Part 2 - Forging forward with GenAI
Introduction
Welcome to the second part of our multipart blog post series on Artificial Intelligence (AI). The first blog post in this series can be found here.
A lot has been said about the benefits of adopting AI technologies such as generative artificial intelligence (GenAI) and none more so than the promise of productivity gains. Improving productivity in government and business is a common theme these days as this could potentially lead to cost savings while still effectively producing the same or better outputs, in theory. Efficiency and effectiveness have been studied in business for many years, resulting in new manufacturing methodologies and new business models.
In this blog post we examine the potential impact of the increasing demand for AI technologies, such as GenAI, could have on the world. We will use commentary of prominent business leaders from the technology and AI industries to shape this discussion.
Productivity gains
The saying “A rising tide lifts all boats” is generally associated with improvement in general economic terms that benefits all. This expression feels like it fits the excitement and hyperbole surrounding the promise of overall productivity gains that generative artificial intelligence (GenAI) will inject into the world. Satya Nadella, CEO of Microsoft, made an important observation regarding this. Mr. Nadella believes that GenAI and the technologies that it inspires must result in global gross domestic productivity (GDP) growth of up to 10% to justify all the investment (Cf the video).
Every few weeks or months a new breakthrough is announced that hails a more impressive ‘smarter’ large language model (LLMs). This can be seen with the leaderboards that track how these LLMs perform in a variety of tests such as solving mathematic problems, generating code, demonstrating reasoning capabilities, etc.1 2 3 4 At what point are the creators of these LLMs just gaming or hacking to get their models to score high in these rankings. The gains will be short-lived as these models reach a theoretical maximum with only marginal gains here and there. A breakthrough will be required, and this is where the next big promise of AI will have to step in – namely artificial general intelligence or AGI.
1 Chatbot Arena Leaderboard - a Hugging Face Space by lmarena-ai
2 LLM Leaderboard 2025 - Verified AI Rankings
3 LLM Leaderboard 2025
4 Aider LLM Leaderboards | aider
More than human
Dario Amodei, the co-founder and CEO of Anthropic, predicts that AGI will be like achieved by 2026/2027. Google DeepMind founder Demis Hassabis has a different view on this prediction stating that AGI has a 50% chance of being achieved by 2030 at the earliest. Hassabis also defines AGI as a system that can exhibit all cognitive abilities that humans can. Amodei’s AGI definition varies slightly in that he defines a model that can perform tasks human can with the competence of an expert in minutes or seconds.
Mr. Amodei also published a blog post that describes what he believes the benefits will be for countries that posse a smart and capable ‘super AI’. The benefit for those countries that achieve AGI first will be long-lasting but will require billions of dollars and access to millions of powerful specialized chips. The US wants to reach that milestone first to ensure that they have economic and military dominance over other countries, especially China.
It is not clear if AGI is required to fulfil Mr. Nadella’s benchmark of 5% real GDP growth in the developed world, but any system or technology that can augment a human to be more productive across all industries will have major implications. It is likely that this new technology will be used as a catalyst for good and bad. The short-term impact on society is not yet clear as the impact may be far reaching. We do know that tech companies are doubling down on their AI investments. In early 2025 Meta started laying off 5% of the lowest performing staff members to set the stage for further AI growth.5 6 7
This is not necessarily a reflection of what will happen across the board, but it does show how aggressive the push is for AI to rely less on humans.
5 Meta Is Planning to Cut 5% of Lowest Performers, Memo Shows - Bloomberg
6 meta firing 4000 employees: Meta Layoffs: 'Have a really, really difficult time...', says employee as Facebook parent firm starts firing thousands of workers - The Economic Times
7 I Was Labeled a 'Low Performer' at Meta and Laid Off. I'm Devastated. - Business Insider
We need more power
Users of the internet have been exposed to data and energy hungry systems long before the advent of GenAI. It does feel that the appetite and hunger of these systems have increased in orders of magnitude more than what we were used to. This increase in pace is unlocked by the ease of which platforms and systems are accessible to anyone that is willing to spend money. Many see AI as a must to differentiate themselves and their businesses. This is not just an irrational fear of missing out (FoMo), but a genuine desire for more capabilities.
Mr. Nadella post on X pointing to Jevons paradox in terms of AI seems to be consistent with his theme of GDP growth. Jevons paradox occurs when technological advancements make a resource more efficient. The economist Jevon noted that as the cost of using the resource, coal in the original context, drops as the overall demand will increase. This has implications for more efficient hardware. The true value of AI will be unlocked in terms of GDP growth when AI is commoditized, and the resource efficiency is high to ensure that anyone can use it.
What does efficiency thus mean in terms of GenAI? Mr. Nadella said he expects to see GDP growth because of AI. We know that building and operating LLMs are costly in terms of access to expert staff, specialized hardware, datacenters, energy costs in terms of electricity and cooling. Microsoft is betting on carbon-free nuclear energy to supply its power-hungry AI datacenters as this is perhaps a better short to medium term strategy in the absence of energy efficient hardware.
