Gujarat targets ₹17 lakh crore in data-centre investments — Dholera as anchor

What happened

Gujarat has set a ₹17 lakh crore data-centre investment target, with Dholera positioned as an anchor site — the September 2026 figure rising from the ₹6 lakh crore baseline set when the Viksit Data Centre Policy 2026-29 launched in July, and following the L&T Vyoma MoU (Feb 2026) that first put a named data-centre project on Dholera's map.

The background

The escalation from ₹6 lakh crore to ₹17 lakh crore in roughly two months reflects the speed at which the AI compute build-out has moved from policy ambition to boardroom pipelines. When Gujarat launched its first dedicated data-centre policy in July 2026 — targeting 7.5 GW of capacity with power, land and tax incentives — some 20 firms including Meta, Reliance and Adani were already expressing interest. By September, the state's stated ambition had nearly tripled. Dholera's role inside that ambition is specific: it is the site where the state has greenfield land at scale, a planned power architecture sized for heavy industrial loads, and the digital trunk infrastructure already in the ground plan.

The figure deserves context. ₹17 lakh crore is an investment target — the total capital expenditure the state expects data-centre developers to commit over the policy period if every expression of interest converts into a built facility. India's entire data-centre capacity today is a fraction of what that level of investment would build, which is why the number reads as aspirational. But the trajectory of AI compute demand globally — hyperscale campuses under construction across Asia, capacity constraints in established markets like Singapore and Ireland — gives the ambition a backdrop that earlier data-centre projections did not have.

The details

Why data centres and Dholera fit each other is a matter of inputs. Hyperscale AI campuses are essentially power converters — their economics live and die on electricity price and reliability — and Dholera's design treats power as a trunk utility, planned into the activation area alongside water and roads, with a renewable base including the Dholera Solar Park adjacent to the SIR. Add fibre routes planned into the digital backbone, and land that a congested metro cannot offer at any price, and the anchor-site label has substance behind it.

The specific economics of AI-era data centres make land at scale more important than for earlier generations. Training large AI models requires clusters of GPUs concentrated in single facilities — tens of thousands of chips in one building, drawing tens of megawatts — because the interconnects between GPUs are bandwidth-limited over distance. That means a single hyperscale AI campus needs more land, more power and more cooling than a conventional cloud data centre. Dholera's greenfield plots — large, serviced and planned for industrial loads — are exactly the product hyperscale developers need.

Why it matters to Dholera

Why it matters to Dholera: compute is the newest utility. If even a fraction of the target lands in Dholera, the region gains the power-hungry anchor tenants that justify every watt and fibre route in the trunk plan. Data centres also diversify the demand base — a region dependent on a single fab is concentrated risk; fab plus data centres plus airport-driven logistics is a portfolio.

The diversification matters for the investment architecture. A semiconductor fab has a long construction tail and a concentrated customer base; a data-centre campus has a shorter build cycle and a diversified tenant base. Together, they generate demand for the same trunk infrastructure — power, water, fibre — at different times and for different reasons, which smooths the utilisation curve of the public investments that underpin the city.

The bigger picture

The discipline in reading this number is the same discipline described in our investment-architecture chapter: announced is not the same as existing. A target is a record of intent — the honest milestones are power contracts signed, campuses grounded and capacity energised. What the September 2026 figure does confirm is that the policy launched in July is generating escalations rather than revisions — the direction is upward, which is what a policy designed to attract capital is supposed to achieve.

Label: policy targets are ambitions, not bookings — tracked at 🟠 until sites and construction appear.

Related chapter: Ch. 18 — Investment & economic architecture →