Magnific ARR 突破 $300M,明年或尋求外部融資
Magnific passes $300M ARR and may raise outside money next year
Magnific 表示 ARR 已突破 $300M,創辦人兼行政總裁 Joaquín Cuenca 稱公司目前沒有外部融資計劃,但若引入資金有助加快發展,明年或會考慮。公司自 2010 年後未再進行融資輪,4 月由 Freepik 改名為 Magnific 時 ARR 為 $230M。Magnific 正在上海和新加坡開設辦事處,並計劃至 2027 年初在亞太區投入 $10M。
Magnific, the Spanish AI creative platform once called Freepik, has not raised a funding round since 2010. Its customers have paid for everything, including the move into AI, co-founder and chief executive Joaquín Cuenca told TNW.
“On outside investment, we have no plans today, but nothing is set in stone. The investment this market requires is very significant, and if taking capital lets us accelerate, it may happen, maybe next year,” Cuenca said.
Magnific said on Tuesday that it has passed $300M in annual recurring revenue. It had $230M when it rebranded from Freepik in April. It is opening offices in Shanghai and Singapore and will spend $10M in Asia Pacific through early 2027.
Magnific says that makes it Europe’s fourth-largest generative AI company by ARR, on publicly reported figures. Only Lovable, ElevenLabs and Mistral rank higher. Mistral raised €3bn in September, and ElevenLabs doubled its valuation to $22bn in a share sale.
Paid for by customers
Magnific is still bootstrapped, Cuenca said. EQT took a majority stake in 2020. It bought shares from his partners and never put new money into the company.
Magnific is EBITDA positive, he said. Profit is thinner than revenue suggests, because most of the money goes straight back into AI compute. Every time a better image or video model comes out, users want it that same day, he added.
“We already spend hundreds of millions of dollars generating what our users ask for,” Cuenca said.
An IPO is not on the table today, Cuenca said. Public markets now ask for far bigger companies than they used to. His target is to pass $1bn in revenue by the end of next year. He expects to get there organically.
“What I value is that there is no timer on us,” Cuenca said.
Chinese models
Several of the best video models are Chinese, and users pick them because they win on quality, Cuenca said. Magnific runs ByteDance’s Seedance, Kuaishou’s Kling, Alibaba’s Wan and MiniMax’s models today.
Until last year he had never been to China, and Magnific’s suppliers were American. He has been several times since, once for just 20 hours. China entered Magnific’s top 10 markets this year for the first time.
Magnific is not trying to build its own foundation model to take on Google, ByteDance or OpenAI, Cuenca said. For an independent company, he called that “a lost battle”.
“Think of a MacBook: the models are the CPU, and we build the rest of the machine, the tools, the workflows, the controls,” Cuenca said.
The Shanghai office puts Magnific next to the teams that build those models, he said, so it can add new ones faster and pass on feedback from its users. According to the release, Singapore will be its sales base in the region, alongside existing teams in South Korea, Japan and Thailand.
IP and studios
“Studios need legal guarantees before they use anything in production, and meeting that bar is one of the reasons we are growing with them,” Cuenca said.
Magnific’s Enterprise plans include indemnification against third-party IP claims on what the platform produces, Cuenca said. Customers own what they create, and Magnific does not use their inputs or outputs to train models, he said.
“The exception is bad faith: if you ask for Mickey Mouse and you get Mickey Mouse, that’s on you,” he said.
In his view, responsibility has to move to the output. AI can infringe without the user meaning to, so what matters is detecting it and flagging it to the user.
“With any tool you can infringe: if you can draw two overlapping circles, you can infringe Mastercard’s trademark,” he said.
Asia and the sales push
In Singapore, Hagia Labs made the films for telco StarHub’s Unlimited+ Plans campaign, with its creative partner Superson. StarHub’s green characters, the Hublings, fly across Asia and land in Japan. Hagia built the characters and settings in Magnific’s Spaces tool. Done the conventional way, Cuenca said, that would have meant travel, crews and a shoot in every location.
By Hagia’s own estimate, it took two to four weeks from brief to delivery. A conventional shoot would take two to four months.
Teams start by doing the same work faster, Cuenca said. Then they make what they could not afford before. Brands that used to advertise with a still image and a line of copy now make film-quality spots, and studios are greenlighting fantasy and historical projects that used to die on budget.
The next step is automation, he said. A team builds a workflow once in Spaces, and staff or their agents in ChatGPT or Claude can run it through Magnific’s MCP server.
Enterprise adoption in the region grew 35%, which is why Magnific is hiring there instead of flying people in from Málaga, Cuenca said. The release says Magnific plans more than 20 staff in Asia Pacific by the end of 2026, and twice that in 2027. Product and engineering stay at its Málaga headquarters. Cuenca called the offices “a long-term commitment to Asia” in Magnific’s release.

From self-serve to sales
Magnific says it has more than one million paying subscribers and over 400 enterprise clients. The biggest change in the last year, Cuenca said, is who those clients are: film studios, large marketing agencies and brands’ marketing teams. The company has moved from customers finding it on their own to a sales team, he said. About 45 of its 370 staff work in pre-sales, post-sales, demos and support. It also sends “forward-deployed creatives” to train clients.
Most AI labs do not have that, he said, and the subscriber base is still its pipeline: people use Magnific for years, then bring it into their companies.
“Being funded by customers forces a discipline I don’t want to lose: you only build what people will pay for,” Cuenca said.
來源:TNW · thenextweb.com