The first is problem-solving. The second is problem-avoidance.

Democratic systems are structurally good at the first and structurally poor at the second. The incentive is simple: problem-solving generates visible credit, good press. Problem-avoidance generates nothing visible - its entire success is the absence of a future event, paid for now, with the benefit accruing years later to a government not yet in office.

No headlines. No ribbon-cutting. No electoral reward.

This is the mechanism behind most sovereignty failures. This has become suddenly important post Fable, lack of machine tools for precision manufacturing and other supply chain exposures that announce themselves only in crisis, despite being knowable for years in advance.

Two Cases, Same Pattern

When the Strait of Hormuz closed to shipping in early 2026, India faced a problem with no obvious connection to oil. Sulphur, a byproduct of Gulf refining, is the feedstock for sulphuric acid, which underpins fertiliser, metal processing, and pharmaceuticals. Roughly 25% of globally traded sulphur moves through Hormuz. Within weeks, India’s import-linked sulphuric acid prices rose from INR 13 /kg to 31 /kg, and the government scrambled to organise emergency procurement.

India’s lithium story similar from the other direction. In February 2023, the Geological Survey of India announced some million tonnes of lithium resources in Jammu and Kashmir and a genuine political moment. Three years on, the auction for that block has been cancelled twice, and environmental clearances, land acquisition, and mining licences flowing through separate bureaucratic channels add 12 to 24 months to project timelines. The announcement was the visible part. The refinery is the invisible part and three years later, it still doesn’t exist.

(The problem is not even unique to India. US pharmaceutical supply chain and the decades of cost-optimisation left it dependent on key starting/intermediate chemicals. The country still exports most of the drugs and the dependency was flagged in 2008 early warning sign, and 4 presidents later still remains unresolved.)

So Governments can’t do long horizon work?

1. It would be easy to read this as “governments can’t do long-horizon work. period.” - and that’s both too convenient and not quite true. Singapore’s water security and South Korea’s semiconductor build-out are multi-decade state bets that paid off precisely because the institutions were structured to survive electoral cycles. The honest claim isn’t that this is impossible. It’s that it’s rare, and the rarity is the story. And ordinary citizens do not have any expectations from a rarity.

2. There’s a second serious counter argument here. Not every dependency is a failure waiting to happen. Global supply chains concentrate for real economic reasons, and a world of zero dependency isn’t obviously a safer one - just one with the fragility distributed.

   The right question isn’t about how to “eliminate dependency”. It’s “which dependencies can you tolerate, and which ones, if disrupted, would actually hurt.”

Markets sometimes answer that question faster than governments do. When the commercial case is strong enough on its own, private capital moves without waiting for policy - which is precisely where the opportunity sits for enterprises reading this. Zoho investing into Karuvi Tools is a case in point.

What This Means in Practice

There’s no near-term institutional fix. Election cycles will keep being shorter than supply chain timelines, and that won’t change by argument.

For enterprises in strategically exposed sectors, the move is to treat intermittent state failure as a planning assumption, not a surprise. Map dependencies by exposure, not just cost. Distinguish indigenisation from genuine sovereignty, which means owning the materials, the process knowledge, and the infrastructure.

And build the capability before the crisis, on the correct bet that when it arrives, the company that already has it will be in an entirely different position from the one still waiting for policy support.

Hormuz shock didn’t create India’s fertiliser/energy vulnerability. It simply revealed it.

The problem R&D leaders face is structurally identical.

Which technology trajectories are worth committing to? Where do you allocate capital across a five-to-ten-year horizon when regulation, materials costs, and market structure are all in motion simultaneously?

Graham had financial statements. Technology strategy has signals across society, technology, economy, and environment - patent velocity, TRL trajectories, supply chain exposures and several others. The indicators are different. The discipline required is the same. Our belief is that a well-designed system should help you build a structured view, make your assumptions explicit, and stress-test them before you commit - in capital, time, and attention. The methods work. They need rigour whether the work is moved forward by people or software.

That analogy has limits, and in our own research spirit we name them here.

