In today’s deep-dive, we will look at a problem that is becoming harder to ignore as India’s deeptech startups start winning larger defence, space and industrial orders. The challenge is no longer only about funding research or building the product. It is also about financing the months between receiving a purchase order, manufacturing the equipment and finally getting paid. We will look at why this gap exists, how startups are managing it, and whether India’s banks, investors and procurement system are ready for the next stage of deeptech growth.

In May this year, a two-year-old defence startup called Armory announced something most founders would love to announce. The company had won three contracts worth around ₹100 crore from the Ministry of Defence for SURGE, its counter-drone system.

Armory had raised only ₹35 crore in equity until then. Suddenly, a company founded in 2024 had government orders worth almost three times all the equity it had ever raised. Founder Amardeep Singh said the company would now expand manufacturing, hire more people and invest further in hardware. Armory was also planning another fundraising round.

There is no suggestion that Armory is in financial trouble. In fact, winning ₹100 crore of orders so early is a remarkable achievement. But the numbers show a problem that is becoming more important as India's deeptech startups move from prototypes to large commercial orders.

A ₹100 crore purchase order is not ₹100 crore sitting in the bank.

The startup may have to buy electronics, motors, sensors and other components months before delivery. It has to pay engineers and factory workers. Inventory sits inside the company while systems are assembled and tested. The customer may conduct trials, ask for modifications and inspect the equipment before accepting it. The startup may also have to arrange bank guarantees.

Only then does a large part of the money arrive.

This is very different from the deeptech problem India has spent the last few years trying to solve. Until recently, the main question was whether somebody would fund the science. Increasingly, the question is becoming whether somebody will fund the order.

And the timing could not be more important. Deeptech startups have already raised around $2.4 billion across 213 deals in 2026, compared with about $1.6 billion across 458 deals in the whole of 2025, according to Tracxn data cited by Mint. More money is going into the sector even though the number of deals has fallen sharply.

That tells us something useful. Investors are becoming more willing to write large cheques for companies that have survived the early technology risk. What India has not solved nearly as well is the boring finance required after those companies begin receiving large orders.

Imagine a startup receives a ₹100 crore defence order. The exact payment terms will differ from contract to contract, but suppose it receives ₹15 crore or ₹20 crore upfront and must spend another ₹50 crore or ₹60 crore before enough milestones are cleared.

That company can have a ₹30 crore or ₹40 crore cash hole even though it has a customer, a signed order and technology that has already passed the procurement process.

For a consumer internet company, growth often releases cash quickly. A customer orders food, takes a cab or pays for software and the money reaches the company almost immediately. There may be marketing costs and losses, but a successful transaction normally creates cash.

Deeptech manufacturing can work in the opposite direction. Growth first consumes cash.

The larger the order, the more raw material has to be bought. The more systems being built, the more inventory sits inside the company. More production may require new machines, larger facilities and additional engineers. If the customer takes longer to test or approve the equipment, the working-capital requirement becomes larger again.

This problem is already visible well beyond startups.

CRISIL studied 24 private defence companies that account for around 42% of private-sector defence revenue. Their combined order book was close to ₹50,000 crore, and CRISIL expects their revenue to grow another 15-16% this year. But the same companies are expected to require ₹1,600-1,800 crore of working capital, apart from ₹1,500-1,600 crore of capital expenditure.

More importantly, CRISIL estimates their normal working-capital cycle at around 240-250 days. A six-month delay in execution can add another 35-40 days.

That is eight months of money tied up in the business before we even get into the worst cases.

These are not poorly run startups. Many are established defence companies with profits, credit histories and access to banks. If their cash cycle is this long, the same business model becomes much harder for a four-year-old company whose balance sheet was built for R&D rather than ₹200 crore of manufacturing.

This distinction matters because delayed payment is often described too simply.

Under India's MSME rules, buyers are generally expected to pay eligible micro and small enterprises within 45 days of accepting goods or services. The government has also built mechanisms such as MSME Samadhaan and the newer Online Dispute Resolution system. By December 2025, more than 2.56 lakh delayed-payment applications involving over ₹55,000 crore had been filed on the Samadhaan system.

But imagine a radar company that spends four months making a system and then waits another three months for trials, inspection and acceptance.

The 45-day clock does not finance those first seven months.

Astra Microwave gives us a good look at how this works inside a real defence supplier. In FY26, the company's inventory stood at around 275 days and debtors at about 218 days. Its gross current assets were roughly 429 days, although that was an improvement from more than 500 days a year earlier.

