The cost of building a company is approaching zero
Every generation of technology has reduced the cost of starting a business. We are entering the generation that reduces it to near zero. Not the cost of running a business indefinitely, but the cost of going from idea to operating company. The barriers that once required teams, capital, and months of setup are collapsing one by one.
In 2000, you needed office space, servers, a development team, legal counsel, an accountant, and six months before you had anything to show. In 2010, AWS and Stripe cut the infrastructure cost, but you still needed a team. In 2020, no-code tools let solo founders build products, but operations still required human labor. In 2026, an AI cofounder handles operations, playbooks handle workflows, and the infrastructure runs itself. The only remaining cost is compute, and compute gets cheaper every quarter.
Three eras of companies
The transition from human-run to agent-run companies is not a single leap. It is three distinct eras, each with fundamentally different economics and organizational structures.
| Dimension | Human-Run (pre-2015) | Software-Assisted (2015-2025) | Agent-Run (2025+) |
|---|---|---|---|
| Minimum team size | 5-10 people | 1-3 people | 1 person + agents |
| Time to launch | 6-12 months | 1-3 months | 1-7 days |
| Monthly operating cost | $50K-500K | $5K-50K | $100-500 |
| Scaling model | Hire more people | Better tooling | More agents + playbooks |
| Operating hours | Business hours + overtime | Extended by automation | Around the clock |
| Knowledge retention | In people's heads | In documents + tools | In playbooks + state |
| Quality consistency | Varies by person and day | Improved by processes | Every output graded |
| Failure mode | Key person leaves | Tool breaks | Playbook produces poor output |
We are at the start of the third era. The companies built this way will look nothing like the companies of a decade ago: smaller in headcount, faster to launch, more consistent in output quality, and dramatically cheaper to operate. The full implications of this shift have not been absorbed yet.
Companies for one
The most immediate impact is that a single person can run a real company. Not a freelance practice. Not a side project. A company with a product, coordinated marketing, customer support, financial operations, and strategic planning, all handled by an AI cofounder and a roster of specialist agents.
This is the founder Capx Casa is built for. The founder sets the strategy, makes the judgment calls, and approves the high-stakes decisions. The agents handle everything else. A single person with clear thinking and good playbooks can compete with a 20-person team that has neither.
The micro-company explosion
When the cost of running a company drops below $500 per month, the calculus changes for millions of people. Ideas that were never worth hiring for become worth trying. A niche newsletter serving 2,000 subscribers. A specialized consulting practice in a narrow vertical. A marketplace connecting two small communities. None of these justify a team. All of them justify a playbook. We expect the number of companies in the world to increase by an order of magnitude within the next five years, and most of the new ones will be agent-run.
Companies at scale
The more counterintuitive implication is what happens above the single company. When an AI cofounder can coordinate a roster of specialist agents, the bottleneck shifts from execution to strategy. A founder with a proven playbook library can launch a second company in days, not months. Then a third. Then a tenth.
The cost of failure drops so low that the rational strategy becomes trying more things, not deliberating longer. Launch a company to test a market hypothesis. Run it for 30 days. If the unit economics work, scale it. If they do not, shut it down and start the next one.
Portfolio founders
From there, the portfolio founder follows directly: one person running multiple agent-run companies simultaneously, each targeting a different market or customer segment. The playbooks are reusable across companies. The operational patterns transfer. The founder's job becomes capital allocation and strategic direction across a portfolio, not day-to-day execution in any single company. This is not a venture fund. It is one person running five or ten operating businesses, each with its own AI cofounder.
Companies anyone can own
Ownership is the second half of the thesis. When a company's operations are run by agents and logged end to end, its performance is observable in a way no private company has ever been. That makes a new kind of ownership possible: tokens that trade on open markets, backed by businesses whose operating record is visible rather than promised. Evaluation moves from pitch decks and quarterly reports to live data.
The demand side of this already exists. Seven companies tokenized through Capx, and their tokens cleared more than $65M in on-chain trading volume, generating roughly $100k per month in protocol fees. Those tokens trade on Raydium, on Solana, and are surfaced through Capx Terminal. The next step is connecting that market to companies built on Capx Casa: tokenization will be opt-in and milestone-gated, a graduation a company earns, not a default it starts with.
Democratized access
Traditional startup investing requires being an accredited investor, having deal flow, and committing large minimums. Tokenized companies on open markets are accessible to anyone, and they can be evaluated on operating data instead of a pitch. This is not about replacing venture capital. It is about creating a parallel system where operating businesses are ownable by default, without gatekeepers.
The Capx thesis
This is not a bet on any single model or provider. Models will keep improving. Costs will keep falling. New capabilities will emerge. What will not change is the need for structured operations, quality control, cost management, and financial infrastructure. Those are the layers Capx provides, and they are durable regardless of which model is state-of-the-art next quarter.
Where this goes
The infrastructure for agent-run companies is early, and the direction matters more than any single feature. Three threads we think about constantly:
An economy of agent-run companies
Today, each company operates independently. The interesting future is companies transacting with each other: one company's marketing agent engaging another company's content company for a project, settled automatically. Not just agent-run companies, but an agent-run economy.
A market for playbooks
A playbook that consistently produces excellent output is real operational capital. Founders should be able to share and sell proven playbooks, so a new company can launch with capabilities that took someone else months to refine.
Governance that fits the business
As agent-run companies take on more complex operations, governance needs to evolve beyond approval queues and spend caps. Different kinds of businesses need different approval workflows, risk thresholds, and escalation paths, encoded once and reused.
The market side of this future has already cleared its first $65M. The build side is opening now. The cost of building a company is approaching zero. The question is not whether this will happen. It is whether you will be building the companies or watching someone else build them.