Why India needs longer batteries | The patient investor runs out of patience

ELI5/TLDR

India is finally building battery storage for its grid, but the goalposts keep moving: the grid no longer needs bigger batteries, it needs longer-lasting ones, because solar floods the day and vanishes at sunset just as everyone gets home. Meanwhile, in America, the institutions built to be the most patient investors on Earth — public pension funds — have been dumping their private-equity holdings at fire-sale prices. The reason is a slow-motion accounting trap: chasing optimistic returns pushed them into opaque assets that hid how underfunded they really were. Two stories, same theme — a system whose needs keep outrunning the fix it bought.

The Full Story

The waste problem nobody can store

Between May and December last year, India threw away 2.3 terawatt-hours of solar electricity. Not because it was broken — because the grid had nowhere to put it. Nearly 40% of that waste came in October alone, when the sun was strong, demand was soft, and the coal plants sitting underneath the system couldn’t power down fast enough to make room.

The obvious fix is storage, and India has started. It added 4.6 gigawatt-hours of battery storage in the first three months of 2026 — more than nine times the previous quarter. Impressive, until you see the target.

The central electricity authority reckons India will need 236 gigawatt-hours of batteries by 2031-32. So we have built in our entire history a little over 2% of the batteries we expect to reach within 7 years.

Two numbers you have to hold at once

Here’s the shift worth understanding. To follow it, keep two numbers in your head.

Power (measured in megawatts) is how much electricity a battery can push out at any single instant. Energy (measured in megawatt-hours) is how long it can keep pushing before it empties. Divide energy by power and you get duration.

Think of it like a water tank with a tap. The tap’s width is power — how fast water comes out. The tank’s size is energy — how much is in there. A wide tap on a small tank empties fast; a narrow tap on a big tank trickles for hours. A 500 MW / 1,000 MWh battery delivers full power for 2 hours. Give it the same tap but a tank twice as big — 2,000 MWh — and it becomes a 4-hour battery.

India’s grid has quietly stopped asking how many megawatts of battery do we need and started asking how many hours of stored sunshine do we need.

Why the evening keeps getting longer

Solar’s problem is the clock. Through the day the sun does the heavy lifting, so India buys cheap solar and idles its coal plants down to a minimum. But after sunset, solar falls to nothing within an hour — right as people get home and switch on lights, fans, ACs and stoves. Demand spikes exactly when the cheapest supply disappears, and the half-asleep coal plants struggle to ramp back up.

A 2-hour battery solves this tidily: charge during the cheap sunny glut, discharge into the evening crunch. That was the logic behind SECI’s landmark March 2022 tender — 500 MW for 2 hours. But a 2-hour battery is a peak shaver only. It clips the very tip of the evening spike and is empty after that.

And the spike has stopped being a spike. As solar crossed 100 gigawatts, the evening peak became an elevated zone — roughly 5pm to 11pm in winter, 7pm to past midnight in summer. The more India leans on solar, the less coal carries the load, and the longer the crunch runs. A 2-hour battery can’t cover it. This is why a new piece of jargon, FDRE — firm and dispatchable renewable energy — keeps coming up: contracts where buyers pay for green power delivered in specific scheduled hours, not a vague annual quantity. You can’t make that promise with a 2-hour battery. Every FDRE auction mechanically drags developers toward 4 hours and beyond. A 2025 study by the Power Foundation of India and UC Berkeley pegs 2027 as the year 4-hour batteries become the default design.

Why doubling the duration barely raised the price

Here’s the counter-intuitive bit. Doubling the storage should roughly double the cost — twice the cells. On paper it didn’t.

To read tender data, you subtract a plausible solar-only price (say ₹2.5 per unit) from the winning bid; what’s left is the “storage adder.” A July 2024 auction with a 2-hour battery cleared at ₹3.41 (adder ≈ ₹0.81). A December 2024 auction with a 4-hour battery cleared at ₹3.52 (adder ≈ ₹1.00). Doubling the duration raised the final tariff by barely ₹0.19 a unit. Cratering battery prices, co-locating storage with solar (shared land and grid hookups), and government viability-gap funding did most of the work.

But that flat per-unit number hides two strains. First, the developer carries twice the cells, twice the capex, and a longer wait to earn it back — at tariffs already razor-thin. The result is a pipeline that won’t build: roughly 130 GWh of projects were awarded in 2025 against barely 9 GWh expected to physically switch on in 2026. Second, 4-hour storage needs twice the cells, which turns the whole thing into a much bigger bet on India making cells at home — scaling from ~13 GW today toward 100 GW by 2027.

The battery that does five jobs and gets paid for one

A battery can, in principle, earn money several ways at once: arbitrage (buy cheap midday power, sell dear in the evening), ancillary services (split-second tweaks that keep grid frequency steady), and capacity payments (getting paid just for standing ready). Letting one asset earn from several jobs is called revenue stacking. India mostly doesn’t allow it. A storage project here typically lives off a single contract with SECI or a state discom — no capacity market, thin time-of-day price gaps, so arbitrage alone doesn’t cover costs.

