Recovering a rocket's first stage is a solved problem. It has been done routinely for years. Recovering the upper stage is not, and almost nobody does it — which is what makes this week's agreement more interesting than its value suggests.
On 25 September 2026, the Technology Development Board of the Department of Science and Technology signed an agreement with Agnikul Cosmos Private Limited, Chennai, for ₹200 crore of support under the Research Development and Innovation Fund, to develop the Agnibaan RLV — a reusable launch vehicle — taking the technology from TRL-4 and above towards TRL-8. Support is through Optionally Convertible Debentures, the same instrument used in the eVTOL award we covered in our explainer on the TRL scale and the valley of death.
The release states the objective precisely: to move beyond the conventional approach of recovering only the first stage and work towards full-system reusability.
Why the two stages are not the same problem
The difference is energy, and it follows from a single line of physics.
Kinetic energy rises with the square of velocity. A first stage separates at a few kilometres per second and on a largely vertical trajectory; it is high, but it is not going fast enough for the atmosphere to destroy it. It can flip, relight its engines, decelerate and land.
An upper stage has done the opposite job. Its task is to accelerate the payload to orbital velocity — roughly 7.8 km per second in low Earth orbit. Bringing it home means shedding all of that energy, and because the energy scales as velocity squared, an object returning from orbit carries many times the energy of a first stage on a suborbital arc.
That energy has to go somewhere, and it goes into heat. Re-entry from orbit produces temperatures that require a thermal protection system — heat shielding that must survive the descent, and for reuse, survive it repeatedly without extensive refurbishment. The mass penalty is severe: every kilogram of heat shield and landing hardware on the upper stage is a kilogram of payload not delivered, and the upper stage is exactly where mass is most expensive because it is carried all the way to orbit.
This is the reason upper-stage reuse has defeated almost everyone who has attempted it, and why an Indian start-up proposing a reusable upper-stage architecture with a descent propulsion system is making an ambitious claim rather than an incremental one.
Deep throttling and restart
The release names two propulsion capabilities as central: deep throttling and restart. Both are landing requirements, and both are genuinely difficult.
Restart is the simpler to state. A rocket engine designed to burn once, continuously, from ignition to shutdown is a different machine from one that must shut down in vacuum and light again reliably later. Relighting requires an ignition system that works after cold-soaking in space, propellant that can be settled at the tank outlet in microgravity before the pumps are fed, and a combustion chamber that tolerates the thermal cycling. An engine that cannot restart cannot perform a landing burn.
Deep throttling is the more demanding. An engine is sized to lift a fully fuelled vehicle off the pad. By the time that stage returns, it is nearly empty — it may weigh a small fraction of its launch mass. To hover or descend slowly, the engine must produce a correspondingly small fraction of its rated thrust. That means running far below the design point, where combustion becomes unstable, injectors work outside their efficient range and the whole system is prone to oscillation. Designing an engine that throttles deeply and remains stable is one of the harder problems in liquid propulsion, and it is why landing rockets is not simply a matter of pointing the engine downward.
Semi-cryogenic, and what Agnikul has already flown
The Agnibaan RLV uses a semi-cryogenic liquid propulsion system. The term is worth being precise about, because it is asked.
A cryogenic engine uses propellants that must be kept at extremely low temperatures — typically liquid hydrogen as fuel with liquid oxygen as oxidiser. It offers the highest performance but liquid hydrogen is bulky, extremely cold and difficult to store and handle.
A semi-cryogenic engine uses a cryogenic oxidiser with a non-cryogenic fuel — liquid oxygen with a refined kerosene or similar hydrocarbon. Performance is lower than full cryogenic but the fuel is dense, storable at ordinary temperatures and far easier to handle. For a reusable vehicle that must be turned around quickly and repeatedly, that operational simplicity matters more than the last few seconds of specific impulse.
Agnikul is not starting from a blank sheet. The company, incubated at IIT Madras, flew Agnibaan SOrTeD — a single-stage suborbital technology demonstrator — on 30 May 2024 from Satish Dhawan Space Centre, Sriharikota. That flight carried three distinctions worth holding: it used the world's first single-piece 3D-printed rocket engine, the Agnilet; it was India's first semi-cryogenic engine flight; and it launched from 'Dhanush', India's first privately developed launchpad. The vehicle was about 6.2 metres tall, 575 kg, with an engine producing about 6.2 kN of thrust for a 70-second burn on liquid oxygen and aviation turbine fuel.
