AI cloud deployment startup Zeabur will shut down its platform after two major security incidents in roughly two months. Founder Lin Yuanlin told users in an Oct. 17 email that the second intrusion, which took place on Oct. 14, had a broader impact than the first breach in late August. Zeabur said it will start winding down services on Dec. 1 and will fully close dedicated servers and shared clusters on Dec. 31.
On the same day, Ling Wu, an early team member who previously served as Zeabur’s Founding Marketer, published a long retrospective that described the company’s internal difficulties. He said Zeabur had already laid off about half of its staff after the first security incident. He also wrote that the company had spent the past year dealing not only with security and operations issues, but also with U.S. market expansion, product positioning, and organizational decision-making problems.
Two security incidents in two months
Zeabur was founded about three and a half years ago and positioned itself as a platform that simplified cloud deployment, allowing developers to launch applications without having to master complicated server configuration. As AI development tools such as n8n, OpenClaw, and Vibe Coding gained traction, Zeabur also rode the wave of AI-native development and became one of the more closely watched startups in Taiwan’s tech scene.
That changed in the second half of this year, when the platform was hit by back-to-back security problems. In the first major incident in late August, an attacker exploited a vulnerability in a user-deployed NextChat service to gain internal access to a shared cluster. From there, the attacker used an older API key with insufficient permission controls, an AWS root key, and over-permissioned internal operations tasks to break into the production database and read environment variables tied to some customer projects.
The exposed sensitive data included API keys for AI services such as OpenAI, Anthropic, and OpenRouter, along with cloud service credentials, GitHub tokens, and database passwords. Zeabur said at the time that it would help affected users rotate credentials and provide compensation for verified extra costs.
Then, on Oct. 14, Zeabur suffered another external intrusion. This incident involved internal systems and some customer services. Lin said the team had contained all known attack paths, but the full investigation was still underway. The company had not disclosed the number of affected users or the scale of losses.
Lin said Zeabur began as a quickly built project three and a half years ago. As the business expanded and product direction kept shifting, the system accumulated complexity and maintenance burdens. He said the two incidents exposed shortcomings in the platform’s security architecture and operations management, forcing the team to reassess the resources needed to keep operating and whether it could bear the corresponding responsibility.
He also said advances in AI were lowering the barrier to application deployment, putting pressure on Zeabur’s original product value and its positioning. After internal discussions, the company decided to cease operations.
From a $2 million seed round to layoffs and growth strain
The shutdown stands out in part because Zeabur had been viewed less than a year earlier as a representative case of a Taiwan AI startup reaching international markets. The company originated from Lin Yuanlin’s graduation project while he was studying at Zhejiang University and later developed into a commercial cloud deployment platform. At the end of 2025, Zeabur completed a $2 million seed round led by U.S. venture capital firm 500 Global. At that point, it had accumulated more than 100,000 registered users, about 5,000 paying users, and had topped Product Hunt’s daily ranking.
But the funding and user growth visible from the outside did not fully reflect the pressure inside the company. In a post published on his personal website on Oct. 17, titled "Looking Back at Zeabur: Riding the Wind," Ling Wu said that after the first security incident in late August, Lin decided after weighing multiple factors to lay off about half of the team. Wu said he himself received a layoff notice in mid-September.
Wu also wrote that Zeabur benefited from the popularity of n8n in the second half of 2025, with revenue and user numbers reaching a period high. The company used that momentum to push into the U.S. market, and its marketing team at one point expanded to five people. But as product strategy shifted and staff members left, the original marketing team was eventually almost entirely dissolved.
Former founding team member points to management and decision-making issues
While Zeabur’s official explanation for the shutdown focused on the security incidents, technical debt, and market changes, Wu’s retrospective offered another angle: the company’s management structure and decision-making process may not have matured at the same pace as its growth.
Wu described Lin as an engineer with strong product instincts, combining low-level technical ability with a sense of product aesthetics. In Wu’s view, that was one reason Zeabur had been able to capture market opportunities more than once. At the same time, he said Lin had acknowledged that he was not good at handling people-related issues, and that weakness became more visible as the company expanded.
Wu said he had repeatedly suggested since the beginning of the year that Lin recruit a chief marketing officer in Silicon Valley who understood the U.S. market and had local networks, but no suitable candidate was found. Without a dedicated CMO, Lin was simultaneously acting as founder, CEO, and marketing lead. Wu said that split the founder’s attention across different functions and left the company without enough independent decision-making and execution capacity in market strategy.
He added that Zeabur remained, in essence, an organization where final decisions were led by the founder. Although the company used workshops and team meetings to seek consensus, some views raised by team members did not necessarily show up in final decisions. Wu said that created gaps between product development, marketing, and external communication.
Nuphos changed business models several times within months
Zeabur had already started exploring a direction beyond its original deployment platform in the first half of this year. In May and June 2026, the team launched a second product line called Nuphos, positioned as an AI DevOps Workspace for enterprise operations teams. The idea was to use the technical and operational experience built through running Zeabur to enter the enterprise market.
According to Wu, Nuphos went through several business model changes in just a few months. In June, the team decided to adopt a sales-led growth model and put resources into marketing and customer support. In July, it shifted to product-led sales while planning a launch event and different pricing plans. By September, the product had moved to an open-source model and removed its original pricing page.
Nuphos was formally launched in August through X and Product Hunt, but Wu said the launch results did not provide meaningful help for subsequent growth. He wrote that it is normal for an early-stage startup to keep adjusting its business model, but Zeabur’s problems in shaping product strategy and coordinating across departments often left the team struggling to understand the direction of decisions, making later marketing and go-to-market work harder.
Wu also said he believed Zeabur’s decision to stop operating its original platform may have been intended to concentrate resources on the new product line. He stressed, however, that this was only his personal judgment based on past team experience and should not replace the company’s official statement.
Threads helped growth, then accelerated the crisis
Wu said Zeabur had spent the past year actively building its presence on Threads, using the platform’s traffic and the founder’s personal brand to raise market visibility. That strategy did bring substantial exposure and helped a developer tool with limited name recognition quickly enter Taiwan’s tech community.
But when the security incidents hit, the same distribution dynamic also sped up the spread of negative sentiment. Wu described Threads as a double-edged sword for brand building: short-term traffic can drive growth, but when a company runs into a major controversy, the speed and emotional intensity of discussion on social platforms can move beyond what the team expected.