CRISIL explains why the numbers are so large. Defence products have long gestation periods, require multiple customer tests and may need third-party inspections before the bill itself is cleared.

This is an important correction to the usual argument that the government simply takes too long to pay an invoice. Sometimes the startup is waiting months to reach the point where it can raise that invoice.

The cash journey can look something like this: buy components, manufacture the equipment, hold inventory, conduct trials, make changes, obtain inspection approval, complete acceptance, submit the bill and finally collect the money.

Financing only the final receivable solves one part of that journey.

Tonbo Imaging offers another useful example.

Tonbo makes imaging and vision systems used in defence and security. CRISIL says its gross current assets reached 515 days in March 2026, partly because of high debtor balances and the inventory required to maintain supplies.

But Tonbo also has something a young startup usually does not have.

CRISIL has rated ₹500 crore of bank facilities for the company. These include cash-credit lines, working-capital demand loans, term loans and about ₹95.5 crore of bank guarantees.

This tells us something about what a defence company eventually has to become.

The obvious moat may be the technology: better sensors, better autonomous systems, better radar, better drone software. But once orders become large, another moat begins to matter. The company needs banks willing to give it ₹100 crore of cash credit and another ₹100 crore of guarantee limits while customers take months to approve and pay.

Two startups may have equally good technology. One has ₹150 crore of sanctioned banking lines and established lender relationships. The other has ₹20 crore in the bank from its last VC round.

The first company can safely accept a ₹200 crore order. The second may have to raise money before it can celebrate winning the same order.

This is not unique to startups either. In 2019, Hindustan Aeronautics disclosed that around ₹9,500 crore of receivables from customers including the Indian Air Force and Army had been delayed. HAL had around ₹12,000 crore in reserves and surplus, but still had to borrow from banks to meet working-capital requirements.

HAL could do that because it was HAL.

A young company may have the same government customer and a much smaller receivable, but nowhere near the same ability to borrow against it.

This week, NewSpace Research and Technologies made the issue unusually visible.

NRT, which builds autonomous drones and high-altitude systems, is looking to raise another $35-40 million through a mixture of equity and debt. Its financial history looks strange if viewed like a normal technology company.

Revenue was about $13.5 million in FY23, when it made a small profit. Revenue then fell to just $765,000 in FY24, when losses reached $5.6 million. In FY25, revenue recovered to around $12.2 million, but losses increased to $7.9 million. Earlier this year, the company reportedly won an approximately $18.6 million government contract for solar-powered UAVs.

An investor quoted by Mint said the revenue promised by the company had not materialised consistently. But another person familiar with NRT's position made an equally important point, saying it was “not necessarily their fault” because government contract payments had not come through.

Both statements can be true.

A company can miss its forecast because management was too optimistic. It can also miss the forecast because a ₹100 crore contract moved from March to June after testing or government approval took longer than expected.

For investors, separating those two situations is difficult.

That is why annual revenue growth can be a misleading way to look at defence deeptech. A company may report ₹100 crore one year, ₹20 crore the next year and ₹160 crore the year after without its underlying competitive position changing nearly as much as the income statement suggests.

The better questions may be about the order book, milestone completion, cash collected, inventory, receivable days, customer concentration and how much money is tied up in guarantees.

This requires a different investing skill.

Debraj Banerjee of Fundamentum Frontier Advisors recently described the problem as the “SaaS metric mindset” among some domestic investors. He argued that if a deeptech company is viewed like SaaS, investors may never make the bet. Airbound founder Naman Pushp described something similar after trying to raise a large round in India. One fund's growth team could not evaluate the company using normal growth metrics, while the early-stage team was uncomfortable writing a cheque of that size.

Deeptech companies can therefore get stuck in a strange place. They are too capital-intensive for an early-stage VC and too uneven for a conventional growth investor.

Then the bank asks where the equity round is.

India's venture-debt market has grown quickly. Deployment reached around $1.3 billion in 2025, up from $1.2 billion in 2024, even as the number of transactions fell from 238 to 187.

This is useful capital. Companies such as NRT have already borrowed from venture lenders, and its previous $52 million round included debt from SBI's startup hub and SIDBI.

But its current fundraise reveals the limitation. According to Mint, NRT's debt investors are likely to proceed only after the company secures the equity portion of the new round.

There is nothing irrational about that. Venture lenders have to protect their money and a fresh equity round provides a larger cash cushion.

But it also shows why venture debt is not always a substitute for proper working-capital finance.