The single-piece 3D-printed engine is the detail most worth understanding. A conventional rocket engine is assembled from a large number of separately manufactured parts, and every joint is a potential leak, a mass penalty and an assembly step. Printing the engine as one component removes the joints, compresses manufacturing time from months to days, and makes iteration cheap — which for a company that needs to test repeatedly is a strategic advantage rather than a manufacturing convenience.
Where ISRO's work sits alongside
India's reusability effort is not confined to start-ups, and the two strands should be distinguished.
ISRO's RLV-TD programme is built around a winged vehicle named Pushpak, after the mythological flying chariot. Its Landing Experiment (LEX) series tested autonomous landing: in LEX-03, on 23 June 2024, Pushpak was released from an Indian Air Force Chinook helicopter at about 4.5 km, executed cross-range correction manoeuvres autonomously, and landed precisely on a runway centreline at over 320 km per hour, decelerating with a brake parachute and landing-gear brakes. It was the third and final flight of that series.
The two approaches differ in method. Pushpak is a winged, runway-landing vehicle, using aerodynamic lift to control descent. The Agnibaan RLV is described around descent propulsion — powered, retro-thrust landing. Winged vehicles carry wing mass to orbit and back but land gently; propulsive landing avoids wings but spends propellant that could have been payload. Neither is obviously correct, and both are being pursued worldwide.
Why reusability matters beyond cost
The obvious argument is economic: a vehicle flown many times spreads its manufacturing cost across many missions, and the release names reducing launch and manufacturing costs and enabling high-frequency, repeatable launches.
The less obvious argument is the one the release also makes, and it is the better answer in an interview: limiting the generation of space debris.
An expendable upper stage does not vanish after delivering its payload. It remains in orbit, often for years, as a large uncontrolled object. Spent stages are among the most significant debris items in low Earth orbit, precisely because they are massive and numerous. A collision involving one generates thousands of fragments, each capable of destroying another satellite — the cascading process known as the Kessler syndrome, in which debris density becomes self-sustaining and certain orbits become unusable.
A recovered upper stage is a stage that is not left in orbit. As launch rates rise worldwide, that shifts from an environmental nicety to a condition for continued access. India's own orbital ambitions and the wider policy frame are set out in our pieces on the international space summit and ISRO's roadmap and on GSLV-F17 and the case for geo-imaging.
🔑 Revision block
- The agreement: 25 September 2026 — TDB, Department of Science and Technology, with Agnikul Cosmos Private Limited, Chennai
- Amount: ₹200 crore under the Research Development and Innovation (RDI) Fund, through Optionally Convertible Debentures
- Objective: develop the Agnibaan RLV, moving from TRL-4 and above towards TRL-8
- Ambition: beyond recovering only the first stage — towards full-system reusability, including a reusable upper stage with descent propulsion
- Why the upper stage is harder: it returns from orbital velocity, about 7.8 km/s in low Earth orbit; kinetic energy scales with the square of velocity, so re-entry heating is far more severe, requiring a thermal protection system and imposing a heavy mass penalty
- Restart: an engine must relight in vacuum after shutdown — needing ignition after cold-soak, propellant settling in microgravity, and tolerance of thermal cycling
- Deep throttling: an engine sized for a fully fuelled lift-off must produce a small fraction of rated thrust to land a nearly empty stage; combustion becomes unstable far below the design point
- Cryogenic vs semi-cryogenic: cryogenic uses liquid hydrogen with liquid oxygen (highest performance, hardest handling); semi-cryogenic uses a cryogenic oxidiser with a non-cryogenic fuel — liquid oxygen with refined kerosene — denser, storable and easier to handle
- Agnibaan SOrTeD: flew 30 May 2024 from Sriharikota; about 6.2 m, 575 kg, thrust about 6.2 kN, burn 70 seconds, propellants LOX and aviation turbine fuel
- Three firsts on that flight: world's first single-piece 3D-printed rocket engine (Agnilet); India's first semi-cryogenic engine flight; launch from 'Dhanush', India's first privately developed launchpad
- Agnikul: incubated at IIT Madras
- ISRO's RLV-TD: winged vehicle Pushpak; LEX-03 on 23 June 2024, released from an IAF Chinook at about 4.5 km, autonomous runway landing above 320 km/h, decelerated by brake parachute and landing-gear brakes
- Two approaches: winged runway landing (Pushpak) versus propulsive descent landing (Agnibaan RLV)
- Debris argument: spent upper stages are major debris objects; the cascading collision process is the Kessler syndrome
🎯 Practice MCQs
Q1. Recovering a launch vehicle's upper stage is harder than recovering its first stage principally because: (a) The upper stage is larger (b) It returns from orbital velocity, carrying far greater kinetic energy (c) It carries no engines (d) It separates at a lower altitude
→ (b) — energy scales with the square of velocity.
