Greening chips and data centres
25 Sep 2024
In Keynesian prophecy, artificial intelligence (AI) productivity can be projected based on McKinsey’s guesstimate that generative AI would add 2% to 5% to the current gross domestic product (GDP). This figure is optimistic, if compared to a mere 0.66% productivity growth highlighted in a Massachusetts Institute of Technology study, whose projections calculated that not all jobs could unlock greater productivity with AI. Productivity gains could also reset as newer job roles require their own investments to cultivate, thereby diluting economist John Maynard Keynes’ possible future and sharpening the tension between AI’s uncertain gains and unpredictable impact.
Furthermore, as AI is computed in data centres, which are infamous as high energy and water guzzlers, the development and deployment of this contentious technology could be a double-edged sword.
Yet, amid these are the possibilities AI could hold for Malaysia.
AI could impact many aspects of Malaysia’s future. For example, ChatGPT could be seen as groundbreaking and hold immense potential across multiple fields. In learning, it could unlock new dimensions of human intelligence and potential. Conversely, it could be regarded as the mother lode of disruptors, bearing security risks and ethical issues where the cumulative advances in ChatGPT are said to have the potential of rendering many educators irrelevant.
The future of the Malaysian cadre could be shaped by the technology whose impact ranges from unexplainable to well-nigh dangerous. Yet, to restrain development and the economy from AI deployment could risk impacting the nation’s competitive future. After all, management consulting firm Kearney’s projected addition of US$1 trillion (RM4.2 trillion) to Southeast Asia’s GDP is a tantalising goal.
Rethinking AI policy
To achieve the good while mitigating the bad, Malaysia would have to reflect on domestic AI policies, not only on the means and ways the technology might impact society or the economy, but also to direct AI’s development as an industry to advance its goal of becoming a developed nation.
Malaysia is not short of plans to deploy or develop AI. The AI Roadmap 2021-2025 and 10-10 Malaysian Science, Technology, Innovation and Economy (MySTIE) Framework sought to strengthen the AI ecosystem for developers and R&D. Meanwhile, the New Industrial Master Plan 2030 (NIMP) cites AI as a possible sector to boost the country’s semiconductor design ambitions. NIMP hinges on economic complexity as Malaysia’s vision of becoming a high-income nation, whereby complexity is gauged by the nation’s productive capabilities to produce diverse and complex goods. However, the plans are not interlinked in an ecosystem which could kick-start an AI industry. This could take a page out of early AI policies in Japan and China that sought to invigorate the market by encouraging AI production in smart cities or smart-home appliances.
Malaysia should take advantage of the semiconductor industry and cultivate a value chain that has both software and hardware components, thus starting in code and possibly finishing in chips powering computers. This would not be impossible as the country has already displayed ambitions to move to the front end and enhance capacity in the back end of chips. Efforts for an integrated design park, such as the semiconductor accelerator and integrated circuit (IC) design park, are aimed at gathering local and global IC design houses to synergise collaboration. While this is indeed a laudable move to add value along the semiconductor supply chain, it is still too early to tell whether it could reap the low hanging fruit, especially if the complementary ecosystems are not present.
But is that enough? Certainly not, once we get down to the elephant in the data room — computing power — without which Malaysia’s AI future will remain bleak.
Environmental concerns
Calculations of computing power can vary, especially on the AI being trained. Yet, training any AI would consume energy, especially to dissipate heat. Furthermore, training AI produces more data, which means more space. Market intelligence provider TrendForce estimates that it would take 20,000 graphics processing units (GPUs) to train the generative pre-trained model underlying ChatGPT.
At commercialisation, the figure should reach above 30,000, especially because of data generation and user numbers. Chips would have to be produced with sustainability in mind, while data centres need to find ways to keep cool. In other words, the digital economy cannot be cleaved from the green economy, which explains the concerns about the flurry of data centres on Malaysian shores. Data centres account for 1% to 5% of the world’s total greenhouse gas emissions while, comparatively, emissions from the aviation industry make up 2% to 3%. It doesn’t end there, as the consumption of electricity will exceed 5,000mw by 2035 in Malaysia alone. On average, a data centre with the capacity of 100mw uses more than 4,000 cu m of water per day for cooling.
Between 2021 and 2023, Malaysia attracted RM114.7 billion in data centre investments, competing for limited resources in regions where they operate. And there’s the rub: Where do we draw the line between economic growth and environmental degradation?
The pressing question is whether these incoming companies have the necessary skills and resources to minimise their environmental impact and commit to green operations. The usual trade-off contentions apply: stringent regulations might raise concerns about slowing foreign investments. But the emphasis should lie on attracting high-calibre investments that align with the country’s long-term sustainability objectives, such as the National Energy Transition Roadmap.
Now, getting high and mighty about the imperative for going green may look good on paper but reality bites hard. In 2020, 50.9% of the peninsula’s electricity was generated from coal, which raises questions about our ability to supply sustainable energy to the expanding data centre sector. Achieving a balanced energy mix is crucial to support these dual objectives of aligning the nation’s renewable energy goals and digital ambitions. Hence, a cost-benefit analysis is needed to balance the economic gains with environmental sustainability.
Climate solutions
Malaysia could benefit from adopting a model like Singapore’s Green Data Centre Technology Roadmap by allocating capacity to data centre operators that prioritise sustainability alongside economic value. It is worth noting that the Malaysian Communications and Multimedia Commission (MCMC) introduced a technical code for green data centres in 2015. This technical code is now undergoing revision to match present technologies. Although it guides operators in enhancing energy efficiency and reducing carbon footprints, it remains non-binding and voluntary.
Meanwhile, the Ministry of Energy Transition and Water Transformation (Petra) and the Ministry of Investment, Trade and Industry (Miti) have announced that Malaysia can expect a robust framework focusing on energy and water efficiency. Undoubtedly a welcome move, this framework is set to introduce innovative solutions, moving from guidelines to enforceable standards.
Nevertheless, there needs to be active inter-ministry discussions to facilitate communication between the relevant agencies overseeing standards and compliance. For instance, the technical code serves as a baseline in developing a framework based on established principles and best practices. These need to be enhanced in areas, such as reliable and resilient water supply, water resource management and other critical public services. Siloed approaches would be counter-productive, especially in a cross-cutting sector such as information and communications technology (ICT).
Hosting more data centres, Malaysia should fully harness AI and technological innovations to advance transformative climate solutions for mitigation and adaptation in as much as creating a thriving digital economy must stem from a multi-dimensional approach that dynamically seeks to exploit the opportunities for growth and expansion while never losing sight of our planet’s limits.
Datuk Prof Dr Mohd Faiz Abdullah is chairman of the Institute of Strategic and International Studies (ISIS) Malaysia
Source: The Edge Malaysia