Value investing works partly because markets revert - an undervalued stock has gravity pulling it toward fair value over time. Technology trajectories mostly diverge as new developments arrive. There is no equivalent convergence mechanism. A correct directional call can still be wrong by a decade on timing. Technology development indicators are less standardised, harder to compare across domains, and the analyst who forecasts a trajectory often participates in creating it. These differences are real, and they are precisely why we never claim the system is complete.

What we do claim is more modest and, we think, more useful: that a structured first pass - drivers mapped, TRLs estimated, scenarios built, options ranked, uncertainties drawn - is worth more than an opaque consultancy output, and worth more than an ad-hoc prompt to a general-purpose model.

We know the methods and we make the structure testable by the user. Every conclusion traces to a driver. Every driver has a source. When your engineers or management push back, they can see exactly what to challenge. The validation loop is the product. The first pass makes that loop possible.

Graham’s margin of safety was an acknowledgement that even rigorous analysis can be wrong - you build in buffer precisely because the future isn’t knowable. We work the same way. The forecast (Utkarsha AI output) we hand you is a defensible starting point, not a prediction. What you bring to it - your materials reality, your supplier relationships, your served-segment volumes, your engineering capability - is what turns it into something you can take to your board.

What emerged from that wartime pressure became known as the superalloy — and the knowledge of how to make one, reliably, at aerospace grade, at scale, has remained one of the most closely held industrial capabilities on earth ever since.

Eighty years later, a startup in Chennai printed a rocket engine from a direct descendant of that same class of material, in seven days, as a single seamless piece, with a US patent attached to the process. The achievement was real. Speciale Invest’s recent thesis on India’s $44 billion space economy cited exactly this kind of engineering-first momentum as evidence that Indian founders are rearchitecting the space stack from the ground up — and they are right to. But the question nobody asked was the one that matters most for India’s space ambitions: where did the metal come from?

---

Why Superalloys Are the Irreducible Constraint in Propulsion?

A rocket combustion chamber is one of the most hostile environments that engineered materials are asked to survive. Temperatures are extreme. Pressure loads cycle rapidly from ignition to shutdown. The material must resist oxidation, thermal fatigue, and creep — the slow deformation of metal under sustained high-temperature stress — simultaneously, and it must do so across hundreds of test firings before a vehicle is cleared for flight.

Steel fails. Titanium, excellent for airframe structure, loses strength too quickly above 500°C. Ceramics can handle the temperature but crack under cyclic mechanical load. The only materials that meet all three requirements together are nickel-based superalloys — and they account for roughly 50 percent of the materials in a modern rocket engine by weight.

Inconel 718, the specific alloy Agnikul uses for its engines, was introduced at industrial scale in 1965. Its strength at elevated temperatures comes from a precisely controlled microstructure: when the alloy is heat-treated correctly, nanoscale precipitate phases called gamma prime and gamma double prime form within the nickel matrix, locking dislocation motion and preserving mechanical properties at temperatures where other metals have long since softened. The composition — nickel, chromium, iron, niobium, molybdenum, titanium, aluminium — must be held to tight tolerances. The heat treatment sequence that produces the right precipitate structure must be calibrated to the specific manufacturing process used.

This is why superalloys are a strategic constraint rather than a commodity input. The specification is available. The knowledge of how to reliably produce material that meets it, batch after batch, is not.

---

Who Built This Knowledge and How Long It Took

The history of superalloy development is short enough to summarise and long enough to be humbling. Pratt & Whitney and General Electric began proprietary alloy programmes in the 1940s — Waspalloy, M-252, and eventually single-crystal casting techniques developed at dedicated research laboratories employing hundreds of scientists over decades. The alloy composition was always the easy part to reverse-engineer. The processing knowledge — casting conditions, cooling rates, heat treatment curves — was the moat. It accumulated in institutions, not in patents, and it did not transfer.

No country in the past thirty years has demonstrated the cost of ignoring that lesson more publicly than China.

The consequence shows up in service life: Chinese military engines improved from a few hundred hours between overhauls to roughly 1,500 hours with better blades. Western equivalents run roughly twice as long between overhauls and several times longer in total service life. As recently as July 2025, Chinese researchers at Dalian University of Technology were publishing work on new superalloy cooling techniques for sixth-generation fighter engines — an acknowledgement, through the language of state science reporting, that the yield problem remains structurally open after decades of investment.