According to an analysis, BESS projects would make 24% to 70% margins if such revenue stacking was made possible.

The battery does multiple jobs for the grid but gets paid for one of them.

The wall beyond four hours

Even 4 hours doesn’t finish the job — it covers the evening but can’t turn midday sun into genuine nighttime power. As solar’s share climbs, the task becomes shifting bulk energy 6, 8, even 10 hours forward. And lithium-ion stays cheapest only out to roughly 4-6 hours of daily discharge. Its weakness is structural: power and energy are welded into the same cell, so the only way to add hours is to buy more whole cells — fine at 4 hours, absurd at 10.

This is where flow batteries get interesting. The energy doesn’t sit in solid cells; it sits in tanks of liquid electrolyte pumped through a separate stack. Power and energy are pulled apart — the stack sets the power, the tanks set the hours. Want 10 hours instead of 4? Add cheap electrolyte and bigger tanks, keep the same stack. So each extra hour gets cheaper, which is exactly why flow’s case lives at the long end. At the 2-4 hour durations India actually buys today, flow is more expensive and less efficient — a handful of pilots (NTPC’s 3 MWh vanadium flow unit, a 100 MWh tender largely built by Electric, the country’s only flow maker at scale, backed by Zerodha’s Rain Matter climate fund).

The tension underneath: the cheaper lithium gets, the longer a “good enough” 4- or 6-hour lithium battery delays the jump to true long-duration storage. Falling prices may not be solving India’s storage problem so much as postponing the harder version of it.

The patient investor who stopped being patient

Switch continents. American public pension funds are, in theory, the perfect patient capital — they know exactly how much they’ll owe and when, decades in advance. They should never have to sell in a hurry. Yet since 2022 they’ve been dumping their private-equity holdings at big discounts, as if fleeing a fire. None of those investments had visibly failed.

The trigger was partly mechanical: stocks fell sharply in 2021-22, so the funds’ public holdings shrank, leaving them overweight in illiquid private markets — they sold to rebalance. But underneath sits a decades-long drift, and an accounting trap.

The logic of a pension is symmetry: workers lock money away, it goes into long-dated assets, and decades later it becomes their income. The old American defined-benefit plan made hard promises backed by estimates of what the portfolio would earn. In the 1980s that was easy — government bonds briefly paid over 15% for almost no risk, and funds held ~40% in fixed income. Then rates fell, and the hunt for yield began.

Here’s the trap. A defined-benefit fund fixes what it owes in the future, then works backwards to figure how much it needs today — using an assumed return. Crucially, the books don’t have to show the risk taken to hit that return.

So imagine you had to pay 1,000 rupees 20 years from now. If you put your assets into bonds that would definitely yield 3% a year, you would need ₹554 in assets today. If you put it in real estate and private equity, that might earn 7% a year and you’ll only need ₹258.

Optimism becomes an accounting choice. Assume a higher return and — on paper — you need less money today. After the dot-com crash blew holes in equity-heavy funds, the rational move for a strategist trying to balance the equation was to take more risk, not less. The 2008 crisis sealed it: funding ratios collapsed, risk-free rates sat near zero for a decade, and cutting benefits or raising contributions was politically toxic. So funds reached for alternatives — private equity, private credit (which boomed as banks retreated from mid-market lending). By 2010, public pension funds assumed ~2% higher annual returns than corporate funds investing in the same markets, with nearly three-quarters of their portfolios in risky assets.

The returns weren’t actually better — PE roughly matched public markets, and fees ate the difference. So why the stampede? Partly herd behaviour (“monkey see, monkey do” — every 10% peers added to alternatives pulled a fund up ~5%). But the biggest draw was opacity. Public stocks reprice every second, adding visible volatility. A private stake is marked periodically by models and manager judgment — its real value unknowable, but on paper it barely moves. To a pension fund it looked like an extremely stable asset with equity-like returns. An illusion built on how rarely the price was calculated.

Meanwhile the demographics turned. Contributors aged, retirees multiplied, and by 2023 the average public fund had gone net cash-negative, bleeding ~2.4% of assets a year. Those outflows need real cash — but cash and liquid securities were just 9% of holdings. So when COVID-19 made stocks and bonds fall together, the private holdings (which hadn’t marked down) suddenly looked exaggeratedly large in the mix. The sell-off was a release valve.

The structural problem remains: the world’s biggest pools of capital are underfunded — by some estimates $4.6-6 trillion short if properly accounted — with the gap hidden under opaque assets. When payments come due, they’ll likely sell the liquid stuff instead, pushing losses onto wider markets that everyone holds.

Institutions built to hold safe assets indefinitely have morphed into those that might sell opaque assets in a panic.

