Huadong Wang: Small Companies, Short-Lived Companies, VC, and the Young People of the Future

By Wang Ran, Founder of China Renaissance

The aftershocks from OpenAI continue to reverberate, but there's no longer any doubt about AI's inflection-point significance for humanity's trajectory. Today (November 21), China Renaissance founder Wang Ran published an essay sketching out his vision of the AI-reshaped future of business: the rise of what he calls "small yet big" and "short yet big" companies, and how these two new archetypes will transform today's venture capital industry and reshape employment prospects for young people.

Wang has recently been running what he calls a "GPT investment banking experiment" inside China Renaissance. After a period of training, large language models led by GPT-4 have reportedly helped the firm's analysts significantly boost their productivity, and broadened their perspective in matching projects with investors, especially overseas institutions. While AI can't yet replace humans, "within two to three years, replacing half or even two-thirds of foundational analytical work is entirely plausible."

As the founder of a homegrown investment bank, Wang has scarcely missed a beat on any major business topic in recent years. So at a moment when people feel both exhilarated and anxious about AI, his essay may offer some fresh angles worth considering.

The only constant is change.

As the internet made efficient distribution, collaboration, cloud deployment, and computing possible; as blockchain made authentic, effective, and irreversible ownership verification and value transfer possible; as AI and large models make AGI possible — going forward, two types of companies will increasingly have the chance to stand out. One is the "small company" that does big things with few people. The other is the "short company" established to fulfill a short-term mission.

"Small Companies" That Are Small Yet Big

Since a massive amount of future AI models, compute, and data will reside in the cloud, cloud services will become public infrastructure as ubiquitous as water and electricity — infinitely fragmentable and available to all of society. Small teams leveraging such infrastructure to achieve outsized commercial outcomes will become far more common. A team of just a few people can fully harness this infrastructure-ized cloud AI capability, develop distinctive AI agent products and solutions around specific markets or needs, then generate data flywheels through real-world application and achieve continuous, uninterrupted iteration.

Therefore, in domains that don't require labor intensity, "small yet big" companies will spring up like bamboo shoots after rain. Such companies certainly won't have the capacity to build foundational AI models, but they can create enormous value for users or clients by simultaneously understanding AI, understanding scenarios, and understanding data.

Historically, there have been overseas cases where tiny companies of just twenty or thirty people — even a dozen or so — were acquired by platform companies for over a billion or even several billion dollars, but these were rare exceptions. In the future, such cases will likely increase dramatically. Many tasks that previously required substantial manpower — such as developing and testing programs, data analysis, text and blueprint generation — can now be handled by AI, opening up far broader possibilities for small companies to create outsized value.

The greatest challenge these "small yet big" companies will likely face is the acquisition and accumulation of native data. Massive datasets typically sit with governments and large platforms, but this congenital weakness isn't entirely insurmountable. As AI converges with cloud services, platforms that own data must open it up and let it flow to build data ecosystems; increasingly, data is being integrated into various cloud services by the platforms that hold it. Small companies lacking native data scenarios can leverage this infrastructure to compensate for their early-stage deficiencies.

"Short Companies" That Are Short Yet Big

Traditionally, when we spoke of enterprises, we celebrated century-old shops and evergreen businesses. Such enterprises needed to build moats through products, technology, and brand; to ensure cohesion and continuity through culture, values, and mission. Let's call these "long companies." Their core value lay in creating shareholder value across historical and economic cycles. The most common external paean to the best of these: if you'd invested one dollar in this company in year xx, it would be worth tens of thousands today.

Going forward, such enterprises will still exist, and will remain the dream and pursuit of many entrepreneurs and investors. But simultaneously, another type of enterprise will emerge. They will no longer chase evergreen status or aspire to become century-old institutions. They will simply be hunting parties — rapidly and efficiently organized when prey appears, concentrating firepower for a swift strike, then departing with the spoils and not a cloud in their wake, until the next prey emerges. This bears some resemblance to blockchain-based DAOs.

