Is Anthropic about to personally kill Figma?

Author | Hualín Wǔwáng

Editor | Jìng Yǔ

In 2010, a small company called Instagram was acquired by Facebook for 1 billion USD. At the time, many people didn’t understand why so much money was spent buying a filter app. The story that followed is known to everyone—Facebook wasn’t buying a product; it was buying a species that could threaten itself, and turning it into its own weapon as well.

Mike Krieger—the co-founder who helped Instagram grow from zero to hundreds of millions of users. In May 2024, Mike joined the thriving Anthropic, serving as Chief Product Officer.

On April 14, 2026, this Krieger resigned from the Figma board. Three days later, Anthropic announced the release of Claude Design.

This time gap doesn’t look like a coincidence.

01 In three days, the landscape of an industry changed

That day, Figma’s share price fell by more than 7% at most, dropping from $20.32 to $18.84. The market’s reaction is always more honest than press releases.

Claude Design is an experimental product driven by Anthropic’s latest flagship model Opus 4.7, coming from Anthropic Labs’ internal team. What can it do? Prototypes, slide decks, one-page summaries, and all kinds of visual content—these are exactly the things designers and product managers do every day when they open Figma or Canva to get their work done.

But if you only treat Claude Design as “another AI design tool,” you’ll underestimate the significance of what’s happening.

What truly makes industry insiders on alert is the “handover” mechanism between Claude Design and Claude Code.

Claude Design page | Image source: Anthropic

When you use Claude Design to complete an interface prototype, the system automatically packages the full set of design specifications into a “handover package,” which can be handed over directly to Claude Code for continued development.

More importantly, when enabled, Claude Design reads your codebase and existing design files, and automatically builds a design system tailored to your team—fonts, colors, layout standards, and brand governance rules—read once, then used throughout.

On Reddit, developers commented that this solves the “most annoying part” of using AI as a design tool—every new project requires you to explain your brand standards to the AI all over again.

Design used to mean two separate tools, two separate processes, and two sets of people. Now Anthropic wants to turn this chain into a closed loop.

02 A clear strategic rhythm

If you put Claude Design into the timeline of the past few weeks, you’ll see an Anthropic moving at a faster and faster pace.

In early April, Anthropic announced that the Claude Mythos Preview would be limited in release. This model is able to discover and exploit vulnerabilities that have been hidden in key software systems for decades. The security risks are so high that the company decided not to open it to the public—instead, in the form of “Project Glasswing,” it provides targeted access to more than 50 top institutions such as Amazon, Apple, Microsoft, Google, and JPMorgan Chase, and each institution comes with a 100 million USD usage quota, specifically for defensive cybersecurity research.

On April 14, Opus 4.7 was officially released, bringing stronger coding capabilities, clearer visual understanding, and a new “self-check” ability. Anthropic itself also admits that Opus 4.7’s performance isn’t as good as Mythos—but Mythos hasn’t been released publicly yet due to security considerations.

On April 17, CEO Dario Amodei met with White House officials, and the topic was the cybersecurity concerns raised by the Mythos model. On the same day, Claude Design was released.

Also on April 17, foreign media reported that Anthropic’s valuation has reached 800 billion USD. The company is in early discussions with Goldman Sachs, JPMorgan Chase, and Morgan Stanley regarding an IPO, with the earliest possible timing being as soon as October this year to go public.

This isn’t what a “model-selling company” would do. This is a company preparing for a listing, one that needs to make the capital markets understand clearly what makes it worth that price, while systematically filling out its product matrix.

From disrupting the developer tools market with Claude Code to Claude Design entering design workflows, Anthropic’s logic is very clear: identify the tool scenarios that professionals use at high frequency every day, redo them in an AI-native way, and then use “model capabilities” as a moat that makes it hard for competitors to catch up.

03 Challenging Figma—but also facing reality

However, there’s often a gap between an ideal closed loop and actual usage.

After trying Claude Design in practice, some reviewers found that simply building a design system, setting up a prototype website, and making a few adjustments already consumes more than 50% of the weekly quota. After you exceed the quota, you need to pay additional token fees. For design scenarios that require frequent iteration, this cost pressure cannot be ignored.

There are also obvious bugs—for example, the design system preview sandbox can’t correctly read image files, causing image links to fail.

At its current stage, Claude Design is positioned more like an internal demo and a fast prototyping accelerator, rather than a production-grade design tool that can be delivered directly.

Claude Design page | Image source: Anthropic

This also means that Canva and Figma aren’t without ways to fight back. Figma has built a moat over years in collaboration features, fine-grained management of design systems, and professional designers’ workflows; Canva’s advantage lies in its template ecosystem and ease of use for non-professional users. In the short term, production-level content still needs manual fine-tuning.

But the word “short term” is becoming shorter and shorter in the AI field.

Anthropic’s real ambition this time may not be to directly replace Figma—maybe it’s to redefine “who Figma’s target users are.” When an independent developer, a product manager at a small team, or a startup founder who needs to quickly make a demo can build a “good enough” prototype in minutes using Claude Design, and then seamlessly hand it over to Claude Code to implement it, do they still need to spend time learning Figma?

That’s the real reason Figma’s stock price fell by 7%.

04 From selling shovels to going to mine

There’s a metaphor that’s been circulating in Silicon Valley for a long time: during the gold rush, the most profitable people weren’t the ones who went to pan for gold, but the ones who sold shovels. In the early days of the AI wave, OpenAI and Anthropic played the role of “shovel sellers”—providing APIs so developers and enterprises could build applications.

But now, Anthropic has started digging its own mine.

Claude Code and Claude Design are two shovels, and they are also two entry points that consume users’ time. When Anthropic directly builds developer tools and design tools, its relationship with the ecosystem companies that used to build products based on its APIs changes from “partner” to “competitor.”

This path has been walked by Microsoft, Google, and Apple—though the difference is that those companies had platforms first and then applications, while Anthropic builds trust first through model capabilities, and then spreads that trust upward into the application layer.

Back then, Mike Krieger turned Instagram into a platform, and later watched Facebook use that platform to suppress competitors. Two years ago, he joined Anthropic and began making products.

History won’t simply repeat itself, but sometimes, participants do.

Anthropic’s IPO may happen as early as October this year. Before that, it will likely “release” a few more times, so the capital markets can see clearly what kind of company this is trying to become.