Capital Without Labour: A Founder's Guide to the Changing VC Equation

By every measure of the conventional startup playbook, I should be spending my evenings refining pitch decks and my mornings cold-emailing VCs. I should be optimising for a Series A. Instead, I'm trying to understand a structural shift in the economics of building software - one that changes when capital creates value and when it destroys it. Getting this question wrong is fatal for founders and investors alike.

I spent thirteen years inside large corporates and Government, so I have some instinct for institutional inertia - the kind of magical thinking that mistakes activity for outcome. When I look at the Venture Capital industry today, I wonder if that machine is facing a structural challenge that its own incentive structures make extraordinarily difficult to confront.

My thesis is uncomfortable but simple: for twenty-five years, successive technology waves have been eroding the economic equation that justified venture capital's role in software. Pay-per-click advertising, open source, WiFi, cloud computing - each removed a cost that startup capital used to fund. The industry absorbed every wave. Generative AI, I believe, is the wave it cannot fully absorb, because it eliminates the last and largest cost in the equation - the human labour of building the product itself.

What follows is what happens when that assumption breaks - and an honest attempt to work out what it means for founders like me who may need capital for reasons the old model never anticipated.

A Brief History of Venture Economics

The venture capital industry likes to trace its lineage to noble origins - post-war innovation, garage startups, visionary risk-takers. The mythology is appealing. The economics are more revealing.

The Supply of Capital

Through the 1970s, VC was a cottage industry, concentrated in California, of perhaps 40 firms, collectively managing a few billion dollars. What transformed it from a niche into an industry was a regulatory event. In 1978, the U.S. Department of Labor reinterpreted the "prudent man" rule under the Employee Retirement Income Security Act (ERISA), effectively permitting pension funds to invest in venture capital for the first time. The effect was immediate and staggering: annual VC commitments leapt from $39 million to $570 million in a single year - a 1,460% increase. By 1986, pension fund money accounted for over 50% of all venture capital.

The industry didn't grow because of a sudden breakthrough in capital allocation. It grew because a regulatory change unlocked an ocean of institutional money.

Apple's $1.3 billion valuation in December 1980 poured kerosene on the fire. VC firms proliferated from roughly 40 to over 650 by the end of the decade. Capital under management ballooned from $3 billion to $31 billion. The pattern was set: a small number of spectacular exits would generate the "returns narrative" that attracted more capital, which attracted more firms, which deployed more money, which - crucially - required more deals. The industry began optimising for deployment volume, not deployment quality. This distinction matters enormously for what comes later.

The Quiet Deflation

But there is a parallel story the supply-side narrative misses: every decade or so, a technology wave quietly reduced what startups actually needed the money for.

Alan Jones - an Australian tech investor who has been in startups since 1995 - recently walked me through this history, and it is worth recounting in full, because VC's survival through prior waves is both a genuine achievement and perhaps a contributing factor to its current blind spot.

Pay-per-click advertising replaced expensive, unattributable offline campaigns. In the dot-com era, a meaningful fraction of startup capital went to brand-building through channels where you could not measure whether any of it worked. PPC made customer acquisition measurable and scalable - the first major cost item to collapse.

Mobile apps meant product marketers could send push notifications and pull users back repeatedly, rather than spending a fortune on PPC just to recapture users during their limited window on a desktop.

Open-source software eliminated licensing costs for databases, operating systems, and development platforms that had previously run to six figures.

WiFi - trivial as it sounds - eliminated the need for expensive office leases. Jones estimates that up to 25% of a seed round once went to premises, at a time when teams couldn't move around a building without recabling the Ethernet network. WiFi meant three cofounders could work around a table at a coffee shop.

Cloud computing replaced the capital expenditure of buying and managing servers with operational expenditure. Before AWS, a surprisingly large chunk of a Series A went to hosting infrastructure, managed by specialists whose scarcity commanded premium salaries.