Q2. A semi-cryogenic rocket engine uses: (a) Liquid hydrogen with liquid oxygen (b) Solid propellant with a liquid oxidiser (c) A cryogenic oxidiser with a non-cryogenic fuel such as refined kerosene (d) Compressed gas propellant
→ (c) — denser and easier to handle than full cryogenic.
Q3. "Deep throttling" is required for landing because: (a) The engine must produce maximum thrust on descent (b) Throttling improves fuel efficiency at altitude (c) It reduces acoustic noise near the ground (d) An engine sized for a fully fuelled lift-off must produce far less thrust to land a nearly empty stage
→ (d) — and combustion becomes unstable far below the design point.
Q4. Agnibaan SOrTeD, flown in May 2024, was notable for using: (a) A solid rocket motor (b) The world's first single-piece 3D-printed rocket engine (c) A nuclear thermal engine (d) An air-breathing scramjet
→ (b) — the Agnilet engine.
Q5. ISRO's winged reusable launch vehicle technology demonstrator is named: (a) Pushpak (b) Gaganyaan (c) Vikram (d) Rohini
→ (a) — after the mythological flying chariot.
Q6. In the RLV LEX-03 experiment of June 2024, Pushpak was released from: (a) A high-altitude balloon (b) A sounding rocket (c) An Indian Air Force Chinook helicopter (d) A modified transport aircraft
→ (c) — at about 4.5 km, landing autonomously on a runway.
Q7. Orbital velocity in low Earth orbit is approximately: (a) 3.2 km per second (b) 11.2 km per second (c) 7.8 km per second (d) 1.5 km per second
→ (c) — 11.2 km/s is Earth's escape velocity.
Q8. The "Kessler syndrome" describes: (a) Engine failure caused by combustion instability (b) Heat shield degradation during repeated re-entry (c) Loss of signal during atmospheric re-entry (d) A cascade in which collisions generate debris that causes further collisions
→ (d) — potentially rendering certain orbits unusable.
Q9. A single-piece 3D-printed rocket engine offers the advantage of: (a) Eliminating joints, compressing manufacturing time and making iteration cheaper (b) Higher specific impulse than any conventional engine (c) Operation without an oxidiser (d) Immunity to combustion instability
→ (a)
Q10. Consider the following statements: 1. A reusable upper stage reduces the number of large uncontrolled objects left in orbit. 2. Pushpak and the Agnibaan RLV use the same landing method. Which is/are correct? (a) 1 only (b) 2 only (c) Both 1 and 2 (d) Neither 1 nor 2
→ (a) — Pushpak lands on a runway using wings; the Agnibaan RLV is built around descent propulsion.
📋 How this gets asked (PYQ pattern)
Space technology is one of the densest topics in the NDA general ability paper, and launch vehicles carry a large share of the questions.
The vehicle question asks about India's launchers and their payload roles. PSLV for polar and sun-synchronous orbits, GSLV with a cryogenic upper stage for geostationary transfer, LVM3 for heavy payloads and Gaganyaan, SSLV for small satellites. Learning them by orbit and payload class rather than by name is what makes the question easy.
The propulsion question asks the difference between solid, liquid, cryogenic and semi-cryogenic stages. The cryogenic-versus-semi-cryogenic distinction is the discriminating one, and it turns on which propellant is cryogenic.
The facility question covers Satish Dhawan Space Centre at Sriharikota, VSSC at Thiruvananthapuram, LPSC, and the new launch complex at Kulasekarapattinam. Locations are asked as often as functions.
The physics question asks for orbital velocity, escape velocity and the distinction between them. 7.8 km/s and 11.2 km/s — two numbers that are confused constantly and cost a mark each time.
For the SSB interview, reusability supports a genuinely good answer. A candidate who explains that the economic case is obvious but the debris case may matter more — because spent stages are among the largest objects left in orbit and collisions cascade — is thinking about space as a shared and finite environment rather than as a series of launches. That is a more considered position than enthusiasm about cost per kilogram.
Preparing for NDA? Space questions get much easier once launch vehicles are learnt by orbit and propulsion type rather than as a list of names. Build the base with our NDA general ability notes, follow the daily NDA current affairs, and prepare with our faculty in the upcoming Cavalier courses in Delhi.
✍️ Written by Col Vijyanat Thakur — Defence studies faculty at The Cavalier. Reviewed by the Cavalier Faculty Desk.