Svadharma in AI: Why Self-Funded Is Not Second-Tier

A senior partner at a US fund once told an Indian founder I know that his company was "promising for an Indian business." The founder was building infrastructure that two American companies in the same category later acquired their way into. The compliment was meant kindly. It was also revealing. There is a quiet assumption running through the AI economy — that the real path is venture-funded, US-headquartered, model-training, and growth-at-all-costs. Every other configuration is a tier below. Self-funded is plan B. India-based is a labour-arbitrage story. Services-led is what you do until you can productise. Infrastructure is what you build because you cannot raise to build models.

This is wrong in a specific way, and the Bhagavad Gita names exactly why. Krishna's argument to Arjuna in chapter three is not "follow your passion" or "be yourself." It is sharper than that. Better is one's own dharma, though imperfectly performed, than the dharma of another well-performed (BG 3.35, Eknath Easwaran). The translation that lands hardest is: even death in one's own dharma is better; the dharma of another brings danger. The principle is structural. Imitating someone else's path — however well — does not compound the way following your own does. The agent economy is going to make this lesson concrete for a generation of founders who borrowed a dharma that was never theirs to begin with.

The Borrowed Dharma Problem

The clearest case is the Indian founder raising a US-style mega-round to build a US-style foundation model company. The capital is there if the pitch deck is convincing. The talent is reachable. The cloud bills are payable. The work, technically, is possible. But every part of the operating system around it — the burn rate, the hiring tempo, the marketing posture, the exit horizon, the investor-update cadence — was designed for a company that has access to US enterprise procurement budgets, US-scale data sources, and a fund cycle that tolerates seven-year horizons. None of those structural inputs are the same in India. The work is performed well; the dharma is borrowed; the misalignment shows up in year three when the math stops working and the round that the model was supposed to bridge to never arrives at the terms the deck assumed.

The mirror image is the self-funded operator who tries to look like a venture-backed team. The four-person services business that hires a "VP of Marketing" because that is what real companies have. The bootstrapped infrastructure tool whose roadmap is built around imaginary funding announcements rather than real customer commitments. The solo founder who shadows a venture-funded competitor's hiring announcements and quietly feels behind. The work is the work, but the metrics are borrowed from a category that does not apply, and the operator spends more energy maintaining the costume than building the thing.

What Svadharma Actually Says

Svadharma — sva, "one's own"; dharma, "duty" or "the right action" — is the Gita's name for the work specifically assigned to you by what you are structurally, historically, and circumstantially positioned to do. It is not preference. It is fit. Arjuna's svadharma in the Gita is to fight, not because he wants to fight, but because he is a warrior on a battlefield with allies who depend on him and a war that will be lost without him. He cannot fulfil it by becoming an ascetic at the moment of crisis, even though renunciation is a higher path in the abstract. The right action is contextual to the actor.

For a builder, svadharma is the work you are positioned to perform better than someone reaching for the same outcome from a different starting point. A founder with 25 years of enterprise services experience has a svadharma rooted in long customer relationships and category expertise — not in three-month consumer growth loops. A solo operator with a stack of infrastructure ventures has a svadharma rooted in compounding small bets — not in mega-round product launches. An Indian builder with cost structure, talent depth, and infrastructure-building DNA has a svadharma rooted in being the picks-and-shovels of the agent economy — not the next OpenAI. None of these are lower-tier paths. They are different paths, and they compound differently.

The Economics Favour Duration

The agent economy will make the durational path more valuable, not less. Three forces are converging. First, the price of intelligence is falling faster than any input cost in technology history — inference costs per token dropped roughly tenfold between 2023 and 2025 and continue to compress. Second, the application layer is consolidating: every horizontal SaaS company is racing toward feature parity by bolting on AI agents, which compresses pricing power across the entire SaaS category by 2027-2028. Third, the infrastructure layer beneath the applications — identity, attestation, audit, observability — is structurally underbuilt because the venture model rewarded application bets in 2023-2024, not infrastructure ones.