Today, certain MCN companies that build influencer streamers and novel marketing channels on social media platforms like Douyin and Xiaohongshu already exhibit "short company" characteristics, as do some short-video and short-drama production houses. Beyond these, "short companies" will also have substantial room to grow in experiential entertainment and tourism project development, idol economies, art auctions and value-sharing, decentralized finance, and other domains.

This kind of temporary commercial organization, formed to maximize the commercial value of a single endeavor through large-scale socialized collaboration in the shortest possible time — let's call it a "short company." Such short companies no longer need to worry about long-term moats — because none will form. They don't need to worry about long-term incentives — because all incentives are immediate, what you create is what you see and what you see is what you get. They certainly don't need to worry about cultural continuity — because there's no intention of continuing.

Pursuing short-term profit maximization is hardly new under the sun; this hit-and-run approach has existed since antiquity. It's actually those so-called evergreen "long companies" that only emerged after the Industrial Revolution. The difference is that hit-and-run enterprises of the past could rarely achieve scale, and typically required strong interpersonal ties to sustain; to build a big company, you first had to build a long company, or at least a relatively long one. Today, thanks to the internet, blockchain, and artificial intelligence, short companies can instantly organize efficient collaboration and value-sharing among masses of strangers, and can also reach considerable scale. "Short yet big" companies and "invisibly big" companies are both becoming possible.

Venture Capital Won't Disappear, But Will Look Very Different From What Came Before

As "small yet big" and "short yet big" companies proliferate, what changes — and what changes are needed — in the venture capital industry?

Silicon Valley investor Chamath Palihapitiya recently predicted that "the entire venture capital industry could potentially be replaced by a new system that automatically matches capital with targets."

Personally, I find that conclusion somewhat extreme — I don't believe venture capital will vanish entirely. But I do believe the industry will undergo sufficiently profound changes that the venture capital we know in the future will look very different from the venture capital we knew in the past. (In China, in fact, the venture capital we know today and the venture capital we knew a decade ago are already barely the same thing when it comes to capital sources and how the game is played.)

The emergence of large numbers of small-yet-big and short-yet-big companies will lead to —

1. Early-stage investment opportunities becoming increasingly dispersed and fragmented, with fewer and fewer targets suitable for institutional venture capital. Small companies of two or three people, four or five people, may leverage socialized AI infrastructure to create substantial value, and the number of such companies will increase dramatically. Meanwhile, short companies rapidly assembled for temporary purposes will also multiply dramatically. Neither type necessarily suits or requires pure venture capital.

2. Many truly first-tier companies with investment value may never need to raise funding; even when they do, raising from users and customers will prove more valuable than raising from professional financial institutions. Consequently, users and customers who are simultaneously investors and shareholders will become increasingly common. (This is already happening in B2B — many new energy companies, for instance, are only interested in investment from prospective customers. Going forward, this will increasingly occur in B2C companies as well.)

3. Combined equity structures of "community membership + user rights + equity" will become more prevalent. For startups, a user who becomes a community member through purchasing services, then uses services and creates value within that community and gains some share of community rights, will prove more attractive than pure financial investment.

4. As AI and AI-based cloud services bring startups closer to cash flow and make that cash flow increasingly transparent, more and more investment will be based on cash flow rights rather than strictly equity. Cash flow rights will become an important investment concept.

5. The boundary between primary and secondary markets will increasingly blur; sufficiently dispersed primary market fundraising will, in a sense, be not fundamentally different from raising through IPOs in secondary markets. (In fact, for many companies with modest market caps today, their IPOs and subsequent fundraising are already barely distinguishable from primary market rounds.)

6. The barriers to becoming an institutional investor will keep falling, and capital sources will gain ever-stronger ability to direct where their capital goes, leading to ongoing disintermediation. At root, whether funds or funds-of-funds, these are all capital intermediaries. As the information and cognition gaps between them and capital sources shrink dramatically, the space for them to add value will inevitably be squeezed.

Against this backdrop, the venture capital industry will become increasingly dispersed, increasingly focused, increasingly close to industries and successful founders, increasingly adapted to project-fund operating models, increasingly inseparable from AI, and increasingly accustomed to deriving value from cash flow generated by portfolio companies' users and customers — rather than from the next buyer in secondary markets.

