A staggering $10 billion windfall in three years: Unpacking the AI investment logic of the top VC behind Anthropic

There is a very interesting detail in Anthropic's prospectus.

In this hundreds-page-long document, an investment institution named Spark Capital is mentioned more than 50 times in total.

This frequency of mention even exceeds that of many tech giants that have invested billions of dollars in Anthropic.

The reason is simple: Spark Capital is one of the earliest institutional investors of Anthropic, and its partner is still the only representative from a traditional venture capital firm on the company's board of directors.

Many people may not know that when Anthropic was first established, almost no VC was willing to invest in it. Even before its Series C financing, the company largely had to rely on personal connections in the tech circle to seek funding.

Especially after ChatGPT made a sensational debut, investors had more reasons to hesitate: OpenAI had taken the lead, and it also had Microsoft's full backing. What could make Anthropic win?

But Spark chose to place a firm bet instead.

Starting from the Series C round, Spark participated in every financing round of Anthropic, investing a total of 777 million US dollars. Calculated at a valuation of 1.5 trillion US dollars, it has earned a full 10 billion US dollars in just three years.

Why did Spark of all firms dare to bet on its competitor when OpenAI was at the height of its power?

After reviewing the company's investment records over the past decade and public interviews with several core partners, I believe this investment is far more than just a stroke of good luck.

Spark realized very early that AI is changing not only the competitive landscape of the technology industry, but also the rules of the game for venture capital itself:

Those relatively static competitive advantages in traditional investment logic are becoming increasingly unreliable in the AI era. Compared with temporary technological leadership, technological judgment and the ability to respond to changes are far more critical.

A $75 Million Investment That Changed Fortunes

Speaking of Spark Capital, one has to mention its partner Yasmin Razavi.

This woman, who grew up in Tehran, Iran, decided to stay in Canada temporarily in 2006 when she went there to visit her elder brother who was studying abroad. Later, she was admitted to the University of Toronto, joined McKinsey, and then studied at Harvard Business School. She joined Spark in 2017 and was promoted to partner in 2020.

And the most successful investment in her career is none other than Anthropic.

Despite Anthropic's current prominent status, almost no VC was willing to invest in it when it was first founded.

In 2021, Anthropic had no products and no revenue, yet it asked for $500 million in financing right off the bat.

At that time, Anjneyn Midha, a scientist by background, once helped Anthropic find investors. He met 22 VCs in one go, 21 of whom rejected the offer on the spot, and the remaining one did not reach a deal in the end.

Some investors even asked in reply: What even is GPT-3?

With no other options, Anthropic had to turn to established tech tycoons who had already achieved financial freedom.

In May 2021, Anthropic completed a $124 million Series A financing, led by Jaan Tallinn, co-founder of Skype, with other investors including Dustin Moskovitz, co-founder of Facebook, among others.

In April 2022, it completed a $580 million Series B financing, and one of its biggest backers was Sam Bankman-Fried, who later became infamous due to the FTX bankruptcy.

The real turning point came in 2023.

ChatGPT made a sensational debut, and Microsoft poured huge sums of capital into OpenAI. The commercial prospects of large models were finally verified, but new problems also emerged:

Since OpenAI had already taken the lead and had Microsoft's full backing, why would anyone still invest in Anthropic?

Most VCs still chose to wait and see. But Yasmin Razavi thought differently.

In an interview with Forbes in May this year, she recalled that in order to push forward this investment, she specifically wrote two detailed investment memos to convince other partners at Spark.

Her logic mainly consisted of three points:

First, large language models may represent a genuine technological paradigm revolution.

Drawing on the theory of paradigm shift put forward by the philosopher of science Thomas Kuhn, she believed that the development of deep learning, unsupervised learning and Transformers over the past decade has created conditions for large language models to bring about a new technological paradigm.

Second, the application scope of AI will far exceed that of ordinary software tools.

Razavi listed fields including law, healthcare, accounting, software development, mining and agriculture, arguing that AI is redefining the way humans and computers work together.

Third, safety is not only a matter of technological ethics, but also the foundation for developing reliable AI products.

She specifically mentioned the Anthropic team's research accumulation in Scaling Laws, RLHF and Constitutional AI, stating that this team not only has top-tier model capabilities, but also has long-term dedication to safety and alignment issues.

More importantly, she did not believe that the large model industry had been completely taken over by OpenAI. On the contrary, as AI becomes a new technological paradigm, companies with original research capabilities that can continuously iterate their models still have the opportunity to become the next generation of tech giants.

In May 2023, Spark led Anthropic's $450 million Series C financing, contributing $75 million out of its own pocket. This was the largest check in Spark's history, accounting for more than 5% of its $1.4 billion growth fund size.

Facts have proved that this investment was only the beginning.

According to the prospectus, Spark subsequently participated in five consecutive financing rounds of Anthropic, with a total investment of 777 million US dollars. Measured at a valuation of 1.5 trillion US dollars, the book value of Spark's shares may exceed 10 billion US dollars, with a total investment return of about 14 times.

Among them, the shares corresponding to the initial $75 million investment may potentially generate returns of up to 80 times.

