Running a solar business in India in 2026, the trends worth tracking

By Nabeel Tauheed · 8 September 2026 · 7 min read · Market

In short

  • Loan application and loan sanction, not subsidy or grid connection, hold the largest share of one Delhi-NCR EPC's active pipeline between them. Loan sanction alone is consistently the slowest stage in that pipeline.
  • DISCOM paperwork is not portable. Net-meter files at that same EPC are consistently one of the slower stages too, and a template accepted by one board can be rejected by the next.
  • Cash collection before material dispatch decides the month. The EPC's monthly collections swing considerably from month to month, and the gap tracks how much cash arrived before material moved, not how many deals closed.
  • PM Surya Ghar's subsidy is filed at commissioning, on a schedule, so a project field that carries the application deadline matters more than a reminder that relies on someone remembering.
  • An AI telecaller booked more site visits per 100 dials than any of three human telecallers at that EPC (6.5 versus 1.0 to 1.7), while operations headcount has to grow with the project count regardless.

Running a solar business in India in 2026 means living with the bank's timeline, not the customer's. Government subsidy under PM Surya Ghar is now a formal national programme rather than a negotiated discount, which removes one source of delay. What replaces it is the loan stage. In stage-exit data from a Delhi-NCR EPC, since May 2026, loan sanction is consistently the slowest stage in the pipeline. That is the new rhythm. Missing it costs cash flow.

Loan sanction is the first bottleneck for a solar business in India

At that Delhi-NCR EPC, running a large book of active projects, loan application and loan sanction together hold the largest share of the active pipeline, more than any other pair of stages. Loan sanction itself, measured on stage exits since May 2026, is consistently the slowest single stage. The project stalls until money lands. Your operations head gets asked for a status update, and nothing moves until the bank does.

Separate the loan preparation stage, which is yours, from the loan sanction stage, which is the bank's, so a status report shows what is happening inside your four walls versus what is sitting inside theirs. Track loan time by lender if more than one bank works with your customers, so you can see whether one is reliably slower than another. That comparison is worth building once you have enough projects through each bank to trust it.

DISCOM paperwork does not travel between boards

Delhi has BRPL, BYPL and TPDDL. Haryana has DHBVN and UHBVN. Uttar Pradesh has PVVNL and NPCL, and both split further by circle. A solar business working across all three states is really working seven separate DISCOMs, each with its own net-meter form. Running net-meter files across these boards for a Delhi-NCR EPC, the same paperwork accepted by BRPL has come back rejected from PVVNL over formatting neither board writes down anywhere a customer can check. The net-meter stage in that EPC's own history since May 2026 is consistently one of the slower stages too — and that is DISCOM turnaround, not a step the EPC itself controls.

A project for a BRPL customer looks routine until documentation ops reaches for the PVVNL form instead and the DISCOM sends it back. That is a week gone before the resubmission is even in. Before a project moves to net-meter, check the template against what that specific DISCOM currently accepts, not what worked on the last project. Track which DISCOM each project falls under, because the checklist that runs on it (see net-metering timelines by DISCOM) has to branch on that field, not guess it.

Cash collection before material dispatch

A Delhi-NCR EPC's monthly collections swung considerably from month to month, even while deals won moved in a narrow band across that same window. The difference is not the instalment schedule. It is whether the customer had the cash on hand when the instalment came due.

A payment gate that will not let material move without the agreed share collected removes the argument entirely. On a loan-financed project that share is typically 65% before material dispatch; on a cash project it is 95%, both before material ships and 100% before net-meter activation, on the defaults this EPC ran. Operations cannot be talked past the gate because a salesperson promised it or a customer asked nicely. When a cheque bounces, surface it fast. If it is due on the 15th and you discover the bounce on the 22nd, material has already gone to the site; the version that works flags it on the 16th, before the material ships.

