Stocks and startups affected after the release of Claude Design

On Sunday morning, at Zombie Café’s table number six on California Ave, the third shot of espresso was finally refilled. The new tool, Claude Design, launched last Friday by Anthropic, was examined twice by Alan Walker in Silicon Valley.

First, let me break down this tool — what it is, why it’s not just “another AI design tool,” and why it’s being priced as a structural event by the US and Hong Kong stock markets. Once I clarify this tool, I’ll go through the companies it has already impacted and those about to be affected, one by one.

What is Claude Design · A Sketch of a New Tool

To put it simply — when you open Claude and type in a prompt like “I want a SaaS pricing page, dark theme, three tiers,” it directly gives you an interactive prototype. If you say “Mark the second tier as recommended,” it updates. If you ask “Change the color scheme to our brand colors,” it reads your codebase and Figma files, automatically aligns. After finishing, you can package everything with Claude Code, which automatically generates deployable code.

Sounds like a combination of Figma + Gamma + v0? Exactly. But this combination is just superficial; the real power lies in three core details below.

Why it’s a Game Changer · Three Underlying Details

First Tool · Directly Access Codebase and Figma Files

During onboarding, Claude Design directly reads your company’s GitHub repositories and Figma design files, automatically generating a live design system — colors, typography, components, tokens — all aligned. All your future projects automatically apply this system.

This cuts out the entire rationale behind the design system tooling category of the past decade.

Companies like Supernova, zeroheight, Specify, Knapsack, Tokens Studio, Frontify are fundamentally about “helping you maintain consistency between Figma and code.” When Claude reads directly from the source, the need for intermediate sync layers disappears. This entire category resets to zero.

Second Tool · Handoff Bundle to Claude Code

Design is done, and with one click, you package a handoff bundle, give it to Claude Code, and code is generated automatically. There’s a phrase in the release notes that’s seriously underestimated —

"without code review or PRs"

This sentence is a death sentence for the past decade of the design-to-code industry chain.

• Zeplin as an intermediate layer for design-to-dev

• Anima, Locofy converting Figma to React

• v0 transforming prompts into React components

• Bolt, Lovable turning prompts into full-stack apps

These four generations of players have collectively raised over 8B USD over the past ten years, all pursuing the same goal — “Design and code are two systems, requiring a translation layer.” Today, Anthropic puts Claude Design and Claude Code into the same runtime, making design and code no longer just “handshakes,” but “two surfaces of the same object.” The physical foundation of the translation layer has vanished.

$15 Third Tool · Frontier Design · Non-Designers Can Also Create Shaders, 3D, Voice, Video

The most aggressive line in the release notes —

![]###https://img-cdn.gateio.im/social/moments-e15d7a589a-63b0b1b963-8b7abd-badf29(

The word “Anyone” here isn’t marketing — it’s a declaration of war.

In the past, prototyping with shaders, 3D, and voice interactions required a three.js engineer, an effects designer, and a sound designer, starting from two weeks, costing ). Today, a product manager can produce a prototype in half an hour using prompts.

Companies like Spline, Rive, Bezi, Ceros that rely on “abstracting complex tech” for their business, have a moat based on technical barriers. When LLMs flatten those barriers, the moat disappears.

Putting these three tools together, you understand why the market treats Claude Design as a structural event — it’s not just “a better Figma,” but “integrating the upstream, midstream, and downstream of creative workflows into an LLM platform in one go.” Every downstream tool that helps Figma reach somewhere else now needs to prove its value again.

Timing of the Release · A Pre-Meditated Setup

Let me tell you something many people overlooked. Claude Design was officially released on Friday, April 17, but the real moves happened on the previous Tuesday, April 14.

That day, Figma filed an 8-K with the SEC — a very brief note: Anthropic’s CPO Mike Krieger resigned from Figma’s board. This isn’t just anyone; he’s one of Instagram’s co-founders, and last year, Anthropic recruited him to lead Anthropic Labs. He had been on Figma’s board for less than a year.

On the same day, The Information leaked that Anthropic’s next-gen Opus 4.7 would integrate a full suite of design tools. This wasn’t a leak — it was Anthropic signaling itself. Three days later, Claude Design was launched.

![]$50K https://img-cdn.gateio.im/social/moments-e1dabf4324-6cbf8b4179-8b7abd-badf29(

This rhythm is clear after two viewings: Krieger couldn’t stay on Figma’s board even a day longer because the shot was already scheduled. This isn’t “competition getting fiercer,” it’s premeditation. After twenty years in Silicon Valley, I’ve learned that the real big battles don’t start at the launch event — they start with the board seat disclosures a week before.

