Inside Arcade’s marketplace model

When it comes to harnessing AI to increase personalization, one company has gone that extra step. Product creation platform Arcade is giving the power of design to the customer.

Launched in beta in late 2024, Arcade was created by Minted founder Mariam Naficy. Customers can input their idea for a piece of jewelry or home decor and use the site’s AI platform—powered by OpenAI’s GPT Image 1 model—to iterate on the item until the model creates the desired product. Arcade works with an array of independent artisans who are then paired up with the customer to create and price the item.

“We’re serving both sides of the marketplace,” Naficy said. “It’s a marketplace where we’re trying to put people together, so that makers can build more business, and the consumer can actually find the maker and talk to them using images.”

No inventory, no problem

All businesses operate with a degree of risk, but for Arcade, inventory is not one of them. Its marketplace model means manufacturing partners send products directly to the customer.

“It definitely reduces the cash you need to run the business,” Naficy said.

The cost-effective supply chain allows Arcade to offer quality goods to customers at a more accessible price. Arcade works with just over 30 jewelry and home decor artisans and designers across the world, including Odette, a sustainable silk supplier and textile manufacturer based in Hong Kong, and Aura, a family-owned jeweler based in Surat, India, among others.

Marketplace economics

The marketplace business model has been around since the onset of e-commerce, with Amazon being the best known of its kind. It’s a lucrative model for many companies looking to cut inventory costs.

“The major pros for sellers are discoverability, because obviously as a new seller, or a small artisan product maker, their challenge is marketing and finding customers when they’re busy making the products,” said Ben Donovan, insights lead at e-commerce analysis firm Marketplace Pulse. “A marketplace aims to bridge that gap because theoretically it should bring you the customers because of the traffic the marketplace is driving.”

On the downside, the marketplace model can sometimes pose high costs for the seller, namely referral or fulfillment fees.

“Over recent years Amazon, because it has become such a high volume marketplace, and there are millions of sellers on Amazon, it’s become very competitive and a bit more of a pay-to-play environment,” Donovan continued. “Sellers are having to essentially spend money on advertising to maintain their organic visibility. So that’s obviously a very mature marketplace. A new marketplace is not going to be anywhere near that level of take rate, but those are the kinds of the creeping economics that can happen as a marketplace grows.”

Fueling growth

Arcade took several steps last year toward success. Last May, it raised a $25 million Series A, which brought its total fundraising to $42 million. This coincided with its expansion into home decor (Arcade’s 2024 beta launch focused solely on jewelry).

The company then hired Varun Jampani as its chief AI officer in October, tasking the executive with AI research and development, specifically transforming natural language into physical product.

“Arcade is the first AI product creation platform that translates intent into physical reality,” Jampani said in an email statement. “We’ve built a network of suppliers and their multimodal models, trained on human-curated data, to generate precise product designs and pricing in real time.”

In the new year, Naficy doesn’t have any other category expansions planned just yet (although she’s gotten many requests for apparel and accessories), and instead wants to further develop Arcade’s abilities within jewelry and home decor.

“I really think that [2026] will bring more of these kinds of specialty [AI] models because everybody’s using the big underlying foundational models like Midjourney and obviously ChatGPT and Gemini…We’re in this middle layer where we provide these very specialized by taste models to people—taste, texture, and opinion—and that’s where I see us, this place where taste and AI meet,” Naficy said.