Reopening the Three Mile Island nuclear plant is said to contribute to job creating and indirect jobs in the surrounding community. This alone is perhaps not sufficient to boost the respective state’s local economy by much, let alone the world GDP. However, it is possible that this demand for more AI will indirectly require more investment in supporting sectors, thus stimulating the economy. Will this only be localized to developed countries, or will lesser developed countries also invest to feed power hungry datacenters that will host their helpful AI assistants?
The power demand from the datacenters that host AI workloads is considerable. xAI’s recently completed datacenter housing the Colossus AI supercomputer used in the creation of Grok3 LLM has special Telsa power banks that buffer the energy drawn from the power grid as the fluctuations put too much stress on the electricity grid or xAI’s own diesel generators.
Will developing nations be as fortunate to have access to their own datacenters to run their own AI backed systems? South Africa has some of the better infrastructure in Africa, but the nation has struggled with continuous rolling blackouts from 2008 as its electricity generation capability is strained due to poor maintenance of its infrastructure and lack of investment in generation capacity. It is unlikely that the South Africa power utility Eskom will have sufficient spare generating capacity to guarantee uninterrupted electricity supply to power hungry datacenters. Also, two of Eskom’s latest power generating units, Medupi and Kusile, are far over budget and missed several project milestones while suffering setbacks. It seems that Microsoft is not too concerned about this as Microsoft President Brad Smith announced in early March 2025 that Microsoft will invest R5.4 billion (~$280 million) into building data centers and train South Africans in AI skills.
Problematic power grids are not isolated just to Africa. At the end of February 2025 Chile’s power grid nearly collapsed as large areas across Chile was left without power due to a technical fault that rippled through the country. All this to meet the demand of the existing economy, let alone supporting AI initiatives.
Developing countries will require serious investment in infrastructure to keep up with the more affluent developed nations or succumb to being dependent on others for their AI workloads.
Crunching the numbers
To achieve 10% GDP global growth will require a substantial effort considering that the OECD projects global GDP growth for 2025 to be approximately 3.3%. Putting this into perspective, the projected GDP growth of the US for 2025 is expected to reach 2.8%. Countries like India, Indonesia, and China are projected to reach GDP growth of 6.9%, 5.2%, and 4.7% for respectively for 2025. Compare this with Spain, United Kingdom, France, and Germany that have projected GDP year-on-year growth of 2.3%, 1.7%, 0.9%, and 0.7% respectively.
To increase the global GDP by 1% has some real implications. We asked Google’s Gemini 2.0 Flash Thinking Experimental model the following question:
‘What does it mean to increase global GDP growth by 1 percent?’Here is what Gemini responded with:
To increase global GDP growth by 1 percent means increasing the rate at which the world's total economic output is expanding by one percentage point.
Let's break this down and understand what it implies:
1. What is Global GDP Growth?
- GDP (Gross Domestic Product) is the total value of all goods and services produced within a country's borders (or globally, in this context) in a specific period, usually a year. It's a key measure of economic size and activity.
- Global GDP Growth is the percentage change in global GDP from one period to the next (typically year-over-year). It reflects how much faster or slower the world economy is expanding.
2. Understanding a 1 Percentage Point Increase
Imagine the current projected global GDP growth for a year is 3.0%. An increase of 1 percentage point would mean the new growth rate becomes 4.0%. It's not just an increase of 1% of the GDP itself, but an increase of 1% on the growth rate.
3. What Does This Mean in Real Terms?
- Huge Increase in Economic Output: Global GDP is a massive number (in the tens or even hundreds of trillions of US dollars). A 1% increase on this base represents a very significant increase in the actual value of goods and services produced worldwide. Think of it as adding the economic output of a large country (or even several medium-sized countries) to the global economy in a single year.
- More Economic Activity: A 1% increase in growth generally indicates:
- Increased Production: Businesses are producing more goods and services.
- Higher Employment: More jobs are likely being created as businesses expand.
- Increased Incomes: Potentially higher wages and profits for individuals and businesses.
- More Spending: Increased consumer spending, business investment, and potentially government spending.
- Increased Trade: More goods and services being traded internationally.
- Impact on Living Standards (Potentially): Faster economic growth can lead to improvements in living standards on average. More wealth creation can translate into:
- Reduced Poverty: More opportunities for people to earn a living and escape poverty.
- Improved Healthcare and Education: Governments may have more resources to invest in social services.
- Technological Advancements: Economic growth often fuels innovation and technological progress.
4. Factors that Could Drive a 1% Increase in Global GDP Growth:
Achieving a 1% increase in global GDP growth is a complex undertaking. It could be driven by a combination of factors, such as:
- Technological Breakthroughs: Major innovations (like AI, renewable energy advancements, etc.) that significantly boost productivity and create new industries.
- Increased Global Demand: A surge in consumer and business demand worldwide, perhaps due to increased confidence or new market opportunities.
- Reduced Geopolitical Tensions: Improved international relations and reduced conflicts could lead to increased trade and investment.
