A public sector undertaking that wants a new technology has more legal room than its managers often assume, and less practical room than its press releases suggest. The rules that govern central public sector enterprises (CPSEs) already allow boards to enter technology joint ventures, buy know-how, set up subsidiaries and procure from startups without asking for years of turnover. What slows PSU innovation in India is rarely an outright prohibition. It is the combination of lowest-price procurement habits, approval layers for anything that looks like equity, and the fear among officers that an honest bet which fails will be read later as a lapse.
This piece maps the routes a PSU can actually use to build or buy new technology, sets out the constraints attached to each, and proposes a route map that matches the route to the maturity of the technology and to the level of scrutiny the decision will attract. It draws on Department of Public Enterprises (DPE) guidelines, the General Financial Rules, Comptroller and Auditor General (CAG) findings on CPSE research spending, Ministry of Petroleum and Natural Gas disclosures on startup funding by oil CPSEs, and the Central Vigilance Commission's current position on bona fide commercial decisions.
The audience is the chief technology officer, the strategy head or the director (technical) of a CPSE, and the founder of a deeptech venture trying to understand why a PSU pilot that everyone liked has not turned into a purchase order. A separate PanScience Research piece covers government procurement routes for deeptech ventures in general; this article stays with the PSU as builder and buyer.
What the rules already permit
The starting point is the delegation of powers. Under the DPE's Maharatna guidelines (office memorandum of 4 February 2010), the boards of Maharatna CPSEs may incur capital expenditure on new items or replacement "without any monetary ceiling". They may make equity investments in joint ventures and wholly owned subsidiaries up to 15% of the CPSE's net worth in one project, capped at ₹5,000 crore, and up to 30% of net worth across all such projects. The same guidelines allow boards to "enter into technology joint ventures or strategic alliances" and to "obtain by purchase or other arrangements, technology and know-how", subject to government guidelines issued from time to time. The Maharatna powers are exercised under the same conditions that apply to Navratna CPSEs, and smaller CPSEs operate under narrower delegations set out in their own guidelines.
These are substantial powers. The constraint arrives through process rather than through the ceiling. The guidelines require professional appraisal of proposals, advance written circulation to the board, the presence of government directors and the functional directors concerned for major decisions, and a preference for unanimity. A later DPE office memorandum of 10 August 2016 added that proposals to establish joint ventures and wholly owned subsidiaries should be presented to the board, that the administrative ministry should obtain NITI Aayog's concurrence on a case-by-case basis, and that decisions should be taken only when government directors are present.
For procurement, the General Financial Rules 2017, which apply to central ministries, departments and associated CPSEs, carry two startup provisions. Rule 170(i) relaxes the earnest money deposit for DPIIT-recognised startups, and Rule 173(i) relaxes prior experience and turnover requirements. The procurement manuals for consultancy and other services and for works carry equivalent relaxations. The Government e-Marketplace (GeM) applies the same exemptions to startups, subject to meeting quality and technical specifications, according to the Commerce Ministry's February 2019 announcement of its startup initiatives.
So the formal position is permissive. A CPSE can buy from a two-year-old company, can co-develop with it, and in the larger categories can invest in a vehicle that builds the technology.
Where PSU innovation in India actually stalls
Three frictions explain most of the gap between permission and practice.
The first is evaluation. Public procurement in India defaults to a lowest-price comparison among technically compliant bids. That works for commodities and badly for something that has no comparable product. The Commerce Ministry named the problem directly in 2019 when it launched the startup route on GeM: product comparison and price discovery were "the two main obstacles" for government buyers purchasing innovative startup products. The GeM answer was a buyer-feedback mechanism in which innovative products rated three or more on a five- point scale by at least three buyers could be accepted for listing as regular products. That solves listing. It does not solve the deeper issue that a first-of-its-kind product has no market price to benchmark against, and an officer who pays more than a comparable legacy product must be able to defend the premium.
The second is the R&D habit. DPE issued guidelines in September 2011 asking Maharatna and Navratna CPSEs to spend at least 1% of profit after tax on research and development, and other CPSEs at least 0.5%. CAG's review of 21 CPSEs for 2013-14 to 2017-18 found the 1% level was met in 79 of 94 company-years, and that the spread was very wide: BHEL, BEL and ECIL spent multiples of the floor, while NMDC and Power Grid were below 1% in every year, and Power Grid's budget utilisation was as low as 5 to 8%. Most CPSEs reviewed did not transfer lapsed R&D budgets to a dedicated fund as the guidelines required. DPE withdrew the guidelines in July 2019 on the grounds that memorandum of understanding (MoU) parameters had taken over the role. The lesson is not that CPSEs do not do R&D; several do a great deal. It is that a spending floor measured against profit produced compliance at the floor in many companies and little evidence of a pipeline from research to deployment.
