At the AI Ascent 2025 discussion hosted by Sequoia Capital, partners Packer Radio, Sonya Huang, and Konstantine Buhler delved deeply into the latest trends, business potential, and entrepreneurial guidance in the AI industry. They emphasized that AI is no longer a future imagination, but a reality wave of unprecedented scale and opportunity.
Why is AI Important? How is it Reshaping the Economy and Work Patterns?
The AI Market is Larger than the Cloud: A Major Industrial Transformation Across Software and Labor
Packer Radio began by paying tribute to the classic framework proposed by Sequoia Capital founder Don Valentine, raising questions such as “What is AI? What impact does it bring? How should we respond?” He predicted that the market size for AI transformation would be ten times larger than that of the software era of cloud transformation, and that the market development over the next 10 to 20 years would be “absolutely massive,” simultaneously impacting the profit markets of “services” and “software.”
From evolving into tools to becoming “co-pilots” and even “autopilots,” AI will not only reshape services but also disrupt the entire software industry and labor market. He added, “The shift from selling tools to selling outcomes has opened up competition in both total addressable markets (TAMs).”
(CEO Nadella: Majorana 1 Drives Breakthroughs in Quantum Computing, Cloud Services Become the Biggest Winner in the AI Industry)
The Next Wave Created by AI: The Whole World is Already Onboard
Packer reminded everyone that since the moment ChatGPT was introduced, AI has fully entered the practical stage. A groundbreaking technology requires computational power, network coverage, data scale, and a user base, all of which are already in place:
Mainstream Narratives Across Eras
Given that the necessary elements for this revolution have been constructed, the adoption of AI will face almost no barriers when the starting gun fires, marking it as the fastest and widest technology diffusion case in history.
How Startups Can Win the AI Race: The Application Layer as a Value High Ground
Packer’s chart indicated that the most successful tech companies in history have almost all emerged from the application layer (Apps), and the AI era will be no different; true value will accumulate in startups that understand the “customer perspective”:
White Space: The Application Layer Remains an Underdeveloped and Highly Potential Innovation Area
Focusing on specific industries or concrete functions to solve problems through a vertical strategy and establishing competitive barriers through human-machine collaboration is currently the most promising competitive track. He cited, “The first batch of AI ‘killer applications’ has already emerged, including ChatGPT, Harvey, Glean, Sierra, Cursor, and A Bridge.” Sonya specifically mentioned deep applications in “medical diagnosis, voice assistants, educational concepts, and advertising visualization,” providing entrepreneurs with some direction.
Sequoia’s AI Investment Guide: From Revenue Flywheels to Four Technical Filters
When selecting AI-related investment targets, Sequoia emphasized that 95% of the criteria are the same as for other industries, but the unique 5% for AI focuses on four key indicators:
– Substantial Revenue: Distinguishing genuine business growth from short-term “trial” companies by focusing on product adoption rates, user engagement, and retention.
– Trust with Customers: At this stage, customer trust in the company is more important than the product itself; trust is the foundation for long-term success.
– Potential for Gross Margin Improvement: As AI computing costs decrease, companies must demonstrate concrete practices to ensure healthy gross margins.
– Effective Data Flywheel: The data flywheel must be directly related to business metrics; otherwise, it cannot form effective competitive barriers.
These are moats that cannot be easily replicated and will become important criteria for assessing whether an AI startup has long-term competitiveness.
Towards the Era of AI Agency Economy: AI is No Longer Just a Tool, but a Digital Partner
Konstantine then revealed that the training of large models is slowing down, shifting towards seeking new breakthroughs, including OpenAI’s “reasoning” and Anthropic’s “Model Context Protocol (MCP),” which attempts to create new methods for expanding applications.
(AI World’s USB-C Interface: What is Model Context Protocol (MCP)? Interpretation of an AI Assistant’s Universal Context Protocol)
Among them, the next significant wave of AI will be the “agency economy”:
AI technology is evolving from a single-task executor into an “agent” capable of collaboration, communication, and transactions. The future will move towards a true “agent economy,” where these intelligent agents will work alongside humans to create a highly automated and trusted digital economy.
(AI Agency Combined with Stablecoins: How PayPal Rewrites Global Business Models with Its Own Financial Operating System?)
Three Technical Barriers of the Agency Economy: Identity, Communication, and Security
To truly realize the agent economy, AI must solve three major technical hurdles, which will become new battlegrounds for future technological innovations:
– Establishing a Stable and Persistent Identity System: Agents need to maintain consistent personality and understanding, continuously comprehending user needs; significant challenges remain in memory and self-learning.
– Developing Seamless Communication Protocols among Agents: Agents need to establish a TCP/IP-like protocol layer for personal computers to transmit information and value.
– Constructing Trust-Based Security Mechanisms: In a scenario where the trading counterparties are all agents, the importance of security and trust must be elevated, potentially giving rise to a security industry centered around trust and safety.
(From Financial Advisors to Secretaries: The Trust Challenge of AI Agencies—Can We Trust Artificial Intelligence’s Autonomous Decisions?)
Redefining Work and Economy: A New Era of “High Leverage and High Randomness” Brought by AI
Konstantine concluded that AI will not only reshape work patterns but also change people’s thinking models, shifting from the “certainty” of traditional computer science to a “stochastic mindset,” as both AI and humans have uncertain memory and responses. At the management level, managing AI agents will require more complex decision-making, including process blocking and feedback, which itself is a discipline.
It is not hard to imagine that the future will be an era of “a few driving the majority of production”; we have already seen many companies expanding at unprecedented speeds with fewer personnel. We will be able to do more, but we must also manage the uncertainties and risks involved. Ultimately, these processes and AI agents will merge, forming a vast neural network that will reshape personal work, businesses, and even the economy, ensuring that no one is left behind.
Risk Warning
Cryptocurrency investments carry a high level of risk, with prices potentially fluctuating dramatically, and you may lose your entire principal. Please assess the risks carefully.