And the meta-infrastructure of raising itself improved: convertible notes, SAFEs, platforms like AngelList and ProductHunt - each reduced the friction and cost of the capital formation process.

Each wave was real. Each materially reduced the capital required to reach $10 million in ARR. And each time, VC adapted: round sizes adjusted, fund structures flexed, the model absorbed the shock and carried on. The industry has earned its scepticism of "this time is different."

But notice the pattern. Each wave eliminated a peripheral cost - marketing spend, licensing fees, office space, server hardware. None touched the core: the human labour of writing the software. Regardless of how cheaply you could market, host, or distribute your product, you still needed ten engineers to build it and fifty salespeople to sell it.

Boom, Bust, and the Muscle Memory of Excess

The dot-com era was the industry's first real stress test - and it should have been a warning. Between 1995 and 2000, annual VC investment surged from roughly $8 billion to over $100 billion. Much of it went into companies with no revenue, no business model, and no plausible path to either. Pets.com, Webvan, Kozmo.com - the roll call of the dead is familiar enough to be cliche.

When the bubble burst, the NASDAQ lost 78% of its value. Hundreds of funds closed. The rational conclusion would have been that the industry had a structural problem with capital discipline - that the incentives to deploy (management fees) overwhelmed the incentives to deploy wisely (carried interest on actual returns).

Instead, the industry rebuilt. The ZIRP era (2009-2022) turbocharged the cycle. Capital poured into VC. Fund sizes ballooned. Annual deployment averaged about $375 billion globally in the four years to 2021 and one prominent growth-stage fund deployed capital into 315 startups in a single year - a pace so frantic that each investment received roughly 28 hours of analysis before a cheque was written. This wasn't incompetence; it was the logical endpoint of a model that rewards deployment volume. When the results were disclosed, the fund placed in the bottom 10% of all venture funds raised that vintage.

The "spray and pray" strategy was the reductio ad absurdum of a model that had been optimising for capital deployment over capital allocation for forty years. Was it also the peak?

The Returns Distribution

The mythology of venture capital rests on a powerful narrative: that VCs are uniquely skilled at identifying the future. The data tells a more complicated story.

In 2012, the Kauffman Foundation - one of the largest institutional LPs in the United States - published "We Have Met the Enemy... And He Is Us." The findings:

• Only 20 of 100 VC funds generated returns exceeding a public market equivalent by more than 3% annually.
• 62 of 100 funds failed to exceed the S&P 500, after fees.
• Since 1997, less cash had been returned to investors than had been invested in venture capital.

The majority of VC funds would have been outperformed by a passive index fund.

The industry's defence is the "power law": a few extraordinary winners compensate for the losers. Cochrane's NBER analysis found mean VC returns of 57% per year - but with "extraordinary skewness." Kaplan and Stromberg calculated the median Public Market Equivalent at 0.90 - meaning the median venture capital fund loses to the S&P 500. The mean is pulled up by a handful of outliers most LPs will never access.

The problem compounds with scale. Carta's 2024 analysis of recent $1 million to $10 million funds are posting better results than recent funds of more than $100 million:

• Small funds ($1M–$10M): For the 2017 vintage, the median IRR was 13.8%.
• Large funds (Over $100M): For the 2017 vintage, the median IRR was 9.8%.

This trend of smaller funds producing higher IRRs - at the 25th, median, 75th, and 90th percentiles - also holds true across the 2018 and 2019 vintages.

Yet the industry raises ever-larger funds, because management fees scale with AUM, not performance. A GP managing $200 million fund (committed) collects perhaps $3.5 million per year - regardless of returns. This is paid whether the fund returns 10x or zero. The median GP commitment is only 3% of fund size. The people deploying the capital have almost none of their own money at stake.

When your incentive structure rewards deployment over returns, you will find reasons to deploy. My economics training has taught me to always look at incentive structures - as incentive structures are destiny.