One failed blade fails the engine. The problem was never the recipe. It is the process discipline required to hit a sub-part-per-million purity target, identically, across thousands of production casts.

India’s version of this story runs through the AL-31FP engines that power the Su-30MKI fleet. The specific superalloy grades for the hot section — BZL1, BZL14H, and ZS 6Y — were treated as restricted items that HAL sourced directly from Russia, even as it assembled the engines under licence in Koraput. That changed in early 2026, when MIDHANI — Mishra Dhatu Nigam Limited — received CEMILAC certification for domestically produced equivalents. While MIDHANI has cracked the chemistry and casting of these alloys, India still faces hurdles in:

1. Single-Crystal (SX) Blades: While the Russian alloys are advanced “Directionally Solidified” or equiaxed types, the very latest Western-grade (or Next-Gen Russian) single-crystal blades remain a significant R&D frontier for the Kaveri or AMCA engine programs.

2. Volume Production: The ability to match the scale of Russian production at a lower cost-per-unit is still being tested. It was a genuine milestone.

   The caveat is what it does not yet cover.

The Constraints Are Process, Not Chemistry

MIDHANI’s achievement covers cast and forged superalloys for jet engine applications. What Agnikul needs — and what India does not yet have — is something different: aerospace-certified gas-atomized Inconel 718 powder for additive manufacturing. The distinction matters because, as China’s turbine blade experience makes plain, knowing the chemistry of a superalloy and being able to produce it reliably at aerospace grade are entirely different problems. The constraint is always process.

A 3D-printed rocket engine starts as powder — spherical particles, typically 15 to 53 microns in diameter, produced by forcing molten Inconel through high-pressure inert gas jets in an oxygen-free chamber. The particles must have consistent morphology, controlled particle size distribution, and oxygen content low enough to avoid embrittlement in the final part. The powder must meet AMS5662 aerospace specification. A single contaminated or off-spec batch produces a part that looks correct and fails under load. No Indian producer currently makes this material certified to aerospace grade at the volumes a commercial launch programme requires. The powder that goes into Agnite arrives from abroad. When Agnikul built its Large Format Additive Metal Manufacturing facility — inaugurated in 2025 at IIT Madras Research Park, with Germany’s EOS as the industrial partner and an AMCM M 4K printer at its core — it built the most capable metal AM facility in India. The printer is Indian-operated. The powder that feeds it is not.

This is where the distinction between indigenisation and a sovereign supply chain becomes material. Indigenisation means local manufacturing and assembly: making the product in the country, reducing imports, building self-reliance. A sovereign supply chain goes further — owning the IP, the raw materials, and the process infrastructure end to end, such that no single foreign decision can interrupt production. Agnikul’s engine is an indigenisation achievement. The engine is designed in Chennai, printed in Chennai, tested in Chennai. But the powder, the printer architecture, and the heat treatment knowledge that makes the print flight-ready all originate outside India. That is not a criticism. It is a description of where India sits on the spectrum.

The second constraint is post-print processing. A freshly printed Inconel 718 part carries a microstructure shaped by rapid, directional cooling — columnar grain structure, residual internal stress, no precipitate phases. To develop the gamma prime and gamma double prime phases that give Inconel 718 its high-temperature properties, the part must go through solution annealing followed by double aging, with parameters calibrated specifically to the as-built microstructure. Those parameters vary with print settings. Developing and validating the right sequence for a new engine geometry requires instrumented trials, destructive coupon testing, and iteration. The tacit knowledge to do this reliably for rocket engine geometries does not yet exist at depth inside India — for the same reason it took Pratt & Whitney a decade and two hundred engineers to crack it in a different era.

The third constraint is qualification. An aerospace-grade component does not become flight-ready when the engineering is right. It becomes flight-ready when a certification authority has reviewed material traceability, process records, non-destructive testing results, and mechanical test data. Every change to the material supply chain — including switching from imported powder to domestic — restarts portions of that process. This is not an argument against the transition. It is a description of its actual timeline.

---

Who May Build the Bridge

The gap has two plausible routes to closure, and neither is close.

