Paradigm's New Arithmetic: When Crypto Can't Hold $12.7 Billion, AI Becomes the Solution

Let’s start with a math problem.

A venture capital firm manages $12.7 billion in assets. Its previous fund raised $850 million. The one before that was $2.5 billion.

The trend is in the opposite direction.

The size is shrinking, not because they can’t raise money, but because there aren’t enough worthwhile targets to invest in. Now, this company wants to reverse this trend. Where should they look for the next large pool of capital?

On February 28, 2026, The Wall Street Journal provided an answer: Paradigm, a cryptocurrency investment firm, is raising a new fund with a maximum size of $1.5 billion, expanding its investment scope to include artificial intelligence, robotics, and other cutting-edge technologies.

This wasn’t a sudden decision. It was a calculation that had been underway for some time, only now revealing the answer.

First, let’s lay out the numbers

In 2025, global crypto VC investments totaled $49.8 billion. That sounds like good news. But looking at this number alone can be misleading.

In the same year, the number of crypto VC deals plummeted by about 60% year-over-year, from roughly 2,900 deals to 1,200. More money was flowing in, but fewer projects were being funded. Capital entering the crypto space was increasingly concentrated in a few large deals, rather than spread across hundreds of early-stage projects.

For most small and medium funds, this might not be a problem. But for Paradigm, it’s a structural challenge. Managing $12.7 billion, one of the largest crypto-focused VCs globally, their issue isn’t finding projects—it’s finding enough large, early-stage projects to deploy this scale of capital while maintaining their expected returns.

In 2021, Paradigm raised the largest crypto fund in history at $2.5 billion. In 2024, they announced their third fund, at $850 million—only a third of the previous one.

This contraction isn’t a sign of weakness; it’s a strategic adjustment to a narrower market. But it also shows that relying solely on crypto, Paradigm is finding it hard to grow at its current scale.

After FTX, Paradigm started asking a question

To understand the $1.5 billion fund today, we need to go back to November 2022.

That month, FTX collapsed. Sam Bankman-Fried’s empire was reduced to ashes in days, taking countless institutions’ funds with it. Paradigm’s paper investment in FTX was $278 million, which ultimately became worthless.

For a top-tier firm known for research-driven, tech-savvy investing, this was more than a bad debt. It was a public misjudgment, requiring explanations to LPs, the market, and themselves.

What happened afterward seemed strange at the time. In 2023, some noticed that Paradigm’s official website quietly changed: all references to “crypto” and “Web3” were removed and replaced with more neutral terms like “technology investing.”

There was no official announcement, but the community quickly spotted the change and debated fiercely. The biggest concern: Is Paradigm leaving crypto?

Co-founder Matt Huang had to step in to clarify. He tweeted that Paradigm “has never been more excited about crypto,” adding, “The development of AI is too compelling to ignore. Framing AI and crypto as zero-sum competition is a popular but mistaken narrative. We disagree. Both are fascinating and will have significant overlap.”

This was a PR clarification, but it also revealed a truth: internally, Paradigm was seriously considering AI.

Post-FTX, the pressing question was: what to bet on for the next decade?

Matt Huang is already working on the answer

If you only look at Paradigm’s official announcements, the company’s transformation seems to have started today. But if you examine Huang’s actions over the past two years, you’ll see he’s long been more than just a crypto investor.

In 2024, Paradigm invested $50 million in Nous Research, an AI infrastructure company focused on open-source large language models. This isn’t a small exploratory bet; $50 million is a serious commitment for Paradigm.

In February this year, Paradigm co-launched EVMbench with OpenAI, a benchmark tool to evaluate AI models’ ability to detect and fix smart contract vulnerabilities. The core infrastructure of crypto meets AI capability assessment—two seemingly different worlds brought together.

Meanwhile, Huang is building another company: Tempo. A stablecoin payment infrastructure firm, where Huang is a co-founder and board member of Stripe, aligns closely with this direction. Stripe partnered strategically with Paradigm in 2025, and that same year launched a stablecoin payment product.

Looking at all this together, Huang isn’t just “going into AI”; he’s been living at the intersection of AI and crypto for at least two years.

He isn’t betting solely on AI or crypto but on the moment when these two will collide. When AI agents need to execute on-chain transactions, or robots require programmable monetary systems, that intersection will be Paradigm’s next battleground.

Why AI×Crypto, not just a pivot to AI

Paradigm’s move into AI doesn’t mean it’s competing with a16z or Sequoia for the same projects.

A common misconception is to see Paradigm’s new fund as “another VC shifting to AI.” But if that’s all it is, it offers no advantage—general AI is already crowded with well-established, resource-rich traditional VC giants.

Paradigm’s real strategy is: it doesn’t aim to compete in the broad AI market; it wants to invest in the niche intersection that others haven’t yet fully recognized.

AI agents are among the hottest concepts today. These autonomous, task-performing intelligences are already replacing humans in various scenarios: search, coding, data analysis, process management. But there’s one thing they haven’t solved: money.

When an AI agent needs to pay, receive, or transfer funds across services, what does it use? PayPal? Bank accounts? These systems are designed for humans, requiring identity verification and manual authorization—things incompatible with autonomous machine operation.

But stablecoins can. Smart contracts can. Programmable money can.

That’s why Huang is simultaneously working on Tempo (stablecoin payments) and investing in Nous Research (AI infrastructure): he believes these two will eventually merge. Paradigm has the capacity to bet on both sides and profit when they do.

This isn’t a pivot; it’s an expansion—into a space he believes others haven’t fully grasped yet.

LPs need a new story

There’s also a practical aspect to clarify.

Paradigm’s LPs—institutions and individuals who entrust it with their capital—saw a $2.5 billion fundraising ambition in 2021, then a more restrained $850 million in 2024.

The stark difference between these two fund sizes requires an explanation. More importantly, it needs a compelling narrative for the next fund.

“Continuing to invest in early crypto projects” was a viable story in 2024 for raising $1.5 billion. But now, “leveraging crypto’s technological edge to enter frontier tech during the AI and robotics boom” is more convincing.

In 2025, 61% of global VC funding—about $258.7 billion—flowed into AI. This is the largest pool of risk capital today. Paradigm’s $1.5 billion fund aims to tap into this pool, not just hold onto a shrinking one. For LPs, this is a bigger, more credible growth story.

Now, returning to 2023. When Huang clarified the website change, he said: “AI and crypto are not zero-sum competition.”

At the time, this was more of a defensive statement—calming the community, preventing LP panic, and leaving room for AI exploration. But if you read it in today’s context, it reads more like a prelude.

Paradigm has spent three years recovering from the FTX wreckage. It didn’t choose the easiest route of shrinking and waiting for the next bull market. Instead, it took a more challenging path—betting on the convergence of AI and crypto, building positions in both, and waiting for their intersection.

Today, this $1.5 billion fund marks a milestone in that journey.

Huang hasn’t publicly responded to The Wall Street Journal’s report. But his Tempo is still in development, Nous Research is ongoing, and EVMbench has been released.

He doesn’t need to explain anymore. Those actions speak louder than any statement.