The Last Treadmill

Strip away the TED talks, and venture capital was solving a legitimate economic problem whose components have been eroding for decades. Marketing costs fell with pay-per-click. Infrastructure costs fell with cloud. Office costs fell with WiFi. But one component stubbornly remained: technical complexity required large teams, and large teams required upfront capital before revenue.

Building software in 2010 meant hiring backend engineers, frontend engineers, DevOps, QA, a product manager, a designer, and - once the product existed - an army of Sales Development Representatives to drag prospects through a pipeline. Reaching $1 million in ARR typically required 10-15 employees. Reaching $10 million required 50-100.

By 2020, the hiring treadmill was the last major cost that justified equity financing - and it was still enormous. Raise a seed round for your first engineers, a Series A for your sales team, a Series B to scale both, and a Series C because you were burning $3 million a month and profitability was still a distant aspiration. The treadmill was real. The capital need was real. VCs were the only source of risk capital willing to finance it.

So, what happens when building software no longer requires an army?

GenAI and the Core Cost

Each previous cost-reduction wave removed a line item from the startup budget. GenAI does something qualitatively different: it collapses the cost of the product itself - the engineering labour that was the last and largest component of the equation. An industry that comfortably absorbed the elimination of server costs and office leases is now facing the elimination of the thing those servers and offices existed to support.

A new species of startups

Consider the decoupling of headcount from revenue:

• Midjourney: ~$200M revenue, 11 employees, zero VC.
• Cursor: ~$100M ARR, fewer than 20 employees.
• Bolt.new: $0 to $20M ARR in two months, ~15 people.

Even when VCs invest (as a16z did with Cursor), they aren't funding a hiring treadmill. They are funding compute or velocity.

These companies are exceptional, and I owe you honesty about that. Just as I'll argue later that VCs suffer from survivorship bias - seeing only companies that seek VC and concluding VC is necessary - I risk the same error by celebrating the winners.

For every Midjourney, there are thousands of AI startups that wrapped an OpenAI API, charged a subscription, and were crushed when the platform moved upmarket. Roughly 90% of AI startups fail - significantly higher than the ~70% rate for traditional tech. When Microsoft embedded Copilot into PowerPoint, it rendered an entire category of presentation AI tools obsolete overnight. Tome, which had raised $81.6 million, abandoned its core product entirely.

Midjourney itself is a dangerous template - a B2C creative tool with viral Discord dynamics, structurally different from the B2B enterprise products most founders are building. Its founder, David Holz, was burned by his previous venture (Leap Motion raised $100 million and sold for a fraction of that). Using Midjourney as a blueprint for B2B SaaS is like citing Apple's garage as a startup strategy: true, but statistically irrelevant to the median founder.

My thesis rests on the structural shift, not on cherry-picked survivors. The cost of technical execution is collapsing toward the marginal cost of compute. A single experienced engineer with AI coding tools can now produce the output of a 10-person team. That shift is real even if most founders who ride it will fail - just as 75% of venture-backed companies never return cash to investors.

I should also throw some caution about the maturity curve. I'm building at the cutting edge of AI, and the frontier models are very, very impressive - but, as Gibson liked to remind people, the future is already here; it's just not evenly distributed.

In the broader sense, AI agents are gaining significant traction, though quality remains the biggest barrier to deploying them into production. Data from LangChain shows that more than half of surveyed professionals already have agents running in production, with an additional 30% actively developing them. While exact failure rates vary, organizations with over 10,000 employees specifically cite hallucinations and consistency as their biggest challenges. Instead of generalized team-shrinking metrics, current evidence shows that large enterprises are prioritizing agents for internal productivity, with coding agents currently dominating day-to-day use. Ultimately, the industry is moving past the proof-of-concept phase; as we enter 2026, the question for most organizations is no longer whether to build agents, but how to do so reliably and at scale.

If the reliability curve flattens at 90% rather than 99%, "one founder replaces ten engineers" collapses to "one founder replaces three." The current evidence suggests 2-5x productivity improvements are real - enough to shrink the Minimum Viable Team from 15 to 3-5, which still dramatically reduces capital requirements.