The result: in 2027 and 2028, the venture-funded application companies built on the assumption of premium pricing will be repricing themselves into a commodity. The self-funded operator who built the same thing without raising a round will have lower burn, no exit pressure, retained equity, and customer relationships that survive the price compression. The infrastructure builders — many of them Indian, many of them self-funded, many of them invisible to US tech press — will be in the position Verisign was in during the 2001 dot-com collapse: charging steady rates for a service that did not get any less essential when the bubble repriced. None of those operators failed at someone else's dharma. They followed their own.

India's Svadharma in the Agent Economy

There is a national-scale version of this argument and it is the spine of chapter six of the book. India did not produce the global model labs. It did not produce the consumer AI breakout. It is not going to produce a credible competitor to OpenAI or Anthropic on the model-training axis, and the founders who try will burn capital learning that lesson. But India did produce UPI — twelve billion transactions a month on infrastructure no other country has built. It produced Aadhaar, the world's largest digital identity system. It produced ONDC, a protocol-layer commerce rail. It produced Zerodha, a self-funded brokerage that outscales venture-funded competitors. The pattern is consistent: India's svadharma is infrastructure that scales, not consumer products that go viral.

The agent economy needs infrastructure that no one is building. Identity layers for agents. Attestation systems for agent actions. Audit trails that work at machine speed. Cost-allocation infrastructure for businesses running thousands of agents. None of these are model companies. All of them are durable, infrastructure-shaped, and economically aligned with how Indian companies are structurally built. The founder who picks one of these and works on it for ten years will outlast three venture-funded application companies that try to do everything at once. The work is not glamorous. It compounds. That is the point.

The Three Tests

There is no algorithm for svadharma — the discernment is the work. But three tests filter most of the obvious cases. The first is vocabulary. If you keep describing your work using metaphors that were minted in a different category — "our flywheel," "our moat," "our 10x," "our category-defining round" — you are probably operating with borrowed language, which means borrowed expectations. Real svadharma is usually describable in plain customer language. We help X people do Y faster. If you need a deck to explain the work, the dharma is not yet clear.

The second test is cadence. Are your six-month milestones tied to customer outcomes or to a fund's reporting cycle? If the next decision is contingent on a round closing rather than on a customer signing, the operating tempo has been imported from a financing structure that is not yours. The third test is the absence test: if no one were watching — no Twitter, no LinkedIn, no investor updates, no peer founders comparing notes — what would you still build? The work that survives the absence of an audience is your dharma. The work that needs the audience to motivate it is not.

The Hardest Verse to Live

The Gita's most quoted line is BG 2.47: you have a right to the action, never to the fruits of action. The Sanskrit term nishkama karma — action without attachment to the result — is the operational form of this. It is the hardest verse to live in technology building because every part of the founder ecosystem is engineered to make you attached to the fruits: the metric, the round, the exit, the comparison. Svadharma without nishkama karma collapses back into outcome-chasing, just with a different vocabulary.

The self-funded operator who follows their svadharma but stays attached to looking successful to peer founders has not actually escaped the borrowed dharma — they have just changed the costume. The Indian infrastructure founder who follows their svadharma but measures themselves against US application valuations is still operating in a borrowed frame. The full move is: do the work that is yours to do, with the quality that the work deserves, without needing it to look like anyone else's work, and without needing it to produce the outcomes that someone else's work produces. The Gita's promise is not that this is easier. It is that this is what compounds. The other path does not.

Chapter 8 of The AI Agent Economy is the long version of this argument — dharma, nishkama karma, and viveka as operational frameworks for the agent economy, with the case studies and verse citations the format of a blog post cannot carry. The shorter version is this: stop building someone else's company. There is no second-tier path. There is only your dharma, performed with full attention, in the only category where you can credibly compound. That is the work. Everything else is a costume.