But what Spark gained is not just financial returns. Razavi has served as a director of Anthropic since March 2023, even earlier than the official announcement of the Series C financing.

After joining the board of directors, Razavi began to help the company identify which specific applications can create the greatest commercial value.

Daniela Amodei, co-founder of Anthropic, revealed to Forbes that before Claude Code matured, Razavi repeatedly advised the company to position AI programming as an independent and important market.

As we all know what happened later, Claude Code became one of the fastest-growing products of Anthropic.

Spark Only Favors Two Types of AI Application Companies

If you only look at Anthropic, it is easy to interpret Spark's success as a random investment. But if you extend the timeline, you will find that the company's layout in AI is actually quite systematic.

Judging from its public investment portfolio, Spark has bet on almost every key link in the AI industrial chain.

In the field of foundational models, in addition to Anthropic, it has also invested in Profluent, an AI protein design company, and Generalist, a foundational robotics model company.

At the computing power layer, Spark has invested in Lightmatter, MatX and Enfabrica, focusing on optical interconnection, AI-specific chips and high-performance network interconnection respectively.

At the infrastructure layer, it has invested in data labeling company Scale AI, vector database Qdrant, model inference platform Baseten, and AI data center company Crusoe.

The application layer is also fully covered, including AI medical company Abridge, audio and video editing platform Descript, AI video generation company Pika, and AI meeting note tool Granola.

These investments seem scattered, but the logic behind them is highly consistent: to find companies that are truly capable of building long-term competitive barriers in each link.

Especially for the AI application layer, Spark divides AI startups into three categories:

The first category is Adaptation, which means adding AI features to existing software, such as AI search and AI assistants.

Although such products can improve user experience, it is difficult to form long-term barriers solely through AI features.

The second category is Evolution, that is, using AI to redesign workflows.

Typical representatives are AI medical company Abridge and AI meeting note tool Granola.

The former uses AI to automatically organize doctor-patient conversations and generate clinical documents, while the latter combines users' own records with meeting content to generate notes that better meet individual needs.

The third category is Revolution, which means AI creates products and markets that did not exist at all in the past.

Hyatt admitted that Spark's most successful investments are mainly concentrated in the third category, as well as some in the second category. It has basically no interest in the first category of companies that simply add AI features to existing products.

Because in their view, truly valuable technological innovation is not just to make original things cheaper and faster, but to enable users to accomplish things that were impossible before.

Hyatt also specifically warned that there is a trap in AI Agent entrepreneurship.

Many companies use large models to automate traditional manual services and quickly generate revenue by reducing costs. But if competitors can all call the same model, the cost advantage established today may disappear tomorrow.

The real competitive barrier must come from the product itself, as well as the continuously accumulated data, user feedback and workflow capabilities.

This is perhaps the key to understanding Spark's AI investment map: it invests not only in companies that create new technological paradigms, but also in companies that use new technologies to redefine products.

In the AI Era, the Rules of VC Investment Are Also Changing

Spark is not only re-evaluating what kind of AI companies are worth investing in, but also rethinking the traditional VC investment methodology.

In 2025, Spark partner Nabeel Hyatt put forward a very interesting point of view in an interview.

In the 2025 interview, Hyatt stated that SaaS investment over the past decade has increasingly become a math problem with a standard solution.

Investors can assess the growth potential of a software company through indicators such as ARR, customer retention rate, sales efficiency and market size.

But investing in AI is different. No one has answers to questions like what capabilities the models will have in a few months, what product forms will emerge, and which positions will be redefined at the moment of decision-making.

So he made a distinction: in the past, VCs were good at solving Puzzles, which are difficult problems with known conditions; AI investment is more like Mystery, an exploration where the answer has not been written yet.

This distinction seems very abstract, but it is very intuitive when applied to Anthropic.

In 2023, any investor who tried to use traditional financial models to estimate the value of Anthropic would find it difficult to find a convincing reason to invest in a company with limited revenue, high computing power costs and OpenAI already leading the way.

But what Spark valued was precisely those things that cannot be measured by financial models: a top-tier research team, and the entirely new market that large models may create.

Of course, betting on technological change does not mean that investment risks can be ignored. Especially in the AI industry, technologies that seem leading today may be caught up or even surpassed by competitors in just a few months.

So the question arises: if technological leadership can be broken at any time, what exactly is the long-term competitive barrier for AI companies?

Hyatt was also asked this question in a previous interview.

At that time, the emergence of DeepSeek made many investors start to question how much competitive advantage large model companies can still build after investing tens of billions of dollars in model training.

But Hyatt does not believe that huge capital investment alone is the only barrier for model companies. Compared with the leadership of a certain generation of models, Spark values the technical judgment of top teams, as well as their continuous execution and iteration capabilities more.

From this perspective, those relatively static competitive advantages in traditional investment logic are becoming increasingly unreliable in the AI era. Compared with temporary technological leadership, whether an enterprise can continuously adapt to changes and even take the initiative to lead the next round of technological transformation is the key to determining its long-term value.

This article is from the WeChat official account "Silicon-based Observer Pro", written by Silicon-based Jun, and published with authorization from 36Kr.