I am projecting a trajectory, not describing a completed transition: the direction is clear; the magnitude and pace are not.

The Scope Caveat: Software, Not Atoms

There is an important objection I should address early: my thesis is about software, not atoms.

While software VC struggles for relevance, "Deep Tech" VC is booming. You cannot "vibe code" a nuclear reactor. Building physical things still requires large teams, specialised equipment, and years of development before revenue. VC remains well-suited to atoms. VC as an industry isn't dying. It's migrating - from software to hardware, defence, and deep science.

For a solo software founder in 2026, that migration is worth noting. If the best VC firms are pivoting to defence tech and biotech, the remaining software-focused VCs are a smaller pool. The question becomes not just whether you need capital, but whether the capital that remains in software VC is the right capital for what you're building.

Second order effects

It's not just engineering. The administrative overhead of running a company - legal, compliance, HR, finance - is collapsing too. SOC 2 compliance used to take months and cost tens of thousands. AI-powered platforms now automate evidence collection, policy drafting, and continuous monitoring for a monthly subscription. The entire "back office" is collapsing into API calls.

This matters because the VC pitch was never just about engineering - it was about the total cost of operating a company at enterprise scale. When that total cost falls by an order of magnitude, the economic justification for dilutive equity financing weakens with it.

The death of the SDR - long live the SDR

If engineering was the primary cost of building, sales was the primary cost of scaling. The SDR - responsible for cold emailing and lead qualification - was a staple of B2B SaaS. VCs poured billions into hiring SDR armies to brute-force growth. AI agents can now prospect, personalise, and qualify leads at near-zero marginal cost.

But there is a ceiling, and I won't pretend otherwise.

Enterprise buyers do not buy mission-critical infrastructure from a solo founder, no matter how elegant the code. They buy accountability. A CIO at a Fortune 500 company buys Salesforce not because it's the best CRM, but because if it breaks, there is a $30 billion company with D&O insurance, cyber liability coverage, and the financial substance to honour an indemnification clause. Procurement frameworks explicitly assess vendor financial stability. A solo founder cannot indemnify a bank against a data breach.

I spent thirteen years inside the kinds of institutions that buy enterprise software - like the ATO and later Qantas Loyalty. I have sat on the buyer's side of procurement. I know that a CIO's signature on a contract is a career risk, and that no amount of product elegance overcomes the question: "Will this vendor exist in three years?" This is not an abstract objection to me. It is the specific challenge I am navigating.

In an era of deepfakes and synthetic content, the demand for verified, capitalised vendors may actually be hardening. This is perhaps the strongest argument for capital's continued relevance to enterprise software.

But the assumption that VC backing solves this trust problem is questionable. Over 3,200 VC-backed companies shut down in 2023. A company with $50 million in the bank and $4 million in monthly burn is eighteen months from death - and the CIO who signed a three-year contract discovers this only when the emails stop being answered. Mercuri International's research finds that 99% of enterprise buyers cite trust as crucial, and the top signal is reliability and consistency - not funding status. A profitable bootstrapped company with stable pricing may outperform a VC-backed competitor on a ticking clock.

Bootstrapped companies can and do sell to enterprise - but it takes longer. Atlassian bootstrapped from Sydney for eight years before taking outside capital and now serves 83% of the Fortune 500. Zoho generates over $1 billion annually without a dollar of external investment. In Nassim Taleb's framework, these founders had skin in the game - their wealth directly tied to the company's success, creating alignment that no term sheet replicates. But the path is measured in years, not quarters. Anyone who tells you otherwise is selling something.

The Founder's Calculus

This changes the fundamental arithmetic. Why sell 20% of your company for $3 million to hire engineers when AI tools can do the work for $500 a month? Why give up board seats to finance a hiring treadmill that no longer exists?