Figma · From $60B to $10B in One Year

When Figma went public last year, Dylan Field declared on stage, “Figma is where teams come together.” At that time, the company was valued at $60 billion, and the IPO was a success, with a strong first day. Less than a year later, on Friday, its market cap fell to $10 billion, an 80% drop from the peak.

This isn’t just fluctuation — it’s a paradigm shift in valuation.

On Friday, Figma dropped 7.28%, which was relatively mild — the real blow came last Tuesday when Krieger resigned, and that day, the stock already fell 6%. Combined, Figma’s market cap evaporated by over a billion dollars just this week.

Some ask, doesn’t Figma have a moat? Multi-user collaboration, plugin ecosystems, designer identity — these are things Claude can’t replicate overnight. That’s true. So Figma won’t crash 20% in a day like LegalZoom; instead, it will de-rate gradually — over two years, its forward PE might fall from 60x to 20x, and its market cap could halve again. That’s the classic SaaS story — when a company’s growth outlook shifts from “high growth” to “steady cash flow,” valuation multiples are compressed to a third.

A friend at Benchmark called me last Wednesday, saying they’ve cut Figma’s target multiple from 12x sales to 4x sales. It’s not just his view — it’s the consensus in the growth equity circle. The crossover funds that invested at IPO will have a tough year.

Figma’s only way to survive is to rely not on “growth,” but on “institutional moats.” Millions of designers worldwide store their work, teams, and versions on Figma — switching costs are high. Just like Slack survived Microsoft Teams’ encroachment, Figma will become a mature, dividend-paying software company. That’s the valuation you see.

Adobe · An 18-Year Veteran Hands Over at the Right Time

Adobe’s story is even more complex. Shantanu Narayen, whom I met at Adobe Summit in 2015, has been CEO for 18 years, leading Adobe from $30B to $200B. On March 12, he announced he will step down, and the stock dropped 7.6% that day, down 25% year-to-date, approaching a three-year low.

I see Adobe differently from many. Many believe Narayen was pushed out by AI — that’s not true. He’s seen what Adobe must do next — and it’s something he can’t accomplish.

Adobe’s next move is “AI-native reconstruction” — essentially tearing down Creative Cloud and rebuilding it as an agent-first product line. Narayen can’t do this because he’s the hero of the subscription transition; his entire mental model is based on subscriptions. To burn that old model, a completely AI-native CEO without baggage is needed.

The market is waiting for that person. The internal favorite is David Wadhwani, and externally, several SaaS and AI executives are being vetted. My judgment is —

![])https://img-cdn.gateio.im/social/moments-e815ae2deb-b5b0b247dc-8b7abd-badf29(

But regardless of who it is, Adobe’s long-term story is no longer “subscription invincible.”

After Friday’s Claude Design release, Adobe only fell 1.5%, which looks decent — but that’s just because the news of Narayen’s resignation has already been digested. The real pain is yet to come, and we’ll have to wait for Q2 guidance.

The Hidden Risks in Canva, Wix, and a Long List of Less Obvious Failures

Canva CEO Melanie Perkins gave a brief endorsement in the Claude Design release — “bringing Canva to wherever ideas begin.” Silicon Valley folks understand what this means — it’s a CEO openly admitting that the top of the funnel is no longer in her company’s hands. Last year, Canva’s internal valuation was $60 billion, but now it’s dropped to $42 billion. If they go public next year, the expected price range is around $30 billion, unless Melanie can come up with a sufficiently aggressive agentic strategy.

Wix has been on my radar. It fell 4.7% on Tuesday, and its performance this year isn’t great, but the stock is now only $20, with a forward PE of 13 and free cash flow of $570 million — cheap. Motley Fool has already started calling “buy the dip.” My view is different — Wix is a classic “deep value” and “value trap” candidate. Its DIY website builder business is directly targeted by Claude Design + Claude Code. SMB customers tend to churn high, but if they find they can build a better-looking website with Claude chat than Wix templates, migration becomes easy.

GoDaddy also dropped 3% that day. The logic is similar to Wix — older, more sticky users. Short-term, it’s hard to fall further, but long-term, it will de-rate.

And a whole list of companies you might not expect:

![])https://img-cdn.gateio.im/social/moments-b4ddce4672-f1108edd12-8b7abd-badf29$42B

This isn’t just isolated — it’s an entire industry chain.

