The startup paradox: how small teams are racing to build the world's most powerful technology

"Image synthesis assisted by Mai Image 2.6 (t2i), an AI partner within the Global Future Nexus ecosystem."

The race to Artificial General Intelligence is often framed as a contest between a handful of trillion-dollar corporations—OpenAI, Google DeepMind, Anthropic, and their peers. Yet beneath the surface of this narrative, a quieter and perhaps more consequential dynamic is unfolding. Startups, many founded by a single visionary or a small team, are raising unprecedented sums of capital, targeting domains the giants have overlooked, and forcing a fundamental question: can the most transformative technology in human history be built by companies that did not exist a year ago?

The Capital Tsunami

The financial scale of AGI startup activity is staggering. In the first quarter of 2026 alone, AI companies raised US$242 billion globally. OpenAI secured US$122 billion, Anthropic US$30 billion, and xAI $20 billion. But the concentration extends deep into the startup ecosystem. In China, DeepSeek raised a 51 billion RMB A round, while StepFun and Moonshot AI raised 23 billion and 19 billion RMB respectively—three companies accounting for over 70% of the top-ten funding total.

This capital is flowing into increasingly specialized domains. "Physical AGI" and "world models"—AI systems that understand and act in the physical world—have become the hottest investment themes. GigaAI, a world-model company, raised 3.5 billion RMB in just three months across three funding rounds, becoming China's first world-model unicorn with a valuation exceeding 10 billion RMB. Across the ecosystem, 25 world-model-related funding events raised over 2.2 billion RMB by late March 2026. Fei-Fei Li's World Labs raised US$1 billion, and Yann LeCun left Meta to found AMI Labs, securing a record US$1.03 billion seed round.

The Physical AGI Frontier

The strategic logic behind this shift is a growing consensus that text alone cannot reach AGI. As LeCun has argued, "the existing LLM route is completely wrong; simply predicting text, AI will never touch human-level intelligence. We need models that can understand physical reality". The Beijing Academy of Artificial Intelligence's 2026 trends report formally endorsed this view, proposing a paradigm shift from "Next Token Prediction" to "Next State Prediction"—predicting the next state of the world rather than the next word.

This has catalyzed a wave of startup activity. Meta acquired Assured Robot Intelligence, a 20-person startup building AI models for humanoid robots, as part of its push toward "physical AGI". RJ Scaringe, founder of Rivian, raised US$500 million for Mind Robotics. In China, startups like Octopus Dynamics, Lingchu Intelligence, and Boundless Dynamics raised tens of millions of dollars each for embodied general brains and physical AGI.

The Governance Reckoning

But the startup explosion has collided with an uncomfortable reality. In September 2026, a researcher named Jacob Coxon resigned from Anthropic, warning that the leading labs were "racing toward self-improving superintelligence without acting responsibly." He wrote: "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt."

This warning triggered an unprecedented response. Anthropic CEO Dario Amodei published a 4,000-word essay calling for a coordinated slowdown, and was swiftly endorsed by Sam Altman, Demis Hassabis, and Elon Musk. Amodei's proposal included a three-step plan to slow advanced model development and a call for a new self-regulatory body—modeled on FINRA, the financial industry's private regulator—to impose safety standards across the sector.

The tension is structural. Startups are incentivized to move fast, but the technology they are building may require a level of caution that is incompatible with venture capital timelines. Amodei himself acknowledged that the industry would need a "limited exemption" from antitrust laws to coordinate safety measures—a request that raises profound questions about who governs the governors.

A new class of startups has emerged to address this governance gap. The Artificial Intelligence Underwriting Company, founded by an early Anthropic hire and former METR COO, has built a third-party audit and certification layer for AI agents, using SOC 2 as its model. Companies like Zenity have raised US$125 million to police autonomous AI agents, while AI Score raised US$5.4 million to help organizations govern the safe use of AI.

The Path Forward

For Global Future Nexus, the AGI startup ecosystem represents both the engine of innovation and a critical governance challenge. The concentration of capital, the shift toward physical AGI, and the emergence of a safety-and-governance startup sector all point toward a future where intelligence is not built by a single entity but distributed across a complex, competitive landscape.

The question is not whether startups will continue to play a central role in the AGI race—they will. The question is whether the governance frameworks being developed—voluntary standards, third-party audits, self-regulatory bodies—are sufficient to manage the risks of a technology that its own creators believe could end human civilization. The startup paradox is this: the same nimbleness that allows small teams to pioneer new frontiers also makes them structurally resistant to the caution the moment demands.

Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)

Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)