If a startup has a confirmed government order worth ₹100 crore, a working-capital lender should ideally be able to study the order, the customer, the production schedule, milestone payments, delivery risk and expected collection. The decision should partly depend on whether the contract itself can support a ₹30 crore loan.

A venture lender may still care more about whether the next VC round closes.

That means a company can end up using expensive equity to make a lender comfortable enough to finance a low-credit-risk government receivable. The founder gives away ownership partly because the credit market does not know how to finance the waiting period.

For a software startup, using VC money to hire developers and acquire customers is normal. For a defence manufacturer, using a ₹100 crore equity round to buy inventory against a government order is a much more questionable use of equity.

Equity is permanent and expensive. Inventory finance should not have to be.

The interesting part is that none of this has come as a surprise to policymakers.

India's draft National Deep Tech Startup Policy, prepared in 2023, explicitly identified payment delays causing working-capital problems as one of the gaps in the ecosystem. It also said growing deeptech startups need better access to credit and face difficulties meeting collateral requirements. The document proposed specialised debt products and even suggested a debt fund backed by the Credit Guarantee Scheme for Startups.

The procurement section becomes even more specific.

The policy document says government advances can be 15% or less, with milestone payments coming later. It then describes the problem of an Advance Payment Bank Guarantee. According to the draft, a large company may obtain such a guarantee while placing only 3-5% of the guarantee amount in a fixed deposit, whereas a startup can be asked for a 100% cash margin. The policy recommended bringing startups closer to the economics available to large companies.

Consider what that means in practice.

A government department may say it is helping the startup by paying an advance. The bank may then ask the startup to lock almost the same amount of cash to issue the guarantee required for that advance.

The startup has technically received money without gaining much usable liquidity.

This is why bank guarantees deserve much more attention in the deeptech discussion. A founder may talk proudly about having ₹50 crore in the bank, while ₹15 crore or ₹20 crore of that money is effectively blocked against guarantees and cannot be used freely to manufacture products.

Large industrial companies spent decades building the balance sheets and bank relationships that make these guarantees cheap. Startups are being asked to compete for some of the same orders without having that financial history.

The government has made genuine progress on delayed payments.

From June this year, all operating Central Public Sector Enterprises are required to route settlement of MSME invoices through RBI-authorised TReDS platforms. Invoice discounting through TReDS has already grown from around ₹40,000 crore in FY23 to ₹3.47 lakh crore in FY26.

This matters. Once a CPSE has accepted an eligible MSME invoice, a bank or financier can discount that receivable and give the supplier cash instead of making it wait until the buyer finally pays.

But TReDS mainly attacks the receivable problem.

The hardest period for some deeptech companies begins much earlier.

If an order requires six months of component purchases, production, testing and certification before an invoice can be accepted, the startup still needs money during those six months. TReDS does not automatically buy the motors in month one simply because an invoice might exist in month seven.

India actually has a small example of what the missing product could look like. GeM Sahay allows eligible sellers on the Government e-Marketplace to receive collateral-free working-capital loans against purchase orders, rather than waiting until an invoice has been accepted. The current government description says the programme is available to eligible proprietorship sellers on GeM.

The idea is more important than the current size of the programme.

If a lender can safely make a small collateral-free loan against a verified GeM purchase order, why could a similar system not eventually finance a recognised deeptech company against a verified defence or space order?

That would move the financing point much earlier.

India has also expanded the Credit Guarantee Scheme for Startups. The maximum guarantee cover per eligible borrower was increased from ₹10 crore to ₹20 crore in 2025. Loans up to ₹10 crore can receive 85% guarantee cover, while larger eligible loans receive 75% cover.

For many startups, ₹20 crore is meaningful.

For deeptech manufacturing, it can become small surprisingly fast.

Armory has already won ₹100 crore of contracts after raising ₹35 crore of equity. A spacetech or defence company that moves from prototype quantities to a ₹300 crore production programme can easily need working capital far beyond ₹20 crore.

The irony is that India has also formally accepted that deeptech businesses are different from conventional startups. In February, the government created a separate Deep Tech Startup category, extending startup recognition from 10 years to as long as 20 years and raising the turnover ceiling to ₹300 crore. The policy explicitly cites long development periods, heavy R&D and capital intensity as the reasons.

We have therefore accepted that a deeptech company needs twice as much time to mature.

The credit system has not yet fully accepted that it may also need much more working capital while doing so.

The most interesting comparison comes from the United States.

The US Department of Defense does not assume that every contractor should finance production entirely from its own balance sheet until the final product arrives. Under current defence procurement rules, the customary cost-based progress-payment rate is 80% for large companies and 90% for small businesses on contracts where these provisions apply.