PM Surya Ghar's subsidy runs on a filed application, not a promise

PM Surya Ghar (pmsuryaghar.gov.in) put rooftop subsidy on a national portal with a defined application process rather than the state-by-state negotiation it replaced. That is good for predictability, and it means the subsidy is only as reliable as the paperwork behind it: filed correctly, on the right sequence relative to installation, or not paid at all. How to apply for it, step by step covers the sequence in full.

The failure mode is not the amount, it is the timing. A project that reaches the subsidy stage with no one tracking its application deadline can sail past it. Recovery means asking the customer to absorb the gap after the fact, a harder conversation than the one you could have had during the sale. A project field that carries the deadline, checked the moment a project enters the subsidy stage, catches this before it becomes a customer conversation. A whiteboard note does not survive a slow week.

Task velocity tells you where to hire next

At that same Delhi-NCR EPC, the checklist engine had generated nearly every task on record automatically, against only a handful created by hand. But velocity is not flat across the pipeline. Measured on stage exits since May 2026, doc collection clears quickly, typically within a few days. Loan sanction, the same stage covered above, is by far the slowest. A project can move through documentation in days and then sit in loan sanction for two months, and the two delays look identical on a whiteboard that only tracks "in progress".

Track task closure by stage and by assignee, in addition to tracking it by project. Know whether documentation is closing on time or quietly slipping. Know that loan sanction is consistently the slowest stage, so a project already well past your own median is worth a look, and one running deep into the slowest quarter is not merely unlucky. That distinction tells you which stage needs a second person, which needs a process change, and which is blocked by the bank or the DISCOM and needs forecasting rather than fixing.

AI calling and team composition

At the same EPC, an AI telecaller live since 27 May 2026 booked 322 site visits from 4,922 dials, 6.5 per 100 dials, in the months since. Three human telecallers booked 110 to 168 visits each from 10,040 to 10,682 dials over a longer window, 1.0 to 1.7 per 100 dials. The AI works fresh Meta ad leads, and a lead that used to wait for a callback now gets dialled within five minutes. This is not the AI replacing the telecallers. It is filling a queue that used to sit.

The sales team at that EPC grew by one seat later in the year, a genuine hire rather than an import artefact, and it now works alongside the AI caller rather than instead of it. The AI's efficiency is measured against ad leads; the human team's against purchased lists and referrals. Those are different jobs, and the comparison only holds if you keep the two apart.

Operations scales non-linearly

At that Delhi-NCR EPC, operations headcount grew by one when a third documentation-ops hire was added; the sales team's own hire came later, so the two moves were not simultaneous. What connects them is that operations load does not track headcount in a straight line. The project pipeline seeds up to 57 checklist items per project across 17 project stages, branching on whether the project has a loan, a subsidy, and which DISCOM it sits under. With a large book of active projects running at once, that is thousands of open tasks at any moment, each with a due date and each capable of going overdue.

Every task one day late moves up the org chart overnight, which means someone reads it and decides what happens next. Three projects going late on the same day means three of those decisions land before the operations head has had coffee. You know whether your team has the capacity for that because you are tracking closure by person and by stage, not because you are guessing. Hire while there is still time to train someone properly; hiring after the team is already drowning means training gets skipped, because nobody training a new person has time to do it well. Forecast the load a few months out from how many projects sit in which stage now, since that tells you what is coming before it arrives.

What to measure

The trends outside your control are bank approval times, DISCOM requirements and customer payment behaviour. The things you can measure yourself are task velocity by stage, cash collected before dispatch, the PM Surya Ghar deadline on each project, and team capacity by role.

A project cannot stall silently between two departments when every open checklist item has a date and an owner. That does not mean the bank moves faster, but it means you know whether the bank is the reason a project is late, or something inside your own process is. A reminder asks for a status. A gate decides when a project can move forward. Operations software like Solar Spine puts both in one place, not in a report run at month end but in the queue that lands on your desk every morning. Where the trends above move, your hiring and process changes can move with them, ahead of the problem rather than weeks behind it.

Run one real project through it this month. You will know by the end of it.

30 days free, no card, no call. Then ₹10,000 a month for the whole company, however many people you put on it.