Tidbits

India extended its deep-water oil and gas bidding deadline for the sixth time since February 2025, trying to lure foreign investors as the Strait of Hormuz crisis sharpens the case for domestic production. Moody’s flagged India’s heavy credit exposure to water mismanagement — the riskiest rating (5) on water, and a 4 on physical climate risk for heat stress and erratic monsoons. And Hindustan Zinc signed an MOU to explore hydrogen fuel in underground mining — by its own claim, a first — as part of a net-zero-by-2050 push.

Key Takeaways

India wasted 2.3 TWh of solar between May and December last year because the grid had no storage and coal couldn’t ramp down fast enough; ~40% of that waste was in October alone.
India’s cumulative battery storage hit 5.9 GWh by March 2026 — about 2% of the 236 GWh the CEA says it needs by 2031-32.
Power (MW) = how fast a battery discharges; Energy (MWh) = total stored; duration = energy ÷ power. Same battery can be “2-hour” or “4-hour” depending on tank size relative to tap width.
The grid’s design question shifted from “how many megawatts” to “how many hours” of storage — the key conceptual change in the last 18 months.
Solar collapses within an hour of sunset just as residential demand peaks; the more solar India adds, the longer the evening crunch gets (now a 5pm-midnight zone, not a single spike).
A 2-hour battery is only a “peak shaver.” Covering the full evening zone needs 4+ hours.
FDRE (firm and dispatchable renewable energy) contracts pay for power in specific scheduled hours, which mechanically forces developers toward 4-hour-plus batteries.
Doubling duration from 2 to 4 hours raised tariffs by only ~₹0.19/unit (storage adder went from ~₹0.81 to ~₹1.00) thanks to collapsing battery prices, co-location with solar, and viability-gap funding.
The pipeline is stuck on construction, not awards: ~130 GWh awarded in 2025 vs ~9 GWh expected to switch on in 2026 — thin tariffs plus a bigger asset put developers in the red.
4-hour storage needs twice the cells, raising the stakes on India’s domestic cell capacity scaling from ~13 GW today to a 100 GW target by 2027.
Revenue stacking (one battery earning from arbitrage + ancillary services + capacity payments) is largely banned in India; allowing it could lift BESS project margins to 24-70%.
Lithium-ion is cheapest only to ~4-6 hours because power and energy are welded in the same cell; adding hours means buying more whole cells.
Flow batteries separate power (the stack) from energy (liquid electrolyte tanks), so extra hours get cheaper — economical only at the long end (8-10+ hours), still a pilot-stage tech in India.
Cheap lithium may be postponing India’s long-duration storage problem rather than solving it.
US public pension funds hold ~$7 trillion; since 2022 they’ve sold PE holdings at discounts despite no investments visibly failing.
The accounting trap: defined-benefit funds can assume a higher return to need less cash today, without showing the risk on their books — turning optimism into an accounting choice.
Each crisis (dot-com, 2008, COVID) pushed funds toward more risk, because under-funded books made optimistic assumptions the only way to balance the equation.
PE returns roughly matched public markets but fees ate the edge; the real draw was opacity — infrequent, model-based marking that hid volatility.
By 2023 the average US public pension fund was net cash-negative (bleeding ~2.4%/year) with only 9% in liquid assets.
Estimated underfunding of these pools is ~$4.6-6 trillion if properly accounted; when cash is needed they’ll sell liquid assets, pushing losses onto markets everyone holds.

Claude’s Take

The battery segment is the stronger of the two — it takes a story everyone thinks they understand (“India needs more batteries”) and reframes it around a single sharp insight: the metric changed from megawatts to hours, and the grid’s needs are outrunning the hardware on a treadmill. The water-tank framing of power-vs-energy is doing real work here; once you have it, the whole FDRE-pushes-toward-4-hours logic clicks. The flow-battery-postpones-the-real-problem ending is genuinely good — it resists the tidy “and so India is solving it” bow that most explainers reach for.

The pension segment is more ambitious and slightly more familiar — the “private markets hide volatility through infrequent marking” critique has been making the rounds (Cliff Asness coined “volatility laundering” for exactly this). But the framing of optimism-as-an-accounting-choice, and the demographic turn to net-cash-negative as the thing that forces the panic selling, is a clean causal chain that most coverage doesn’t bother to assemble. The numbers are cited to named economists, which is more rigour than the format usually offers.

Marking it 8/10. Both stories teach a transferable mental model rather than just narrating news, the sourcing is specific, and the writing earns its length. It loses a point because the pension half leans on a thesis (the underfunding crisis cascading to everyone’s portfolio) that’s plausible but unfalsifiable on this timescale, and the transcript’s number-mangling (terawatt/gigawatt typos, garbled tariffs) means you’d want to verify the figures before quoting them anywhere that matters.

Why India needs longer batteries | The patient investor runs out of patience | The Daily Brief #493