This does not mean the Pentagon simply hands 90% of every contract to a small company on day one. The rules are more complicated, eligibility varies, and the government protects itself through contractual controls.

The important difference is the principle.

Contract finance is treated as part of procurement.

If the government wants a small supplier to manufacture something expensive over a long period, the financing of that production is not automatically treated as the supplier's private problem.

Europe is moving in a similar direction through the banking system. The European Investment Bank increased its programme for financing defence-supply-chain SMEs from €1 billion to €3 billion. Its first deal under the expanded programme involved a €500 million loan to Deutsche Bank that can support around €1 billion of financing, including working capital, for defence and security companies.

This is particularly relevant to India because Europe is not trying to replace commercial banks.

It is helping the banks lend.

That may be the more practical model here as well.

India is now putting serious public and private money behind innovation. The ₹1 lakh crore Research, Development and Innovation programme is aimed at strategic technologies. Fund of Funds 2.0 has a ₹10,000 crore corpus with a stronger deeptech focus. iDEX, ADITI and other defence programmes are helping startups move through R&D and validation.

These programmes matter because somebody has to take the technology risk.

But once that risk falls and a government customer places an order, the type of money required changes.

A grant is useful when an engine may not work.

Venture capital is useful when the company itself may become huge.

A term loan is useful when a factory or machine needs to be bought.

Invoice discounting is useful once the customer has accepted the bill.

But a large missing category sits between the purchase order and the accepted invoice. That period needs ordinary working-capital credit, purchase-order finance, milestone finance and affordable bank guarantees.

The government does not need to fund every rupee directly. It could instead make a verified sovereign purchase order much easier for banks to lend against.

A specialised guarantee could cover part of the risk. Lending limits could rise with confirmed orders instead of being capped at a fixed number that becomes irrelevant once the startup scales. Production milestones could unlock financing before final acceptance. Guarantee requirements could be adjusted for recognised deeptech companies that have already cleared technical qualification.

Banks would still assess the company. The government would still test the product. Founders would still carry execution risk.

But the startup would no longer need to finance an entire strategic procurement programme out of VC equity.

There is also a counterargument here, and it is important.

Government defence customers cannot simply pay every startup immediately. Equipment has to work. Trials are necessary. Quality checks matter, particularly when the product is going into a military system. Some delays will come from failed tests, design changes, poor documentation or supply-chain problems inside the startup itself.

CRISIL's analysis makes exactly this point. Higher-value indigenous defence orders involve more testing and approvals, and execution delays of five or six months can occur.

So the answer cannot be to remove scrutiny.

The better answer is to finance credible work while that scrutiny is happening.

A bank can lend against inventory. A lender can finance completed milestones. A government guarantee can absorb part of the risk on a verified order. A procurement contract can make progress payments as manufacturing advances. None of this requires the government to accept defective equipment or abandon financial discipline.

The deeper issue is about who carries the cost of time.

Today, a young company may carry a large part of it.

That creates an odd situation. The ultimate customer may be the Government of India, one of the strongest possible counterparties in the country, but the bank is mainly underwriting a startup with three years of financial history.

The sovereign is unlikely to disappear.

The startup might.

And that is why the current deeptech funding boom can hide the next bottleneck.

India has become much better at giving scientists money to start companies. Investors are becoming more comfortable financing drones, rockets, robotics, semiconductors and advanced manufacturing. The government is increasingly willing to become the first customer.

That is real progress.

But as these companies grow, survival will depend less on whether they can raise another seed round and more on whether the financial system can turn a ₹100 crore purchase order into ₹30 crore of usable working capital without first demanding ₹30 crore of founder or VC money.

Armory is already winning orders larger than its historical equity base. NewSpace is trying to combine another equity round with debt while investors study lumpy revenues and delayed government payments. Astra carries hundreds of days of inventory and receivables. Tonbo has built a ₹500 crore banking umbrella around a business whose working-capital cycle can exceed a year.

Those companies show where Indian deeptech is heading.

The successful startup of the next decade may still need world-class science, excellent engineers and patient investors. It will also need something much less glamorous: cash credit, receivable finance, cheap guarantees and a bank willing to understand a defence purchase order.

India has spent years asking who will finance the invention. As more of those inventions finally become products, the harder question is moving to the factory floor: who will finance the months between winning the order and receiving the money?

If the answer continues to be the founder's latest VC round, India will have built a deeptech funding ecosystem that works surprisingly well until the customer actually says yes.