Some Preliminary Results From Our Internal Experiments — What Work Gets Replaced

In recent weeks we've spent some time conducting small-scale experiments. I've personally tried using GPT-4 to help complete tasks we regularly encounter in daily work, and have had multiple exchanges with relevant colleagues. While this doesn't constitute anything like definitive conclusions, my preliminary impressions from the experiments are as follows:

— On market and industry understanding: It can help us rapidly form a basic framework and provide effective information sources, and can serve a suggestive role in the comprehensiveness of thinking angles. For most domains where we already have some knowledge and understanding, the incremental gain here is in the 10-30% range; for relatively unfamiliar domains, the incremental gain is higher.

— On data analysis: It can save our colleagues doing foundational analytical work at least half their time, and substantially improve work quality on certain dimensions, with incremental gains of at least 50%, likely higher.

— On counterparty discovery: It's still difficult for it to achieve precision and comprehensiveness today, but can play a gap-filling role in certain situations, with incremental gains of 10-20%.

— On report and document reading and information summarization: It can't fully replace human reading, but can at least improve our reading efficiency and accuracy, with incremental gains of at least 30-50%; in extreme cases it can compress report or document reading time by 80-90%.

— On document translation: Used in combination with translation software, it can save at least 80% of original manpower.

— On hypothesis validation and retrospective model verification: It still needs considerable training and personalized data integration; I haven't been able to form a definitive conclusion, but my intuition is that incremental gains here will be enormous — likely not multiplicative but exponential efficiency improvements.

To summarize: at least in our industry, I believe it's entirely plausible that within two to three years, AI could replace half or even two-thirds of foundational analytical work currently performed by mid-to-junior level colleagues. Definitely not 10-20%, but definitely not 100% either.

I believe similar dynamics will play out in many other knowledge-work industries. Even in physical labor domains that seem less replaceable today — such as delivery and food delivery, healthcare and nursing, transportation and logistics, construction and infrastructure — gradual large-scale replacement of traditional human labor by robots will also occur.

This raises another question — if a substantial portion of jobs in the workplace will be replaced, what future do young people have?

Future Opportunities for Young People — The Creator Economy in the Broad Sense

Whether domestically or internationally, commercial value in the future economic landscape will increasingly flow toward two poles, with enterprise value distribution gradually shifting from a spinning-top shape to a dumbbell shape.

One end of the dumbbell is super-large companies, responsible for providing infrastructure for industries and even society as a whole, indirectly helping governments tax all commercial practitioners by collecting infrastructure usage fees.

The other end is entrepreneurial small companies — not necessarily with many employees, but capable of leveraging infrastructure provided by large enterprises to create substantial commercial value.

In the middle, there will continue to be some medium-sized companies focused on vertical domain services or playing secondary infrastructure provider roles in vertical domains, while also giving rise to short companies that rapidly assemble for short-term commercial opportunities and disperse after capturing short-term results.

While those infrastructure-providing super-companies will also have demand for young talent, it's precisely these two new types of companies I mentioned — "small companies" and "short companies" — that represent the real employment opportunities for young people going forward.

More and more foundational analytical work based on established methodologies and specific information sources will be replaced, but more and more creator economy opportunities will open to young people with imagination, creativity, and vitality. The "small companies" and "short companies" I mentioned earlier will precisely become the backbone of the creator economy.

The creators referred to here are not merely creators of artistic works, but also include generators of various new content and short-form content, creators of AI agents, designers and inventors of novel smart hardware, explorers and providers of new data services, builders of new content-driven marketing channels, and so on. AI will give creators' creative work wings; the internet will help creators' works and products spread widely; blockchain can provide institutional arrangements for creators' rights verification and long-term value-sharing.

Thus, each generation will have its generation's opportunities. But the alternation between old and new opportunities will also mean dramatically reduced demand for young people in many traditional industries, with traditional career advancement paths becoming narrower and steeper. Young people will need to manifest interests, imagination, and creativity at earlier stages, and continuously use AI for self-validation, training, and iteration around these — only then will they have a decent chance of finding paths and methods for self-realization without needing to climb the long ladder of traditional careers. This clearly poses new and unavoidable challenges for future education, and for parents with high hopes for their children.