AI infrastructure firm Firmus has signed a multi-year deal with OpenAI to supply computing power from two data centres in Malaysia. As demand for the specialized computing systems required to train and operate increasingly sophisticated AI models grows, so too does OpenAI become a significant buyer of such equipment from Firmus.
The deal is particularly timely for Firmus, which has been making strides to expand its operations in the Asia-Pacific region. The company has garnered a lot of investor interest and was valued at over $10.5 billion in its last round of investments. Among the big investors were Nvidia, Jane Street, Blackstone-backed funds and Coatue Management, indicating that many companies are banking on the rise of companies that offer the necessary infrastructure for the AI industry.
Firmus’ signing of OpenAI as a customer is a significant commercial advancement. The models used by OpenAI are some of the most popular AI systems in the world, such as ChatGPT, and they need a tremendous amount of computing power. The AI systems are becoming more powerful and companies are building them require access to significant quantities of specialized hardware, stable power, data centre space. These deals enable AI firms to access data centre infrastructure for the long haul, and also offer the centralized data centre operators more transparency into future demand.
The Malaysian deal also helps to help to boost Firmus’ guaranteed computing power significantly. Since the deal, the company’s contracted capacity in its customer base has exceeded 900 megawatts. This is a testament to the sheer amount of computing power needed for today’s AI systems. The need to run COPS, networking gear, and cooling equipment with its high-speed processors all the time can require substantial amounts of electricity for large data centres that are serviced by AI.Large data centres powered by AI require considerable electricity, as they are required to run powerful processors, networking hardware, and cooling systems continuously.
Firmus is already represented globally with data centres. It has a portfolio of operational AI data centres in Australia and Singapore with several more under construction in the Asia Pacific region. The addition of Malaysia to the expansion makes another important addition to that network and puts the company within one of the fastest-growing data centre markets in Southeast Asia.
Malaysia’s strategic position, expanding digital economy and infrastructure accessibility for tech firms make the country a promising place for data centre investment. The country’s closeness to Asian markets has also enabled businesses to bring in extra computing power. In particular, Johor has emerged as an important data centre hub, due to its proximity to Singapore and the increasing demand for cloud and AI services.
But, the exponential growth of the data centres has also raised questions. They use a significant amount of electricity and water, especially as the hardware for advanced computing can generate a lot of heat and require complex cooling systems. The energy issues, water usage and the environmental footprint of data centre construction are increasingly on the agenda for more large operators.
The question is especially applicable in the midst of a period of strong infrastructure growth in the AI sector. To train and run sophisticated AI models, a set of specialised processors, known as GPUs, are required to handle large numbers of calculations. Nvidia is one of the biggest producers of these processors and Firmus’s association with the chipmaker is another way to prove its expansion strategy.
At scale in Asia-Pacific, Firmus will use Nvidia’s new Vera Rubin processors. The technology is expected to be able to handle the computing needs of more sophisticated AI applications. Supplying these systems to Malaysian facilities would further enhance the region’s position in the global AI infrastructure ecosystem.
The agreement also reflects the general race among technology companies to buy the physical machines necessary for artificial intelligence. Some of the most widespread discussions about AI are about software, models, and applications, but the industry relies more and more on a less visible layer of infrastructure. To operate AI services, data centres, semiconductor processors, power grid configurations, cooling solutions and high-speed connectivity are all necessary.
It’s not just about companies in the United States, this is an infrastructure race. All technology groups around the world are bidding for appropriate sites, power and computers. Data Centre capacity has become an increasingly valuable commodity as demand continues to grow, especially in areas where large data centres can be built.
OpenAI’s growing infrastructure needs demonstrate the dynamic growth of AI developer needs. The more complex and larger the model, the more computing power it requires. Companies are then required to plan their capacity in advance, not just short-term, to avoid the need for infrastructure arrangements.
This is good news for data centre operators like Firmus as they can build long-term relationships with significant AI service providers. An anchor customer can bring increased confidence in knowing there is a guaranteed demand for the capacity and facilitate investment in extra capacity. Meanwhile, users are able to access infrastructure optimized for their computing needs.
Another interesting aspect of the Firmus agreement is the timing, as the company is reportedly in talks with the public offering. The partnership with OpenAI might further enhance Firmus’ investor profile for its growth potential, should it go ahead with a listing. Its infrastructure business is being assisted by a large contracted customer and over 900 MW of contracted capacity.
The firm hasn’t revealed any monetary terms for its deal with OpenAI. Not having a contract value renders it difficult to see how much the deal will directly contribute to Firmus’ revenue right now publicly. OpenAI has also yet to comment on the deal.
