˂  Back



Artificial intelligence (“AI”) has been the talk of the town for more than a year now. The hype surrounding AI and the potential (think increased revenue and commercial advantages) that it may bring pretty much “convinced” many companies to now also brand themselves as “AI companies”, looking to deploy their own AI models.

It is an established fact that training and deploying AI models require a vast amount of computing power. Graphics processing unit (“GPU”) is a type of electronic circuit traditionally used for videos rendering, image creation and to process games with high quality graphics. Owing to its ability to perform and process many operations in parallel, GPU is also very suitable for AI training and generally any tasks that would require high computing power. The increase in the number of AI companies inevitably also translates to an increased demand for advanced GPUs, which in turn results in a supply shortage – like it or not, there are only that many companies globally capable of producing and supplying advanced GPUs.

GPU-as-a-Service, or “GPUaaS”, is an attempt by some companies to address the GPU shortage faced by the industry. GPUaaS is essentially a cloud-based solution that rents out access to GPUs to organisations that need them, on-demand. In this article, we will be breaking down some benefits of subscribing to GPUaaS and laying down some of the key considerations to take note of for companies thinking of resorting to GPUaaS.


Benefits of GPU-as-a-Service

Apart from allowing quicker access to GPU, GPUaaS also provides several other benefits to its subscribers, allowing easy justification to the stakeholders of companies in its adoption.

Cost efficiency is often one of the main benefits cited by companies when opting for GPUaaS as opposed to its on-premise counterpart. Instead of having to invest in and maintain physical infrastructure and specialised hardware, which also attracts other operational costs caused by energy consumption and cooling requirements, companies utilising GPUaaS only pay for subscription fees based on their project requirements.

Just like other cloud services, GPUaaS also offers the same flexibility and scalability to its subscribers. Users are often given the option to scale up (or scale down in some instances) their computing needs based on their project requirements, all without the need for having to invest in physical infrastructure and hardware only for that temporal increase in usage need.

Given the technicalities and specialised know-how that may be required to operate and maintain an on-premise GPU facility, it may not be worthwhile and commercially viable for companies that are not traditionally involved in this space to set up a new business unit just for this purpose. This is also one of the reasons why GPUaaS became the preferred choice of many companies traditionally involved in other businesses but now decide to venture into the AI space. By paying professional service providers for GPUaaS, companies can better focus their resources and attention on building their business and expanding their topline.


Legal Concerns of GPUaaS

We certainly cannot talk about all the advantages of GPUaaS without highlighting some key legal considerations that companies looking to offer or adopt GPUaaS should take note of.

  1. 1. Data Security
  2. For companies that may have concerns on storing their data on cloud or companies that are under strict regulatory requirements on data security, GPUaaS may not be the most suitable option. Granted that it provides flexibility, scalability and cost-savings, companies subscribed to GPUaaS are essentially relying on the GPUaaS providers to take charge on the security of their data. Companies with data security concerns should ensure that there are adequate data security assurances provided in the GPUaaS agreement with the service providers so that risks are allocated appropriately. Companies may also want to consider retaining the contractual rights to conduct audit on the security measures put in place by the GPUaaS providers.
  3. Where regulations impose data localisation requirements, companies should then enquire about the location of the physical facilities of the GPUaaS providers to ensure that the data localisation requirements can be met.
  5. 2. Termination Assistance
  6. It is of no surprise that companies subscribed to GPUaaS may actually store a vast amount of data on the cloud infrastructure of the GPUaaS provider. Considering the possibility that these data may be of mission-critical to the companies, it is crucial that the companies secure some commitments from the GPUaaS provider for the rendering of termination assistance covering the migration or transition of these data, either to the companies’ own GPU facilities, or a third party outsource service providers in the event of a termination or expiry of the GPUaaS agreement, including stating clear timelines and responsibilities for the termination process to ensure a smooth transition and minimize the risk of data loss or downtime.. This is all to ensure that in the event of a termination or expiry of the GPUaaS contract, the companies will not suffer any unplanned interruption of its business operation.
  8. 3. Service Levels
  9. In a GPUaaS arrangement, given that the operation of the GPU is beyond the direct control of the companies, the agreement for GPUaaS should address the service level that the GPUaaS provider is committed to. It would be of utmost importance that the GPUaaS agreement at the very minimum provides for the service levels of remedial action that the GPUaaS provider should take in the event of an unplanned service interruption or downtime.
  11. 4. Licensing Requirement
  12. GPUaaS at its core is essentially a form of infrastructure-as-a-service (“IaaS”). Some countries may actually require the providers of IaaS to obtain certain licence(s) before they can operate within the jurisdiction. Malaysia for one, imposes a legal obligation on either the IaaS provider with locally incorporated company, or a foreign IaaS provider that utilises a local data centre, to obtain an Application Service Provider (Class) licence before it can offer its services here in Malaysia. As such, it is important for companies looking to deploy GPUaaS in any jurisdiction to ensure appropriate due diligence is conducted prior to commencing operation. Conversely, companies looking to subscribe to any GPUaaS should also conduct simple verification to ensure that the service provider indeed has the required licences to conduct its business, so that unwanted interruption to the subscribed services can be avoided.
  13. .


GPUaaS is certainly a creative way to address the GPU crunch suffered by the industry. That being said, companies considering subscribing to GPUaaS should not dive headfirst, but should instead work with internal stakeholders and external advisers to evaluate the needs of the business against what GPUaaS could offer, in order to ascertain whether GPUaaS is the right fit for the organisation, or whether the organisation would be better off securing its own physical infrastructure and hardware.

Considering the nuances of GPUaaS, companies should conduct a holistic review of the GPUaaS agreement offered by the service provider to ensure that the companies’ needs and requirements are sufficiently addressed in the agreement.


If you wish to enquire more about GPUaaS, or if you are thinking of subscribing to GPUaaS, please feel free to reach out to the lawyers from our Technology Practice Group. We would certainly be delighted to assist in this exciting endeavours.

About the authors

Lo Khai Yi
Co-Head of Technology Practice Group
Technology, Media & Telecommunications, Intellectual
Property, Corporate/M&A, Projects and Infrastructure,
Privacy and Cybersecurity


Ong Johnson
Head of Technology Practice Group

Transactions and Dispute Resolution, Technology,
Media & Telecommunications, Intellectual Property,
Fintech, Privacy and Cybersecurity

More of our Tech articles that you should read:

Our Services