The third is personal risk. Officers who approve an unusual purchase or an investment that later loses money worry about how the decision will be read years later. The Central Vigilance Commission's master circular on the definition of vigilance angle (No. 01/MC/2026) addresses this. As summarised by the government employee news service that published it, the circular treats commercial risk-taking as a normal part of how these organisations operate, says good-faith decisions that end in financial loss should not automatically trigger vigilance action, and frames the test as whether the decision was made honestly, within rules and procedures, and in the organisation's legitimate interest. Routine procedural lapses without an integrity concern fall under conduct and discipline rules rather than vigilance proceedings. This is a meaningful statement, but it protects decisions that are documented as honest, rule-compliant and in the organisation's interest. The practical implication is that innovation routes need paperwork designed to show exactly those three things.
Evidence that PSUs can fund startups at scale
The most concrete public example of CPSEs funding startups as a group is in oil and gas. The Ministry of Petroleum and Natural Gas reports that ten oil and gas CPSEs (ONGC, OIL, IOCL, BPCL, HPCL, EIL, GAIL, NRL, MRPL and Balmer Lawrie) committed a startup corpus of ₹320 crore over three years. The ministry's page reports 167 projects selected, ₹195.30 crore committed and ₹120.65 crore disbursed, and a further planned fund of ₹291.91 crore for the following three years. The ministry does not say, on that page, whether the money is grant or equity.
Other examples are smaller but instructive. The Startup India public procurement page lists PSU-specific measures: HAL earmarks Make-II projects of up to ₹250 lakh for startups; BEL has created special categories for areas such as AI, machine learning and cyber security with Make-II relaxations; NTPC has issued vendor assessment guidelines and a list of non-critical activities open to startups; EIL has eased vendor enlistment; the Airports Authority of India uses a grand challenge model with counter-bidding; and HPCL runs a programme for proof of concept and commercialisation. In defence, Make-II financial caps for startups are set by each defence PSU.
Read together, these show the pattern that recurs in PSU innovation: the startup gets in through a non-critical category, a challenge, or a co- development scheme with a capped value. The harder question is what happens after the capped first engagement.
The scale of the procurement channel matters too. A PIB document dated 28 February 2025 reported that startups had fulfilled GeM orders worth ₹35,950 crore, against total GeM order value of ₹4,03,305 crore in the previous financial year. That figure aggregates all government buyers, not only CPSEs, and mixes commodity supply with innovation. It shows the channel is real; it does not show how much of it is new technology.
The PSU innovation route map
The routes available to a CPSE differ in what they cost, who must approve them, how long they take, and how much scrutiny they attract. The table below is the original contribution of this piece. It is an analytical classification built from the rules described above; the "scrutiny" column is a judgement about how visible and how open to later challenge each route is, not a legal assessment.
The route map suggests three rules of thumb.
Match the route to the maturity. A technology at prototype stage belongs on routes 1 to 3. Pushing it straight into an open tender (route 5) asks a procurement committee to compare it with mature products on price, which it will lose, or to accept it without comparison, which the committee will not do. Pushing it into equity (route 7) asks a board, a ministry and NITI Aayog to approve a business case for something that has not yet worked in the field.
Plan the hand-off between routes, not just the first route. The common failure is a successful route 2 or route 3 engagement that has no defined path to route 4 or 5. The oil CPSE disclosures show 167 funded projects; the public record says much less about how many became suppliers to the CPSEs that funded them. A CPSE that funds a proof of concept should decide at the outset which business unit will own deployment, what performance would trigger a purchase, and which procurement route the purchase will use.
Build the documentation for the vigilance test into the route. Each route should generate, as a by-product, a record that the decision was honest, followed the applicable rule, and served the organisation's interest. For routes 2 to 4 that means a written problem statement from the user department, a documented evaluation of alternatives (including the legacy option), pre-agreed performance criteria and a record of who decided what.
Building versus buying inside a PSU
Most PSU innovation decisions are framed as procurement questions, but the strategic choice is between building capability and buying it.
Building means internal R&D (route 1), co-development (route 3) and, at the far end, a subsidiary (route 7). The case for building is strongest where the technology is core to the CPSE's mission, where the operating data needed to develop it exists only inside the CPSE, or where security considerations rule out dependence on an outside supplier. The CAG findings suggest that where building is done well, it is concentrated: a few CPSEs with engineering cultures account for most of the high R&D ratios.