1. MIDHANI is the most natural candidate. At Aero India 2023, the company announced intentions to produce nickel and titanium-based alloy powders for aerospace additive manufacturing — a logical extension of its VIM capability and its existing CEMILAC relationships. Three years on, no certified aerospace AM powder has shipped from Hyderabad. The announcement established intent. The capex commitment, the atomization infrastructure, the inert-atmosphere powder handling lines, and the qualification programme that would follow are a different matter.

   With an order book hovering near the ₹2,500 crore mark, MIDHANI’s growth is no longer just speculative. It is anchored by the material logic of India’s two largest defense spenders: HAL’s pivot to indigenous engine alloys and the Navy’s expansion into advanced underwater platforms. The commercial pressure to prioritise that backlog over a long-horizon AM powder programme is real. Any AM related joint programme could change that and worth watching.

2. The second route runs through Agnikul itself. At a valuation above $500 million and with the most capable private rocket manufacturing facility in India, the company has the clearest incentive to close the feedstock gap. Vertical integration into powder atomization is a different business from engine design — the skills do not overlap, and the capex is not trivial. But if India’s launch sector reaches the cadence its ambitions require, the imported powder dependency will become an operational constraint before it becomes a strategic one. That tends to concentrate minds and unlock investment decisions that look premature in advance.

3. A third possibility, less discussed, is a specialist materials company that reads the demand signal early and builds atomization capability ahead of the market. India has produced this kind of bet before. A niche material, a stated deadline, a capability gap, and eventually a private actor who concludes the moat is worth crossing. The optical glass post tells the same story in a different sector.

---

Speciale Invest’s thesis frames India’s space ambition correctly: the founders competing in this space are doing so on engineering depth, not cost arbitrage, and the structural tailwinds including Space liberalisation, growing domestic satellite demand, a $44 billion addressable market.

The distinction between indigenisation and sovereign supply chain that sits at the base of every layer of the space stack still cannot be ignored. And underwriting both the tech and commercialization risk is necessary.

Agnikul has crossed the indigenisation threshold. The engine is designed, printed, and tested in India, and the US patent confirms that the process innovation is genuine. What it has not yet crossed is the sovereign supply chain threshold — where the powder, the printer architecture, and the process knowledge are also Indian. That gap is not a failure. It is the next problem. And it is the harder and painful one, because it requires building the kind of deep materials and process institution that Pratt & Whitney and MIDHANI built over decades.

But why is a material as seemingly common as glass treated with the same strategic gravity as a missile guidance system?

1. Why Borosilicate Glasses are Critical Targets?

In the theater of modern warfare, the difference between a hit and a miss is measured in microns. Borosilicate Crown (BSC) glass is the bedrock of military optics. Unlike commercial glass, defense-grade BSC-510644 is engineered for extreme clarity, specific refractive indices, and thermal resilience.

These glasses are critical indigenization targets because they represent a single point of failure in the supply chain. If an adversary or a global crisis cuts off the supply of these optical “blanks,” the production of sighting systems for tanks, snipers, and rocket launchers grinds to a halt. Indigenizing this glass isn’t just about saving foreign exchange; it is about ensuring that the Indian soldier’s point of aim remains true, regardless of global geopolitics.

2. Legacy Expertise and How it Grew

For decades, the world of high-end optical glass was a closed club, dominated by a few giants: Schott (Germany), Hoya (Japan), and various state-run institutes in Russia.

Digression: Story of the 41 Glassmakers

The history of this expertise is steeped in espionage and strategic evacuations. In 1945, as World War II drew to a close, the U.S. military realized that the technical brilliance of the Schott factory in Jena was too valuable to be left behind. In a mission known as “The Odyssey of the 41 Glassmakers,” they evacuated the top scientists and their families to West Germany. This move ensured that the “Trade Secrets” of the melt—the exact stirring speeds and temperature curves—remained in the West. It proved that in this industry, the process is more valuable than the patent. You can see this more in material tech than in say software. (Another example is the jet engine blade crystals with long learning curves in the metallurgy process over decades.)

These companies developed their edge through a century of trial and error, moving from ceramic pots to platinum-lined crucibles, slowly mastering the “fluid dance” of molten silica.

