Quick answer
Economic Survey 2025-26 models 24% baseline and 32% high domestic demand growth; easier finance does not remove persistent GPU supply bottlenecks.
Compute-demand scenarios are conditional stress tests of how infrastructure responds when demand, financing and hardware access change. Charts XIV.3–5 show that strong demand does not guarantee capacity expansion: bottlenecks migrate, but global GPU access remains a recurring constraint.
Concept diagram
Plan for more than one demand path
Three unscaled scenario dials distinguish demand sources without implying quantified forecasts.
- Baseline demand differs from demand linked to foreign and domestic users.
- Capacity planning should adapt as actual usage and constraints become clearer.
Basics: assumptions are not predictions
Demand and operational capacity
- Demand is the desired use of compute; operational capacity is the infrastructure actually available to deliver it.
- Both training and inference create demand, but additional spending cannot instantly manufacture GPUs or accelerate grid construction.
Borrowing cost and hurdle rate
- A borrowing rate is the cost of debt. An investor hurdle rate is the minimum required return used to judge a project.
- The baseline 9% borrowing cost must not be confused with the baseline 15% investor hurdle rate.
Box XIV.2: scenario assumptions, not observed national outcomes
| Input | Baseline | High foreign GPU demand | High domestic demand + liberal finance |
|---|---|---|---|
| Simulation horizon | 40 quarters | 40 quarters | 40 quarters |
| Domestic demand growth | 24% annually | Baseline unchanged | 32% annually |
| India share of global GPU demand | 4% | Baseline unchanged | 6% |
| Investor hurdle rate | 15% | Baseline unchanged | 10% |
| Baseline debt cost | 9% annually | Other factors unchanged | Liberal financing; no new debt rate stated |
| Starting power tariff | ₹8/kWh | Other factors unchanged | Other factors unchanged |
| New grid-capacity lag | 12 quarters | Other factors unchanged | Other factors unchanged |
| Foreign demand | Normal conditions | Almost double baseline growth pace | Returns to baseline |
Charts XIV.3–XIV.5: documented simulation results
| Survey chart | Early-stage result | Persistent bottleneck | Policy reading |
|---|---|---|---|
| XIV.3 — baseline | Finance initially filters expansion | GPU waiting dominates much of horizon; grid episodic | Demand is not the binding constraint |
| XIV.4 — high foreign demand | Costlier GPUs weaken project economics and financing | Delivery uncertainty survives financial recovery | Domestic credit easing alone cannot secure hardware |
| XIV.5 — high domestic demand | Financing wait is two quarters shorter than baseline | Firms reach GPU bottleneck faster; grid intermittent | Liquidity does not ensure proportional capacity growth |
How to read the three scenario charts
Source and limits
- Source: Economic Survey 2025-26, Box XIV.2, Charts XIV.3–5; author’s calculations. The charts track counts of operators waiting for finance, GPUs or grid access.
- The table preserves the results stated in the text. It does not fabricate quarter-by-quarter operator counts from plotted lines.
Feedback mechanism
- Higher foreign GPU demand raises upfront capital costs, compresses margins and weakens debt service capacity, extending financing stress.
- Even when improved revenues ease finance, probabilistic supply and delivery delays can leave viable projects unbuilt.
Policy sequence implied by the stress test
Diagnose the binding input
Distinguish financing delays from hardware scarcity and grid readiness.
Coordinate and diversify
Combine finance with diversified access to global compute and infrastructure planning.
Build longer-term capability
Develop domestic semiconductor capability over time without assuming immediate advanced-chip self-sufficiency.
Common exam traps
Statements examiners like to twist — learn the correct version.
32% is India’s realised compute growth.
It is an assumed annual domestic demand growth rate in scenario 3.
Cheaper finance eliminates all queues.
It reduces financing delays but accelerates exposure to hardware constraints.
Foreign demand is exactly 48% in scenario 2.
The text says almost double the baseline pace; it does not specify an exact rate.
Prelims focus
- Memorise 24% versus 32%, 4% versus 6%, and 15% versus 10%.
- Know the difference between 9% debt cost and investor hurdle rates.
Mains angle
- Use the three scenarios to explain complementary-input bottlenecks.
- Assess the limits of subsidy-led AI infrastructure policy under external supply dependence.
Flip & recall
Answer in your head first, then tap the card to check.
Practice MCQs
Original Prelims-style questions with explanations.
0 / 3 correct · 3 left
Q1.The baseline borrowing rate and investor hurdle rate are respectively:
Q2.Scenario 3 assumes annual domestic demand growth of:
Q3.Across scenarios, the recurring long-run constraint is:
