On May 28, the development company of the AI large model Claude, Anthropic, announced the completion of a $65 billion Series H financing, with a post-financing valuation reaching $965 billion, surpassing its competitor OpenAI ($852 billion), becoming the world's most highly valued private AI company, once again highlighting the global capital's fervent pursuit of AI.

When trillion-dollar giants wrestle closely on the compute foundation, what opportunities remain for ordinary startup teams located in the application layer? Where will the true division of labor in the AI industry between China and the United States head? With these questions in mind, recently, Jenny Yang, founder and CEO of Starlabs Consulting, spoke with Steven S. Hoffman, founder and CEO of the leading American startup incubator Founder Space, who is known as the "godfather of venture capital" in Silicon Valley.

Hoffman is a serial entrepreneur and venture capitalist, as well as a bestselling author, having written several acclaimed bestsellers, including "Let the Elephant Fly," "Survival Rules," and "The Five Forces of Innovation." He is also a sought-after keynote speaker worldwide and has long served as a strategic advisor to governments, well-known corporations, and incubators in multiple countries.

As a seasoned venture capitalist who has deeply mentored thousands of startups globally, Hoffman provides an extremely calm, honest, and visionary business deconstruction of the current AI frenzy.

Here are excerpts from the interview between Jenny Yang and Hoffman:

The true turning point for autonomous agents will arrive within two years at the earliest

Jenny Yang: You just concluded a trip to China. Please share your overall impression of China's AI technology, AI companies, and the current state of AI applications. What differentiated roles do you think Silicon Valley and China will play in the next phase of AI competition?

Hoffman: My overall impression is that China is moving forward rapidly, extremely fast. The Chinese startups I spoke with are integrating AI into every aspect: payments, logistics, customer service, human resources, marketing, sales, procurement, manufacturing, and more.

Meanwhile, I believe Silicon Valley will continue to dominate the foundational research of cutting-edge large models. The concentration of computing power, top talent, and capital in the United States remains unmatched for now. However, China will excel in application deployment. Chinese companies are exceptionally good at scaling a technology at an astonishing speed and transforming it into commercial products that have real users and real scenarios. This pragmatic attitude and efficient execution are where China's strengths lie.

China also boasts powerful top AI laboratories, including Moonscape, Alibaba, ByteDance, and DeepSeek. These labs will play an extremely sharp "fast follower" role, closely following the pace of their American counterparts. Although these labs may not have particularly abundant capital, they always find innovative methods to minimize costs, thereby promoting global expansion of platforms.

Additionally, China holds an absolute leading position in the robotics field. No place in the world possesses such a complete supply chain, infrastructure, and talent pool to support the large-scale production of robots. The next phase of the AI competition is by no means a winner-takes-all scenario. Silicon Valley will continue to build the most powerful technological engines, while China will build the most outstanding commercial ecosystem and robotic hardware. Both are equally important.

Jenny Yang: Do you think AI has national boundaries? In the current context of tightening global data sovereignty and AI regulatory policies, do you prefer companies that deeply cultivate the local market or those that are globally oriented from day one?

Hoffman: From a technological perspective, AI knows no boundaries; but from reality, global regulatory policies are rapidly defining boundaries. Data sovereignty laws, national security reviews, model export restrictions… these are reshaping the global compliance framework.

Some founders, seeing this trend, conclude that they should cultivate a single local market. I understand this logic, but I absolutely do not agree.

I firmly support Global from Day 1; the reason is simple: those companies that seek to establish a foothold in the local market first and later pursue overseas expansion have almost universally found themselves in difficulties. This is due to differences in distribution channels across countries, varying compliance requirements, and even the need to completely reshape brand positioning, which is not only costly but also slow to progress.

In contrast, global-first companies build modular and highly adaptable systems from day one. They design underlying architecture directly addressing regulatory differences rather than waiting for problems to emerge for remedial action. They can attract international teams that understand diverse markets, which translates to a lasting structural advantage.

Admittedly, the difficulty of compliance is increasing day by day, and enterprises also need to introduce localized compliance systems. However, the key to breaking through is to build flexible architecture rather than to be content with a single corner. Market opportunities are global, and the ambitions of every tech entrepreneur should be the same.

Jenny Yang: You have pointed out that we are still in the very early stages of the AI revolution and that the eruption of autonomous agents will completely overturn existing business paradigms. Based on your observations, how far are we from that day? In the face of structural unemployment challenges triggered by AI, what preparations can we make in business models or systems?

Hoffman: That day is very close to us—closer than most people imagine, but a bit farther than the media hype suggests. Autonomous agents capable of handling independent, clearly defined specific tasks have already emerged, such as automated customer service, code review, data analysis, research compilation—these are no longer just demos; they are already in commercial use.

The real turning point—when different agents can achieve self-coordination, handle ambiguous multi-step goals, and operate across systems without human oversight—will likely be another 2 to 4 years away, perhaps even sooner.

When that wave truly arrives, labor replacement will be cold and real, and it is no exaggeration.

The solution is certainly not to slow down the pace of AI but to ensure that social mechanisms keep up with the iteration speed of AI technology.

• On the business model front: The smartest founders are designing their companies around "Human-AI Collaboration," rather than "pure automation." In their models, humans handle decision-making, creative output, and accountability, while agents are responsible for workload and efficiency. This model is more resilient to risks and better supports the development of team personnel.
• On the policy level: We need to face the issues of retraining, social security systems, and education reform candidly. The roles being replaced are not merely low-skilled positions but lawyers, analysts, copywriters, consultants, and virtually all knowledge-intensive roles. This fundamentally alters the underlying logic of social governance.

Jenny Yang: You have pointed out that traditional consulting, intermediary, and "Humans as a Service" business models are difficult to achieve real scalability due to high marginal costs. But now, with AI increasingly replacing and automating professional intelligence services, does this mean that AI-driven knowledge services will break the curse of scalability for HaaS businesses?

Hoffman: The traditional consulting industry has long faced a conundrum: to grow the business, you must increase headcount; as headcount increases, costs rise, ultimately squeezing profit margins and stalling scalable expansion. This is the inherent trap of the HaaS model.

However, AI has fundamentally changed this underlying equation. Nowadays, a senior consultant fully armed with AI agents can provide analytical work that previously required a small team to accomplish, meaning that the marginal cost of adding new clients plummets dramatically. This is unprecedented.

So, yes, AI-driven knowledge services finally have the capability to break the curse of scalability. But the premise is that enterprises are willing to restructure their organizational framework. The companies that will thrive in this transformation will not treat AI merely as an efficiency tool but will comprehensively reshape their entire business system around the underlying AI framework.

Startups should focus on scenario innovation

Jenny Yang: Regarding open source vs. closed source, from the perspective of Founder Space and venture capital, do you prefer to support applications that are deeply tied to giant closed-source ecosystems or independent projects built on an open-source ecosystem? Why?

Hoffman: In the United States, I am optimistic about applications built on the ecosystems of leading cloud providers (including AWS, Azure, and Google Cloud). These platforms have complete distribution channels, enterprise-level trust, and deep integration capabilities, which are necessary for business scalability. Developing on these large platforms allows inheriting many inherent advantages: security compliance, stability commitments, and global infrastructure support. Open source is certainly exciting, but "excitement" won't help you secure enterprise-level contracts.

However, the situation in China is different. The cloud ecosystem there is primarily shaped by Alibaba Cloud, Tencent Cloud, and Huawei Cloud, and policy and regulatory environments determine which platforms enterprises can choose. In China, open source models like DeepSeek are garnering significant market attention because they allow Chinese companies to operate independently without relying on foreign infrastructure. In this context, open source is not just a concept; it is an inevitable strategic choice.

Thus, the correct answer entirely depends on where you build your product and whom you intend to sell it to.

Jenny Yang: In a context where compute power and algorithms are monopolized by giants, how can early AI startup teams effectively identify and capture those demand pain points that truly have scaled commercial prospects and are not easily diminished by the giants?

Hoffman: Tech giants will inevitably commodify generic underlying technologies; that is beyond doubt. If what your startup is doing can be implemented within six months as a new feature by OpenAI, Anthropic, Google, or Microsoft, it is not a business but a feature point in someone else's product roadmap.

To survive in such an intensely competitive environment, startups must focus on highly specialized areas with deep scenario relevance. For example: a workflow requiring acute understanding of a particular industry, a compliance solution relying on specialized knowledge not possessed by the foundational large models, or a customer relationship that takes years to establish trust.

Diving deep into a specialized field vertically serves as a defensive barrier for startups. The more a solution relies on practical experience from industry experts (surgeons, supply chain managers, actuaries, etc.), the more difficult it becomes for industry giants to replicate quickly.

Ultimately, speed is the most critical moat for early-stage companies. Your iteration speed must exceed that of giants completing competitive product initiation and budget approvals internally. By the time those giants react, agile startups will have already established their own brand and solidified their market leadership—this means they have built a rapidly growing user base, exclusive data, and mature products that truly fit the market.

Jenny Yang: With the development of generative AI, there is a surge of AI forgery and fraudulent information. From the perspectives of cybersecurity and anti-AI fraud, does this present a highly potential avenue for entrepreneurs?

