How we think about supply-side moats in a world where code costs nothing
When you book a blood test in an Indian city, your sample's journey typically moves through several hands: a local franchisee collects the sample for a commission, who then passes it to an aggregator, who finally sends it to a third-party lab. Because of this, the brand on your report rarely maintains direct control over your sample from start to finish. In this unorganised segment, which handles most of India's testing, the entire experience hinges on freelance collection agents and labs of uneven quality.
Orange Health, our portfolio company in the at-home diagnostics space, looked at this supply chain and decided not to participate in it. It delivers a jaw-dropping customer experience because it runs its own accredited labs and employs its own rider fleet. A phlebotomist employed by them reaches you within an hour; the report has a ~6-hour turnaround time; and because every step from your arm to the report happens inside one company, the result can be fully trusted by the consumer. A wrong or late report has exactly one owner - and that accountability is what the consumer is paying for.
In developed markets, reliability and speed are seldom treated as competitive advantages, as they are baseline expectations. In India, however, that assumption of a seamless infrastructure is not inherent. It has to be earned. This is because the Indian ecosystem is a highly fragmented landscape of millions of disconnected local operators.
And thus, for years, our thesis move has been the intermediary one: take a trusted middleman, make them more capable with technology, and let the network become the moat.
We had seen this across our portfolio - SolarSquare grew on homeowner referrals - rooftop solar is an expensive decision made in a low-trust market, so a neighbour who can point to a working system on their own roof does more to close the next sale than any advertisement (90% of its customers say they would recommend it, and that word-of-mouth has carried it to roughly 50,000 homes). Meesho initially grew by leveraging existing social trust - in its early model, Meesho recruited housewives and homemakers as resellers - so the trust was borrowed from an existing relationship rather than built by the brand from scratch. (We did a deep dive on Meesho’s journey)
While that thesis still holds, it’s worth noticing that those are demand-side solutions. They solve how you win the next customer: through accumulated trust. However, when we look back at what has actually compounded in our portfolio and in the ecosystem, a less glamorous and far harder-to-copy pattern sits underneath, ie controlling & integrating deeply into the supply side. If you try to shortcut India’s operational landmines, you eventually blow it. The companies that we’ve seen compound are the ones that stopped looking for a workaround and just built the infrastructure themselves.
SolarSquare's referrals and Orange Health's repeat customers came after each company had done the unglamorous work of controlling its supply side: like owning the install teams & after-sales service in SolarSquare’s case, and owning the labs & the rider fleet in Orange Health’s. The satisfied customer who recommends a brand is thus the output of a supply chain that delivers an experience that doesn't fail them. Demand-side love is earned on the supply side. For Meesho as well, the housewife-reseller network surely made them a household name in the initial years, but it was fixing the fulfilment that made the value proposition of price sustainable at scale.
For most of the last 15 years, global public markets paid a persistent premium for asset-light business models - the capital-efficient, software-like companies that scaled without owning much. Think Uber or Airbnb, which grew to enormous scale without owning a single car or hotel room. This was rightly seen as the smarter model: renting the assets meant you could grow faster, burn less capital, and earn software-like margins. Many became hugely successful, which is precisely why the market paid up.
India has also run this experiment repeatedly. Consider the 2015 hyperlocal grocery wave: TinyOwl and PepperTap were demand-first businesses that spent millions to acquire customers while renting the entire supply side from third parties they couldn't control. Their business models were built on a foundation of discounts and aggressive user acquisition mechanisms, and this soon led to the unit economics collapsing. Because they owned no inventory and controlled no delivery, they had no operational leverage - and ultimately, no reason for a customer to stay once the discounts disappeared. Contrast that with today’s quick commerce giants - Blinkit, Zepto, and Swiggy Instamart. They learned the lesson that you cannot outsource the "quick" in quick commerce. Instead of relying on a fragmented relay of third-party vendors, they built an integrated stack: they own the dark stores, they stock the inventory, and they employ the rider fleets. By owning the full chain, they’re both a software and an infrastructure company with supply-side moats. That is why they are defensible.
This shift - from renting capacity to owning the operational stack - has become a necessary survival strategy in a market where the "software-only" advantage is rapidly evaporating. Today, AI has crushed the cost of building software - what took a funded team a year can now be approximated by 1 person and their agents. When the cost of building falls, so does the defensibility that came with it. But what AI cannot commodify is a heavy yet low-obsolescence operating layer of the physical world: the labs, the installer network, the classroom systems. You can't prompt your way to a phlebotomy fleet.
And this is no longer only an India story. Globally, asset-heavy companies have begun to outperform asset-light ones for the first time in over a decade. We’re seeing a shift in market sentiment that mirrors our own thesis - in an AI world, durability comes from owning your supply chain, not outsourcing it.
You can't consolidate away India's mess; you can only decide who absorbs it
A reasonable question from anyone who hasn't operated in India: why not just fix the chain? Consolidate the suppliers, standardise the process, remove the mess?
After a decade of investing, we’ve come to realise that India's fragmentation is structural, not a phase of development it will pass through. Production is distributed across 63mn+ MSMEs; the average Indian factory employs around ~40 workers, and over 80% of the workforce operates informally - outside contracts, compliance systems, and institutional oversight. Preferences, languages, and regulation shifts happen from state to state. We've written before about why this long-tail structure is a feature of the economy, not a bug.