The most dramatic is Thomson Reuters — a 15.83% drop in one day, the largest single-day decline in its history, caused by Claude Cowork’s legal plugin. LegalZoom fell 19.68% that day. The traditional SaaS sectors of legal, accounting, and data services have been penetrated over the past three months. Gartner also dropped 21% that day; research firms that “organize information manually” are no longer worth as much in the face of LLMs.

Three Waves of SaaSpocalypse · Timeline

Pulling the timeline together, you see what Anthropic has been doing this year.

Four months, three waves of bullets. Each hits different categories, but the underlying logic is the same — LLM-native products replacing SaaS middle layers.

Anthropic’s valuation has already risen from ( to reportedly rejecting an $800 billion investment offer. If the founders believe the valuation will break a trillion next year, it means they still have a lot of ammunition unreleased. Several hedge fund friends around me are now watching closely — where will the next shot land?

Goldman Sachs Prime Brokerage has a chart I keep looking at. Hedge fund positions in semiconductors and software are at record highs — net long in semiconductors, net short in software. This isn’t a short-term trend; it’s industry-wide consensus. You don’t need smart money — just see where the money is, and the story writes itself.

The Companies Hit Hard by Claude

Let’s focus on this part separately. Over the past two years, the hottest category in Silicon Valley has been “AI design tools” or “vibe coding.” Several players have raised huge sums —

Gamma — making AI pitch decks, valued at $600M last year, with strong user growth. Its pitch deck features 100% overlap with Claude Design’s functions. Today, Gamma’s board is probably in a call discussing “How not to become just a feature.” The answer is tough because Gamma uses the Claude API — the underlying models are from others, and now those others are directly involved, with branding and distribution in their hands.

v0 by Vercel — always the most elegant design-to-code product. Today, v0’s core proposition is being directly eaten by Claude Design + Claude Code. Vercel’s parent company has Next.js and deployment infrastructure, so it’s not going away, but v0 as an independent product will need to find a new positioning. The most likely path is — v0 downgrades to a deployment target for Claude Design, from a peer competitor to a downstream executor. Not shameful, but definitely less sexy.

Lovable — I’ve liked this company for a while. It just completed its Series B in December, with a valuation of $6.6 billion, backed by top-tier investors — CapitalG, NVIDIA Ventures, Salesforce Ventures, Databricks Ventures, Atlassian Ventures, HubSpot Ventures. Its ARR jumped from $192837465657483.91T in July to $400 million in February — very aggressive. But Lovable’s core tech route is “using Claude to make apps” — they use Claude themselves. Now, with Claude doing app generation directly, the legitimacy of that proposition is shaken. Lovable might survive a while longer with full-stack deployment and integration with Supabase, but the valuation re-rate pressure is mounting.

Bolt.new (backed by StackBlitz) — in the same category, with 90% overlap. Its WebContainer tech is differentiated, but whether it’s enough against platform giants is uncertain.

Galileo AI, Uizard, Visily — smaller players, highly homogeneous propositions, no distribution, no moat. Their investors are already preparing for next-round valuation cuts or early exits this weekend.

A friend in early-stage investing sent me a screenshot last week — three CEOs of AI design tools in his portfolio sent him urgent emails, all using the same word — “pivot.” No clear direction, but the word “pivot” has appeared. This is the real Silicon Valley scene.

The Hong Kong Market · The February 4th Crash

That’s the story in US stocks. Now I want to focus more on Hong Kong and A-shares — because this segment is less talked about by mainstream financial media, but contains more mispricings and opportunities than US stocks.

Let’s rewind. On February 4, US SaaS was bloodied overnight by Claude Cowork’s legal plugin. The next morning, the sentiment spread to Asia, and the Hong Kong SaaS sector nearly collapsed:

![])https://img-cdn.gateio.im/social/moments-6b6bea2ebe-c565281ea7-8b7abd-badf29$380B

Hong Kong SaaS’s total market cap evaporated nearly HKD 15 billion that day. The sentiment kept spreading: Tencent -3.96%, Alibaba -0.93%, NetEase -3.32%, Baidu -2.97%, Hang Seng Tech Index -1.84%. A-shares fared worse — Sangfor -4.86%, Yonyou -3.27%, HanDe Information -7.41%, and others like GigaMedia and Tiandi Online hit the daily limit down. My first reaction was — overreaction. Not that Hong Kong SaaS has no issues, but their underlying logic differs from US SaaS. US SaaS is a highly mature, existing market; Salesforce, Workday, Atlassian’s clients have already digitized, and AI is eating into their subscription revenue. Hong Kong and Chinese SaaS serve early-stage digitalization clients — Kingdee, Yonyou, many still migrating from Excel to ERP. AI is an accelerator, not a substitute. That’s why I added some Hong Kong SaaS positions on February 5 — not all, selectively. The mispricing caused by emotional sell-off is an alpha opportunity.