Hoffman: Yes. Today, producing synthetic media has become utterly accessible, making voice cloning, deepfake videos, and simulated AI phishing emails an increasingly serious nightmare.

The defensive measures in the cybersecurity industry lag far behind the methods of attacks, and this pain point represents a market opportunity. Detection tools, trace verification, digital watermarks, identity authentication, and other fields all hold immense entrepreneurial potential. Businesses and government agencies need such solutions, and the financial industry needs them more, as they are suffering financial losses from various AI fraud activities.

However, it is important to note that detection models can only defend against known types of attacks; therefore, startups must recognize this adversarial characteristic from the outset of development and ensure that their products are capable of continuous learning and dynamic iteration.

If a startup team excels in both generative AI and cybersecurity, it has the opportunity to create a multi-billion dollar company aimed at addressing the current surfeit of deepfake technologies.

Web3 + AI may be a trap

Jenny Yang: What do you think a founder needs to possess in terms of underlying thinking, which is different from the past, to lead a team to create the next generation of unicorns in today’s era, filled with technological anxiety and capital fervor for AI?

Hoffman: Forget all your past notions about "barriers." In today’s industry environment, a product from 18 months ago may already be obsolete. Entrepreneurs who make it to the end have long recognized this.

First, replace functional thinking with systemic thinking. The next unicorn cannot emerge solely around a clever prompt. It must be built on a network of agents, a data flywheel, and a multi-party integration system, relying on long-term compounding effects for growth.

Second, stay attuned to real user needs. AI significantly boosts development efficiency but can also easily lead products away from practical applications into self-indulgence. Excellent entrepreneurs will always focus on core user demands; blindly iterating off-course will ultimately lead to internal strife.

Third, recruit highly adaptable talent. Currently in-demand skills may no longer be relevant in two years. Firms need to build continuously learning teams rather than merely execution-oriented ones.

Fourth, do not fear technology. Many entrepreneurs see AI as an elusive black box. You must have enough understanding of it to accurately know what it can and cannot do. This awareness itself is your competitive advantage.

Jenny Yang: You have previously mentioned that there is excessive hype surrounding many enterprise applications of blockchain outside cryptocurrencies, while AI is the truly universal foundation that touches every industry. Nowadays, many Web3 companies are trying to merge AI with Web3. Do you think "Web3 + AI" is a promising entrepreneurial direction?

Hoffman: I'll be frank: Web3 does have real value, but mainly for those already immersed in the crypto circle. Decentralized finance, asset tokenization, and intermediary-free cross-border settlements are significant for this specific group, but they constitute a tiny fraction of the global economy.

For ordinary business clients, small and medium-sized merchants, and the general population, the situation is different. I don’t believe Web3 can drive substantial momentum in the mainstream market, and I have never been a proponent of it; my views have not changed after several years of development.

Most consumers and businesses do not need blockchain to achieve their business goals. What they need are stable, reliable products, excellent user experiences, and reasonable prices. Web3 adds friction, increases complexity, and introduces regulatory risks. For ordinary consumers and users, Web3 does not offer what they truly need.

On the contrary, AI is the genuinely universal underlying technology. It can reach into every industry; nearly all sectors can leverage pattern recognition, automation, and intelligent decision-making to solve practical problems. This represents a fundamentally different value proposition.

Forcing the combination of Web3 and AI will not multiply the value of either but will only increase complexity. For most founders, this is not an opportunity but a trap. Of course, AI may benefit those already deeply tied to the Web3 ecosystem, but for the broader consumer market, it will not lead to any substantial change in user adoption or industry development trajectories.

Jenny Yang: We noticed that you announced an ambitious nonprofit plan at the beginning of 2026—to establish research centers in 10 universities worldwide aimed at training future leaders on how to ensure AI reflects core human values. Can you share the current progress of this plan? What "responsible innovation" ideas do you hope to convey to future AI entrepreneurs through these centers?

Hoffman: Our vision is to establish specialized research centers in ten universities globally, and there is still a significant gap to reach this goal.

We are still in the very early phase, with most energy focused on fundraising. Because before we can implement it, we must ensure we have the necessary resource support. Building truly substantive, sustainable projects in universities requires tangible funding investment.

Driving us forward is a simple belief: every young person entering the workforce today will spend their entire career in a world where AI is integrated into every product, service, and business. However, most of them are ill-prepared for this massive change. Our research center aims to change that.

We want the next generation of entrepreneurs to not only understand how to use AI to build products but also to know how to ensure the AI products they create align with human values; to learn to anticipate the various secondary and derivative impacts arising from technological deployment; and to innovate responsibly while retaining their ambition.

That is our mission, and we are moving towards this goal.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。