Enterprise AI in a box: SuperX’s new server ships with GPT preloaded

SuperX AI Technology launched enterprise servers Wednesday that promise to cut months of artificial intelligence (AI) setup work while keeping company data safe through advanced security features.

The Singapore-based company’s new All-in-One Multi-Model Servers come ready to use with OpenAI’s latest open-source language models. Businesses can deploy the systems immediately without lengthy integration processes that typically take months.

“A single model cannot solve the problems of a complex world,” said Kenny Sng, SuperX’s chief technology officer. “Multi-model collaboration is a vital step in the evolution of AI towards Artificial General Intelligence to serve people.”

The servers address two major barriers preventing companies from adopting AI: complex setup requirements and data security concerns. SuperX built the systems with NVIDIA’s Confidential Computing technology. This creates a secure environment that protects business information during AI processing.

Companies can choose from four different models. The basic AI Workstation costs $50,000 for individual use. Medium-sized businesses can purchase the B200 Standard Edition for $500,000.

Large enterprises requiring maximum power can order Cluster Editions starting at $4 million.

Each server comes loaded with GPT-OSS-120B and GPT-OSS-20B models that OpenAI released last week. Testing shows the larger model matches or beats several leading commercial AI systems in key performance areas.

SuperX designed the servers to handle multiple AI tasks simultaneously. They can process text, create images, recognize speech, and analyze documents all at once. This multi-model approach lets businesses tackle complex projects that require different types of AI working together.

Business users can build AI applications through simple interfaces without writing code. The servers include templates for common business tasks like document drafting, legal research, and policy analysis.

SuperX’s previously introduced its XN9160-B200 AI Server in July. That earlier product focused purely on raw computing power, while the new Multi-Model Servers emphasize ease of use and business applications.

SuperX rebranded from Junee Limited earlier this year and hired Sng, Intel’s former Singapore and Malaysia technology chief. The company operates a production facility in Japan that can build 10,000 units annually.

Despite reporting negative earnings of $0.55 per share in mid-2024, SuperX stock jumped 45.5% after announcing the server launch. Investors appear confident in the company’s shift toward AI infrastructure solutions.

SuperX plans to begin shipping the servers immediately to enterprise customers. The company expects high demand from finance, healthcare, and government sectors that need secure AI processing capabilities