Hong Kong Companies I Would Buy and Avoid

$330M I Would Buy

Kingsoft Cloud (03896.HK) — a structural winner. Its stock has surged nearly 6.5x in the past six months, now over HKD 32 billion in market cap. The logic is straightforward — in the AI era, demand for computing power explodes. China’s local infrastructure is in a window of substitution, and Kingsoft Cloud is the backbone for Xiaomi’s AI strategy and WPS AI. Lei Jun publicly said at Xiaomi’s Feb 27 event that R&D budget will exceed 300 billion RMB by 2025, with AI accounting for a quarter, and Kingsoft Cloud is the most direct beneficiary. This isn’t sentiment play — it’s revenue visibility.

Zhipu Huazhang (02513.HK) — the “first global large model stock,” listed on Jan 8, with a 13.17% opening gain, maintaining a valuation around HKD 90 billion. After the Spring Festival, it shot up 42.72% in one day on Feb 21, briefly surpassing HKD 63B in market cap.

![]$100M https://img-cdn.gateio.im/social/moments-ee9507eff2-d258077151-8b7abd-badf29(

This isn’t driven by earnings but by the “domestic substitution” narrative. With Anthropic’s Claude unavailable in China, domestic large models are a must — Zhipu is one of the top choices. This is policy beta, not earnings alpha, but this beta will persist over the next two years.

MiniMax — listed on Jan 9, with an initial 78% rise, closing at 109.09% higher, surpassing HKD 8B in market cap, recently at HKD 304.2 billion. Launched in February, its M2.5 model is positioned against Claude Opus 4.6. Its C-end products — Hailuo AI, Talkie, Xingye — have strong paid user bases overseas, accounting for 71% of C-end revenue. I’ll watch but not heavily invest — C-end AI companionship products face copyright risks, and overseas regulatory windows could tighten at any time.

Meitu (01357.HK) — the most conflicted. On one hand, Claude Design’s image generation threatens Meitu’s AI design tools upstream. On the other, Meitu’s 12.61 million paid subscribers (up 38.4% YoY), and its localized beauty, social, and e-commerce scenarios form a moat Anthropic can’t breach in the short term. Over the past two years, Meitu’s market cap rose from ) to HKD, up 1300% — already pricing in many optimistic expectations. Neutral bias, no new positions, no clearing — wait for Q2 earnings.

Weimob (02013.HK) — small but beautiful. Just announced FY2025 results, with a first-ever annual profit of 42.4 million RMB, and AI-related annual revenue of 116 million RMB. At HKD 1.8, it’s at the SaaS winter bottom. If AI revenue continues to grow in the next few quarters, it’s a 2-3x potential small cap. But risks remain — Weimob’s core SaaS subscription business faces AI disruption; whether AI revenue can outpace subscription decline is uncertain.

Companies I Would Avoid

Kingdee International (00268.HK) — Though it fell 12% on Feb 4, possibly oversold, structurally its cloud ERP is the most direct target for AI agents. The core value of ERP is “workflow and data accumulation,” and Claude Cowork plus enterprise data connectors are swallowing that “workflow” piece. Without rapid AI transformation, valuation multiples will keep declining.

Kingsoft (03888.HK) — WPS Office is the target of Claude Design and Cowork. Domestic protection lasts about 12-18 months; afterward, local giants (ByteDance, Alibaba) will intensify efforts in Chinese office software, rapidly diluting Kingsoft’s moat.

China Software International (00354.HK), Mobvista (01860.HK) — the former is IT outsourcing, directly replaced by AI; the latter is mobile ad SaaS, where Claude Cowork’s marketing plugin will gradually eat into long-tail ad workflows. Both are long-term structural negatives; short-term rebounds are not worth participating in.

Chinese AI Startups · A Completely Different Track

At this point, I must highlight Chinese AI startups separately because their situation with Claude Design is completely different from Silicon Valley’s.

To conclude — Claude Design’s direct impact on China’s “Six Little Dragons” of AI is nearly zero. The reason is simple: Anthropic is unavailable in China. Competitors like Zhipu, MiniMax, Moonshade, Baichuan, Jiayue, and Zero One are not competing with Claude but with each other, and with giants like ByteDance, Alibaba, Tencent.