Mailchimp bootstrapped to a $12 billion acquisition by Intuit without ever taking VC. Basecamp has operated profitably for over two decades without a single outside investor. These were once eccentric choices. In the GenAI era, perhaps they are the rational default.

The Incentive Trap

Here is the part that, as an economist, I find most fascinating. Venture capitalists are professional pattern-matchers - people whose value proposition rests on seeing the future before everyone else. And yet, the structural incentives of fund management create predictable cognitive biases that make this particular shift extraordinarily difficult to confront from the inside.

This is a case of institutional incentive misalignment - the same forces that Kahneman and Tversky documented almost fifty years ago.

The industry has heard "this time is different" at least five times before - and each time, it was right to be sceptical. VC absorbed pay-per-click, open source, cloud, mobile, and the operational efficiencies of SAFEs and AngelList. This track record of successful adaptation is precisely what makes the current wave maximally dangerous. When the wolf finally arrives, the village has been trained not to listen.

Clayton Christensen's work on disruptive innovation established the pattern: incumbents fail not because they are incompetent, but because their competencies are optimised for a world being replaced. Applied to VC: the industry's core competencies - relationship-driven deal sourcing, pattern-matching on founder profiles, concentrated bets, hands-on board involvement - were optimised for a world where startups needed $15 million and 50 employees. In a world where a solo founder with an AI stack can reach significant revenue, these competencies may become impediments. The VC who insists on a "proven team" of 10 risks filtering out the founders who've figured out how to build without one.

Status quo bias compounds the problem. The deal flow machine - incubators, warm intros, pitch competitions - is a filter designed to select founders who fit the 2015 model. A solo founder saying "I don't plan to hire" pattern-matches to "lifestyle business," not "hyper-efficient scale."

I've confessed my own survivorship bias earlier - I cited Midjourney and Cursor while thousands of AI wrappers died in obscurity. VCs suffer a structural version. They only see companies that seek venture capital - a sample pre-filtered to include only founders who believe they need VC. When your entire sample is drawn from companies that chose the VC path, you will naturally conclude the path is necessary. It's like concluding that umbrellas cause rain because everyone you see with an umbrella is wet.

The management fee structure reinforces the dynamic. With over $2.51 trillion in global dry powder, the pressure to invest is enormous. Raise funds, deploy under pressure, find companies that need capital, conclude capital is necessary, raise more funds. I'm describing these incentives from outside the system, which makes them easy to see. If I were a GP with $200 million to deploy, I suspect I'd find compelling reasons to keep deploying it. The incentive structure makes it genuinely hard to ask: "What if founders don't need our money anymore?"

The Compute Capital Shift

Startups are not becoming capital-free. They are shedding human labour costs but accruing a different cost: compute.

For AI companies, inference costs now average 23% of revenue. Traditional SaaS enjoyed 80-90% gross margins. AI-native companies often see margins compressed to 50-60% - looking less like software businesses and more like industrial operations with significant COGS.

But this changes the type of capital startups need. Inference costs are predictable, scale linearly with usage, and represent a known operational expense - not an uncertain bet on product-market fit. Using equity that demands 10x-100x returns to pay a predictable AWS bill is economically irrational: financing certainty with risk capital.

The market is experimenting. Revenue-based financing from firms like Tractor Ventures offers non-dilutive capital based on recurring revenue. I wouldn't be surprised if there's plenty of others.

The Structural Mismatch

The core observation still holds: we are moving from a world where the primary capital sink was human labour (risky, best financed with equity) to one where it is compute (more predictable, better financed with debt once revenue exists). The VC fund structure - fee models, governance expectations, 10-year lifecycle - was built for the former. It fits awkwardly in the latter.

The "missing middle" of venture capital - $10 million to $50 million growth rounds designed to scale the team - is thinning because there is less team to scale.

The Barbell Effect

Where does this leave the industry? A barbell that squeezes the middle.

At the low end, solo founders bootstrap with customer revenue and RBF. The dilution is hard to justify when your burn rate is measured in API costs.