3. Synthesis and Process Hurdles

If the chemistry is the recipe, the synthesis is the “hell’s kitchen” of engineering. Manufacturing BSC-510644 presents three brutal challenges:

• Platinum Problem: At 1400 deg C, molten glass is a universal solvent—it “eats” ceramic containers. To maintain the “Six Nines” purity required for defense, manufacturers must use massive Platinum-Rhodium crucibles. This adds a staggering upfront CapEx and requires specialized induction heating.

• Boron Volatilization: Boron gives the glass its unique properties, but it does evaporate at high heat. If the surface of the melt loses too much Boron, the refractive index shifts, ruining the entire batch. Engineers must use counter-rotating platinum impellers to constantly fold the surface back into the liquid.

• Months-Long Cool Down: Precision annealing is the final hurdle. To prevent internal mechanical stress, a large block of glass might need to be cooled at a rate of just 1 deg C per day. One power failure or a slight dip in temperature, and the million-dollar melt shatters into worthless shards.

4. Who Can Invest in India?

With a CapEx requirement estimated in 100s of crores, this is a game for players with deep pockets and a long-term strategic vision.

For years, India’s CSIR-CGCRI in Kolkata functioned as the “R&D lab of the nation,” successfully melting these glasses on a pilot scale. However, the bridge from lab to factory remained unbuilt until recently. We are now seeing a shift where the government provides the “recipe,” and the private sector provides the “scale.”

Potential Investors:

• Paras Defence & Space Technologies: Currently the most aggressive private player in defense optics, they have already begun securing international partnerships to bridge the “process gap.”

• Borosil Limited: As the masters of commercial borosilicate in India, they possess the foundational chemical knowledge. A move into defense-grade melts would be a logical “value-up” for their specialty glass division.

• Asahi India Glass (AIS): With their massive industrial footprint, they have the balance sheet to sustain the long “learning curve” associated with high-precision melts.

As the December 2027 deadline approaches, the race to master the melt is on. The companies that successfully cross this “platinum moat” will be securing the eyes of the Indian Armed Forces.

Who we are

Spirelia is an AI native company we are building and is currently incubated at IIT Madras. We build agentic AI systems for innovation and R&D team typically at large organizations. They are responsible for figuring out what technology bets to make, and when, we help them make these decisions better with our expertise in innovation methods.

At present (April 2026), we have the first version of our first product, Utkarsha, that aids technology forecasting and roadmapping. The methodologies behind it has been developed and refined over years by researchers and practitioners. We have chosen 2. ATRA (Advanced Technology Roadmap Architecture) was built by MIT Prof de Weck during his time at Airbus. FORMAT was developed by Politecnico di Milano Prof Cascini and team including Dr Bala, our founder while working on an EU funded project for manufacturing technologies. Both of these methods are integrable and are mainly based on the intersection of few knowledge bases across patent intelligence, R&D strategy, and competitive foresight.

Why Spirelia

The view we take is both stage gating and innovation methods are structured enough for applying to a wide range of domains and helpful in the whole innovation process. Application of these methods is nuanced and we are constantly adding these skills to our AI products and helping humans in the loop, who will ultimately take better allocation decisions.

We started this substack because the conversations we have internally — about how AI is changing the way organizations make innovation decisions — deserve a wider audience. We anyway keep speaking to innovation leaders across industries and their views are worth spreading as well.

What to expect here

We write about AI in innovation and R&D strategy. Not the hype layer, but the basic/mundane/ops reality: how agentic systems can change technology decisioning, what it means to build auditable reasoning, what it means to be grounded in a forecasting workflow, where human judgment still matters and where it is being quietly replaced.

Expect one post every two weeks mostly analytics, case related, sector-specific technology reads. Some will be more direct dispatches from building AI products in the field.

If you work in an innovation function, lead R&D at a manufacturing or deep-tech company, or are thinking seriously about where AI fits into long-horizon strategy, this is written for you. We can only promise to keep it short, meaningful and worth the reading time. Being the story tellers we are expect at least one story per post.

If this is relevant to your work, subscribe or know someone who runs an innovation or technology strategy function, share this with them. Thanks for being part of the community.

Subscribe now