But that doesn’t mean they’re without pressure. The pressure comes from another direction — the benchmark for model valuation in capital markets is being reset by Claude’s valuation curve.

Early 2026 in China’s AI scene looks like this:

![]###https://img-cdn.gateio.im/social/moments-5c95e5ad06-ad616376cb-8b7abd-badf29(

This is the 2026 Chinese AI unicorn differentiation map — only Zhipu, MiniMax, Moonshade remain as top-tier independent model companies, alongside giants like ByteDance, Alibaba, DeepSeek — forming a “Six-Strong” pattern.

The core issue for these independent model companies isn’t Claude Design — it’s their own burn rate:

![])https://img-cdn.gateio.im/social/moments-7fc0b358ad-eb2b39fdec-8b7abd-badf29$3B

They need continuous funding, and the benchmark is Anthropic — which has achieved $29B ARR without profit pressure because Google and Amazon have poured billions into it. Can Chinese model companies raise similar amounts? Currently, only through state-backed funds and Hong Kong IPOs.

Zhipu chose a Hong Kong IPO, MiniMax is following, and Moonshade plans to IPO in H2 this year. It’s a race for survival through fundraising.

Claude Design’s product launch isn’t direct competition for them but a signal that global investors are re-evaluating the valuation of “model + product” integrated solutions — Anthropic’s own product efforts have lowered the valuation ceiling for “model-only” companies.

Two weeks ago, I had dinner with a product director at Kimi, who said — “Claude Design’s release is a wake-up call for us, reminding us not to just sell models but to build end-to-end products and ecosystems.” That’s the next step for Chinese model companies — shifting from API sales to scenario-based solutions. Alibaba Cloud’s Baolian at 7.9 RMB/month for Qianwen 3.5, GLM-5, Kimi K2.5, MiniMax M2.5 are all laying the groundwork for this distribution channel.

Another often-overlooked line — application-layer AI companies. Chinese AI application firms like Natural Choice (AI companionship), Future Miao (emotion robots), MarsWave (generative audio) are actually more likely to survive than model companies because they target specific scenarios, not an endless arms race for compute. Claude Design is an opportunity, not a threat — if its design capabilities can be packaged as APIs, these companies’ development costs will drop sharply.

Next Shot · The Thing I’m Waiting For

Claude Design isn’t the end; it’s the third bullet from Anthropic Labs’ product factory (Cowork, Code, Design). Where will the next shot land? Two directions.

( Direction One · Full-stack Marketing and Growth

Claude Cowork already has a marketing plugin. If Anthropic launches “Claude Growth Studio” or similar, HubSpot (HUBS) could see a Figma-level crash in a single day, with Klaviyo, Braze, Marketo all following.

This is a highly probable scenario because marketing automation is the most quantifiable ROI product for Anthropic Labs next. If I want to hedge, I’d buy put spreads before HubSpot’s earnings.

) Direction Two · Analytics and BI

Tableau (owned by Salesforce), Looker (Google), ThoughtSpot. Claude Cowork already does data analysis; next, native dashboard generation and interactive data exploration are natural progressions.

This will benefit data layers like Snowflake and Databricks, but application-layer BI tools will face disruption. Salesforce, as Tableau’s parent, will continue to be under pressure.

( Direction Three · Further Erosion of Enterprise Dev Toolchains

Atlassian (TEAM)’s Jira, Confluence, and ServiceNow’s ITSM could all be restructured. ServiceNow, holding vast enterprise process data, might survive longer; Atlassian, lacking a deep “system of record” moat, is more vulnerable.

![])https://img-cdn.gateio.im/social/moments-c757a95a5a-0d71f57c7a-8b7abd-badf29$10B

The Zombie café staff are clearing cups from the early rush; the Stanford kid at the neighboring table has already left. Now, let me tell you what I did this weekend.

My Portfolio Actions

![]###https://img-cdn.gateio.im/social/moments-e8bc72b621-69f9d37527-8b7abd-badf29###

( The Core of This Market — Two Things

It’s not “AI killing SaaS,” but "LLM platform providers directly building applications, making wrapper companies’ valuations collapse."**

This is a structural, irreversible process.

Where Anthropic’s next shot lands, I don’t know. But I do know one thing — after four months and three bullets, the next shot will likely come in late May or early June. That day, the market will shake again, and we’ll be here watching another string of names.