Buying means startup procurement (route 4), open tenders (route 5) and alliances (route 6). The case for buying is strongest where the technology is general-purpose, where a startup has already proved it elsewhere, and where speed matters more than ownership. For enterprise AI specifically, the build, buy or co-build choice is covered in a separate PanScience Research piece; the same logic applies inside a PSU, with the added constraint that "co-build" has to be expressed through one of the routes in the table.
A middle path that the rules support but that is rarely written down is staged buying. A CPSE can run a capped proof of concept with a startup, agree in advance the criteria for a limited first order under the startup relaxations, and only then move to an open tender for scale. Each stage uses a different rule, and each produces evidence for the next. The startup gets a credible reference; the CPSE gets a benchmark price that did not exist before the first stage.
PanScience reports that Nyaay AI, one of its ventures, has deployments in more than 14 Indian High Courts (company-reported). Courts are not PSUs, but the experience of selling into public institutions in India points the same way: the first engagement is rarely the hard part, and the move from pilot to a budgeted line item depends on an internal owner who can defend the choice later.
What this means for founders selling to PSUs
For a deeptech founder, the route map changes the first conversation. Instead of asking a PSU for a pilot, ask which route the pilot sits on and which route the purchase after it will use. If the answer to the second question is "an open tender", the founder should expect the specification to be written later by people who did not see the pilot, and should work with the user department to document performance in terms a specification can use.
Founders should also understand the relaxation precisely. The GFR relaxations remove barriers on prior experience, turnover and earnest money deposit; they do not remove the requirement to meet quality and technical specifications, and they do not change how price is evaluated. A startup that wins a capped first engagement on merit can still lose the scale tender on price to an incumbent with a weaker product.
Finally, equity from a PSU is a long road. The approval chain for a subsidiary or JV, including NITI Aayog concurrence where the 2016 office memorandum applies, means a founder should not plan a funding round around a PSU equity cheque. Commercial engagements and grant- funded co-development are faster and, for most ventures, more valuable.
What this means for PSU leadership
First, give innovation a portfolio budget across routes rather than a single R&D line. The withdrawn 2011 guidelines measured spending as a share of profit; a portfolio view would measure how many projects moved from one route to the next.
Second, name a deployment owner for every funded proof of concept. The user department, not the R&D wing, should co-sign the problem statement and the success criteria.
Third, use the vigilance circular as a design specification rather than a shield: templates for routes 2 to 4 should capture honesty, rule compliance and organisational interest as a matter of course. Officers will take reasonable risks when the process protects them.
Conclusion
The question for PSU innovation in India is less whether public sector undertakings may build and buy new technology than whether they can move a technology from one permitted route to the next without losing it in the gaps. The DPE delegations give large CPSEs considerable powers over capital, alliances and know-how; the GFR and GeM give startups a way in; the oil CPSEs have shown that pooled startup funding at the scale of hundreds of crores is achievable; and the Central Vigilance Commission has said plainly that honest commercial risk- taking is not in itself a vigilance matter. The work that remains is in route design: matching each technology to the right route, planning the hand-off, and producing evidence at each step that protects the officers who take the decision.
Methodology and sources
This article is based on documents opened and read in October 2026. DPE's delegation of powers for Maharatna CPSEs, including the office memorandum of 4 February 2010 and the office memorandum of 10 August 2016 on joint ventures and subsidiaries, was read from the DPE compilation (https://dpe.gov.in/sites/default/files/power.pdf). The 2011 DPE R&D guidelines, their withdrawal in 2019 and the compliance findings come from Chapter 7 of CAG Report No. 18 of 2019 (https://cag.gov.in/uploads/download_audit_report/2019/Chapter_7_Expenditure_on_Research_and_Development_by_CPSE_of_Report_No_1 GFR startup relaxations and PSU-specific measures come from the Startup India public procurement page (https://www.startupindia.gov.in/content/sih/en/public_procurement.html). The GeM startup route and the buyer-feedback mechanism come from the PIB release of 19 February 2019 (https://pib.gov.in/PressReleasePage.aspx?PRID=1565237), and startup order values on GeM from a PIB document dated 28 February 2025 (https://static.pib.gov.in/WriteReadData/specificdocs/documents/2025/mar/doc202532511101.pdf). Oil CPSE startup funding figures come from the Ministry of Petroleum and Natural Gas (https://mopng.gov.in/en/refining/start-up). The Central Vigilance Commission master circular No. 01/MC/2026 was read through the summary published by Gconnect (https://www.gconnect.in/top-stories/cvc-master-circular-definition-of-vigilance-angle-2026.html); readers should consult the circular itself before relying on its wording. The route map, the scrutiny ratings and the rules of thumb are the author's analytical framework, not observed data. PanScience facts are company-reported. Nothing here is legal advice; delegations and procurement rules change, and CPSE-specific delegations differ from the Maharatna position described.