At the high end, foundation model companies require multi-billion-dollar commitments funded by sovereign wealth funds and hyperscalers. Traditional VCs are too small to play.

The missing middle - $10M to $50M growth rounds - is thinning. If you don't need to scale from 15 to 150 employees, the Series B becomes harder to justify.

There is a compelling investor-side counter-argument here, and it deserves honest engagement. If building is cheaper, a $500K seed cheque buys what $3M used to buy. VCs can make more bets with smaller cheques. The portfolio diversifies. The risk per bet drops. From the investor's chair, AI cost compression may actually be excellent news. Alan Jones put it well: "My venture dollar goes further, and each startup investment I make is risking less of my total capital."

This repricing thesis has real force at the seed stage. But the downstream economics are less clear. If those seed-stage companies reach $5-10M ARR without needing growth capital, the VC's ownership stake never scales to produce fund-returning outcomes. The investor who makes 20 smaller, smarter seed bets still needs those companies to grow into outcomes that justify the fund structure. When the companies are capital-efficient enough to reach profitability on their own terms, the path from seed investment to venture-scale return narrows.

The traditional "value-add" is also facing competition from AI tools that perform competitive analysis, source talent, and generate market intelligence - though how quickly this erodes remains an open question. For the elite firms, the brand signal alone may continue to justify the relationship.

There is a further structural consideration: market fragmentation. As building costs drop, we see an explosion of "Micro-SaaS" - vertical products built by solo founders for specific niches. A Micro-SaaS generating $5 million a year with 90% margins is life-changing for a solo founder. But it is structurally uninvestable for a $500 million fund that needs billion-dollar exits. If the ocean fills with millions of profitable tuna, the whaling ships face a scarcity of whales - even as the ecosystem flourishes.

The Distribution Paradox

There is a demand-side consequence my argument has been glossing over, and it deserves its own reckoning.

Herbert Simon wrote in 1971: "A wealth of information creates a poverty of attention." When anyone can build a CRM in a weekend, 10,000 people will. The scarce resource shifts from the ability to build to the ability to be heard.

The data supports this. B2B SaaS acquisition costs are rising 14% year-on-year. a16z now argues that "momentum," not product quality, is the primary AI-era moat.

The obvious conclusion: capital migrates from the "Build" bucket to the "Sell" bucket. You still need capital. Case closed.

This objection has real teeth.

Enterprise sales is not B2C virality. A CIO at a Big Four bank does not discover software through word-of-mouth; they discover it through Gartner Magic Quadrants, peer CIO networks, and procurement-approved vendor panels. The sales cycles are 6-18 months, involving legal review, security assessments, and reference checks that no AI agent can shortcut. Building the credibility infrastructure for enterprise - certifications, cyber insurance, analyst relations, a small but senior sales team with domain expertise - costs real money. I am building a System of Trust designed to deliver provable, defensible AI analysis, and the irony is not lost on me that selling trust requires demonstrating trustworthiness - which in enterprise procurement is partly a function of capitalisation.

This is the strongest version of the argument for capital's continued relevance in software, and I take it seriously.

But there are important qualifications.

First, the nature of the distribution spend is changing. Not $15 million for 50 SDRs running a cold-email playbook, but perhaps $2-3 million for credibility infrastructure deployed with precision. That is a fundamentally different capital conversation - different in amount, in instrument, and in what the investor should expect in return.

Second, the data on organic distribution is real:

• Word-of-mouth generates 5x more sales than paid advertising, with 2-3x higher ROI
• Product-led growth companies achieve CAC of $100-$500 vs $5,000-$50,000 for sales-led, growing 50% YoY vs 21%
• Google Ads CPC increased across 86% of industries in 2024; nearly 60% of searches end without a click

Third, the history of "growth at all costs" is littered with expensive lessons. Quibi raised $1.75 billion. It died in six months. WeWork raised $10 billion. It collapsed from $47 billion to $2.9 billion in eighteen months. Meanwhile, bootstrapped SaaS companies grow at parity with VC-backed ones while spending one-quarter the CAC, and are three times more likely to be profitable within three years.

