred. A chip fabricated abroad from an Indian design may be a major design achievement. Calling it fully indigenous would still be misleading.
Talent policy is industrial policy
India has excellent chip designers, many working for multinational firms domestically and abroad. The challenge is not proving Indian engineers can design semiconductors. It is creating institutions where they can build products, retain intellectual property and learn through repeated fabrication cycles. Chip development punishes one-off project culture. Teams need stable funding, realistic deadlines and the ability to learn from an unsuccessful tape-out without a parliamentary scandal ending the program.
Universities should receive access to design tools, prototype fabrication and shared testing infrastructure. Public research processors such as SHAKTI and VEGA should become teaching and commercial platforms rather than permanent demonstration pieces. Start-ups need procurement pathways and patient capital, but they also need honest technical review. Subsidizing every company with “AI” in its presentation will produce expensive brochures. Concentrating resources on verifiable engineering will produce fewer announcements and more capability.
Assembly numbers should become supplier numbers
For the broader electronics push, I want the government to publish measures that are harder to game: domestic value addition, local component categories, Indian engineering payrolls, supplier certifications, export diversity and the share of incentives paid after sustained production. Unit output matters, but it can conceal shallow integration. A phone assembled from imported displays, memory, processors, camera modules and machinery is still useful industrial activity; it trains workers and creates logistics. It should not be described as the destination.
Component policy must also avoid creating permanently protected high-cost inputs. Indian device makers should not be forced to use inferior components simply to satisfy a local-content slogan. Support should be time-bound, performance-linked and export-oriented. The best evidence that a component ecosystem is competitive is that firms sell beyond a captive domestic mandate.
The counterargument: two years creates urgency
Officials may say an aggressive deadline mobilizes institutions. They are right. Indian public projects can drift when goals are vague. A two-year target can force decisions, attract talent and prevent endless feasibility studies. My objection is not to urgency but to undefined completion. If “we have our own GPU” can be declared after any laboratory prototype, the deadline becomes public relations. If milestones are explicit, even a delay can yield knowledge.
India should also collaborate internationally without treating partnership as dependence. Licenses, foundry access, packaging expertise and joint research can accelerate learning. Strategic autonomy does not mean reinventing every tool domestically. It means retaining enough knowledge, alternatives and bargaining power that a foreign restriction cannot immobilize critical systems.
Build the stack, then celebrate
I am more optimistic about India’s electronics trajectory than many left-leaning critics who treat industrial incentives as corporate gifts by definition. Factories and supply chains do not materialize from moral suspicion. The government’s willingness to spend, coordinate and set strategic priorities is a reform worth defending. The next stage is harder because it depends less on ribbon-cutting and more on engineering institutions.
An Indian GPU would be a significant achievement if it is defined honestly and supported by usable software. It would be even more significant if the program leaves behind skilled teams, open tools, university capacity and companies capable of designing the next chip. That is how industrial power compounds. A single patriotic launch followed by procurement orders would not be enough.
So yes, aim for two years. Publish what happens at six, twelve, eighteen and twenty-four months. Benchmark the result openly. Tell us which parts were designed here, fabricated elsewhere and packaged locally. Give the government credit when those milestones are met and criticism when slogans outrun silicon. “Make in India” becomes durable only when we can describe precisely what India has learned to make.