So when the complexity doesn't disappear, it has to land somewhere, and there are only three places it can go:
on your customer (the delayed report, the botched install, the failed delivery),
on a chain of suppliers and partners (who can't fully be relied on)
on the business itself.
Most companies dodge the mess because fixing it is expensive and slow, but that’s a race to the bottom. To compound in the long term, businesses should absorb the complexity by solving for the supply side deeply: ie instead of renting whatever fragmented capacity exists, you build or control each link yourself and to your own standard.
The lab is yours, so accuracy is yours to guarantee. The install team is yours, so quality is yours to guarantee. That is what going full stack on the supply side means in practice: owning the chain of production and delivery deeply enough that by the time anything reaches the customer, every failure point has already been handled inside your own walls.
What supply-side ownership buys you
Owning the supply-side is the cost, but it purchases a business three distinct advantages.
1. The build cost scares off the field
The hardest requirement to meet is usually the one that scares everyone away: high upfront capital, deep engineering, and years of patient building. These aren’t "move fast and break things" markets; they are markets where you have to build things that last. The incumbents often don’t have to fight off challengers because very few serious players bother to show up.
Rooftop solar demonstrates this. A residential install sits on someone's roof for twenty years; installation quality is wildly inconsistent across the market, and after-sales service is mostly fictional. SolarSquare chose to own installation, service, and the on-ground trust-building. That choice demands capital, engineering depth, and operational patience up front. Tellingly, the only other company doing well in residential rooftop solar is Tata Power - a legacy conglomerate with decades of brand trust behind it, and even it largely operates through dealer and installer networks rather than owning the install end to end.
2. Capital buys speed, but it can’t buy hindsight
Even when a company manages to enter these markets, they can’t shortcut the process. A new entrant can raise money, buy a fleet of bikes, and launch an app. But they cannot shortcut the "hindsight" that only comes from years of running an operation.
Orange Health didn't invent at-home diagnostics - Dr Lal, Apollo, and others were already there. But while incumbents focused on brand, Orange Health focused on the plumbing: factories, reverse logistics, factory-like lab precision - and then the brand. A second-mover challenger can simulate the service, but they can’t simulate the operational maturity. The real advantage is the years of accumulated knowledge around regulatory clearances, the logistics know-how required to keep blood samples viable in monsoon traffic, and the specific lab workflows that shave hours off turnaround time.
3. Customers who anchor to you.
This is the demand-side payoff of the two supply-side moats. When almost nothing in the market can be relied on, the operator who always delivers stops being just another vendor and becomes essential infrastructure. Customers start building their own routines around it - the way you'd depend on electricity or a bank - because switching to someone unproven isn't worth the risk.
Consider our portfolio company, LEAD. By owning the entire classroom stack - from curriculum and teacher training to assessments - they standardise education in regions where quality is otherwise unpredictable. Because LEAD becomes the school's operating system, removing it would require a complete reconstruction of the school's day-to-day rhythm. This deep integration is exactly why their partnerships endure for years.
AI lifts the operational ceiling
With code no longer the differentiator, AI has simultaneously unlocked the ability to solve the one constraint that has historically prevented supply-heavy businesses from scaling: the coordination tax.
The historical barrier to scaling ops-heavy businesses has been coordination, not capability. Each new city, site, or lab adds a sprawling layer of scheduling, supervision, quality control, and planning. As the physical footprint expands, management overhead balloons and margins erode under the weight of this coordination. By contrast, software models operated without this tax - they scaled to millions of users at near-zero cost, requiring no new physical infrastructure. For two decades, capital naturally flocked to that frictionless model. Today, AI is finally attacking that structural constraint. By absorbing the administrative burden of scheduling and planning, AI can make these hard businesses easier to scale from the inside.
Regular readers will recognise what comes next: over our last few editions, we built two theses for AI in India - Orchestration AI and Vertical Workflow-Integrated AI. Here is why they matter specifically for supply-first businesses:
In plain terms, Orchestration AI is software that runs the planning and coordination that a growing operation used to need armies of managers for - scheduling, sequencing, tracking who does what across many sites at once. It attacks exactly the coordination tax described above, as it provides the coordination layer that complex and multi-part operations always lacked. MyGenie, our portfolio company in construction, is a great example: its AI analyses material costs, worker productivity, and project timelines to create optimised schedules across multiple sites at once - planning work no human project manager could efficiently hold in their head, delivered at a near-zero marginal cost. The contractor still builds, but AI does the coordinating that previously required an expensive management layer.
Vertical Workflow-Integrated AI removes the administrative drag inside professional workflows. CloudPhysician - a market example - deploys a system that monitors ICU video feeds alongside patient records and coordinates alerts across monitoring data and care protocols, effectively automating the connective work between systems that used to consume clinicians' hours. In practice, that lets a handful of specialists monitor many hospitals' ICUs remotely, instead of each hospital needing its own on-site intensivist. The professional keeps doing the expert work; AI absorbs the admin around it.
Owning operations creates the raw material for an AI advantage - the data, the workflows, the coordination surface - but the advantage only materialises for companies that deliberately build the AI layer on top of it. An asset-heavy business that ignores AI keeps the old scaling problem: coordination costs that grow with your operational footprint. This is now a standing conversation across our portfolio - the supply-first companies that pair their operational depth with an AI coordination layer are the ones that convert a defensive moat into a compounding one.
For a decade, the ambition was to build above India's mess - to abstract it away behind an app. The companies that lasted did the opposite: they went down into the complexity and chose to own it.