- Effective Policy Changes: Governments implementing policies that stimulate economic activity, such as:
- Fiscal Stimulus: Increased government spending or tax cuts.
- Monetary Policy Easing: Lowering interest rates to encourage borrowing and investment.
- Structural Reforms: Policies that improve efficiency, reduce barriers to trade, and promote competition.
5. Important Considerations and Nuances:
- Distribution Matters: While a 1% increase in global GDP growth is a positive sign, it doesn't guarantee that the benefits are evenly distributed. Growth can be uneven across countries and within countries, potentially exacerbating inequality.
- Sustainability: The type of growth matters. Growth driven by unsustainable practices (e.g., excessive resource depletion, environmental damage) can have negative long-term consequences. Sustainable growth is key.
- Quality vs. Quantity: GDP measures the quantity of economic output, but not necessarily the quality of life or well-being in all aspects. Focusing solely on GDP growth can sometimes overlook other important societal goals.
- Base Effect: A 1% increase is more impactful when the base GDP is larger. If global GDP is already very high, a 1% increase represents a larger absolute increase in economic value compared to when the base GDP was smaller.
In Summary:
Increasing global GDP growth by 1 percent is a significant economic event. It represents a substantial increase in global economic activity, potentially leading to higher incomes, more jobs, and improvements in living standards. However, it's crucial to consider the distribution of benefits, the sustainability of growth, and the broader societal implications beyond just the numerical GDP figure. It's a complex goal that requires coordinated efforts and favorable global conditions.
What is striking about the Gemini response is that it mentions AI as a technological breakthrough that is required to help boost economic growth. Gemini also highlights that economic growth could be disproportionate, exacerbating inequality even more. Another interesting point is the mention of sustainable economic growth. The latter is something that seems to escape most conversations.
To increase real GDP growth from 3.2% to 4.2%, will require the global economy to grow several trillion dollars more to reach real economic growth when excluding inflation. The IMF forecasts that the nominal global GDP for 2025 is expected to be $115.49 trillion dollars while with only a real GDP growth of 3.2% (while OECD projected ~3.3%). This coming of the back of nominal global GDP of $110.06 trillion in 2024 that saw the same real GDP growth of 3.2%. The response that Google Gemini provided a key insight into this. It stated that 1% increase in the growth rate is equivalent to adding one large country’s GDP to the global economy. Could AI and all the activity supporting AI have such an impact that see a GDP growth of 5-10% even in the face of potential negative impacts of economic growth detractors raging at the same time?
At the start of March 2025 Goldman Sachs estimated that AI generated $340 billion in revenue from 2022 and is projected to increase to $580 billion by the end of 2025. The same reporting goes on to say that “real AI-related investment in US GDP terms only increased by $42bn over the same period”. Goldman Sachs attributed several reasons for this such as high inflationary actions such as semiconductor prices, AI-related sales growth happing outside the US, and archaic accounting practices in how semiconductor purchases are classified by the Bureau of Economic Analysis (BEA). If it’s this challenging to measure AI’s direct impact on the US’ GDP, then how easy would it be to measure the impact of AI on the world’s economy?
What came before
The internet is one technology that has contributed tremendously to the increase in productivity over the past decades. In 2011 McKinsey estimated that the internet accounted for 10% GDP growth over the prior 15 years but accounted for 21% of GDP growth in the prior 5 years as observed in mature or advanced economies (Note that this period overlaps with the global financial crisis of 2009 where the world GDP growth was negative 1.3% meaning a shrinking world economy for that period). Similarly, the Brookings Institute states that productivity increased 3% annually from 1995 due to earlier investment in computers and telecommunication networks. A blog by the World Bank attributes increases GDP growth in the 1990s with increased expansion of internet connectivity. The McKinsey study is particularly intriguing as that comes fresh of the back of a global financial crisis that saw GDP contraction in some developed countries.
The internet has truly become a commodity that has transformed business and society. It birthed various other technologies and enabled many businesses that never would have. During the dotcom bubble, there was a moment when it seemed like we might lose the internet, but it was just a reset on an overhyped concept. Would AI have the same impact, and would the world also have to endure a similar ‘dotcom bubble’ before the real value is unlocked?
Summary
Increasing global GDP growth by 1 percent is a significant economic event, with potential benefits such as higher incomes, more jobs, and improvements in living standards. However, the distribution of these benefits, the sustainability of growth, and broader societal implications need to be carefully considered. The Gemini response highlights AI as a crucial technological breakthrough for boosting economic growth but also warns of the potential for exacerbating inequality. The focus on sustainable economic growth is particularly noteworthy, as it is often overlooked in economic discussions.
In conclusion, the potential for AI to boost global GDP growth is significant, but it comes with challenges and uncertainties. Users, businesses, and governments will need to decide how to measure AI's success and justify investments in this technology. Will there be a ‘killer app’ that makes AI shine, or will AI wield and mold us to do its bidding? The need to balance technological advancement with human values and priorities must not be overlooked.
” To reclaim our humanity in the age of AI, we must prioritize the mastery of self. “