And there is a signalling argument worth making. ZIRP made VC funding a cheap signal - deployed at such volume that raising a seed round signalled little more than "I showed up." Michael Spence's signaling theory predicts exactly this: signals anyone can send cheaply become worthless. Profitability, by contrast, is extraordinarily hard to fake. As Paul Graham observed, the critical question is whether you are "default alive" - whether you'll reach profitability before running out of money. In a market drowning in noise, that may be the strongest signal available.

But let me be clear: organic distribution works on a longer timeline. Atlassian bootstrapped from Sydney for eight years before its IPO. Anyone promising instant enterprise traction without capital is selling a fantasy.

When Capital Creates Value

My argument is not that capital is unnecessary. It is that the reasons for taking capital, the amount required, and the instruments best suited to provide it have all fundamentally changed.

There are situations where equity capital still creates genuine strategic value for a software founder:

Winner-take-all dynamics. When network effects create a land-grab and the cost of being second is death, speed matters more than capital efficiency. If you are racing to establish a standard before a hyperscaler does, the calculus changes.

Credibility capital. When your buyer requires vendor financial substance - and in regulated industries, they often do - capital on the balance sheet is not vanity. It is a procurement requirement. This is a real constraint, particularly for enterprise software in financial services, healthcare, and government.

Speed-to-trust. When the market window for establishing a defensible position is narrow, organic growth may be too slow. Being first to establish a trust standard in a new category - provable AI analysis, for instance - could be a position worth buying speed to reach.

Strategic access. When investor networks provide channel access that no amount of product quality can replicate - government procurement panels, enterprise CIO networks, specific industry verticals - the equity trade may buy distribution that revenue alone cannot.

The question is not "VC or no VC?" It is: "What am I buying with this equity, and is the price rational?" For a $500/month compute bill, equity is absurd. For credibility infrastructure that unlocks enterprise procurement, it may be essential.

So, Where Does This Leave Us?

I started this piece with a question: as a solo technical founder, how should I think about venture capital?

The VC model was built for a world where software was denominated in human labour. That world is ending - not suddenly, but as the terminal phase of a decades-long deflation in startup costs that the industry has previously absorbed. Whether this wave crosses the threshold from "another repricing" to "structural break" is the genuinely hard question, and I don't claim certainty about the answer.

What I do think is that the default is flipping. For decades, the assumption was that a serious founder needed VC. The burden of proof was on the bootstrapper. Now, the burden should increasingly be on the capital provider to explain why equity - the most expensive form of capital, carrying the most onerous governance obligations - is the right instrument for a company whose primary costs are compute and credibility, not headcount.

The most interesting founders in the GenAI era don't look like traditional deal flow. They are solo or micro-team operators generating meaningful revenue without ever entering a pitch competition. They don't need $15 million for hiring - they might need $2-3 million for credibility infrastructure, deployed with a governance model that respects their capital efficiency. The investors who develop pattern recognition for this new archetype will access a class of founder that the traditional funnel systematically excludes.

Here is what I would not buy with equity: engineers I don't need, SDRs running a playbook AI can execute, office space for a team that doesn't exist. Here is what I might buy: the credibility infrastructure to pass enterprise procurement - certifications, insurance, analyst relations, a small senior team with domain expertise and enterprise relationships. The capital conversation worth having is not "how much do I need to hire?" It is "what is the minimum capital required to convert technical capability into enterprise trust, and what is the expected return on that specific investment?"

The deepest irony may be that an industry built on identifying disruption is struggling to recognise it in its own mirror. But I should be careful about declaring victory from the outside. I'm a solo founder in Canberra placing a bet that might not pay off, building Enterprise Software while navigating the very credibility gap that system is designed to solve. The question is whether enough founders make similar bets to shift the equilibrium - and whether enough investors are willing to meet them on new terms.