The economics of AI: does the winner take all or do we all win?
Tech giants are spending hundreds of billions building out their AI capabilities. Will they be as dominant as they have in social media and search, or will the benefits of AI be distributed more widely across the economy? The answer will shape markets and returns for years to come.
Authors
Perhaps the most important trend in finance in the past 40 years has been the emergence of “winner take all” industry structures in technology. This has led to incredible success for a small number of technology companies – often known as the “Magnificent 7.”1 To give some context of where we are today:
- Nvidia and Microsoft are each bigger than the stock markets of the UK, China and Canada combined
- Apple is the same size as the Japanese stock market
- Amazon is bigger than the French market
- Alphabet is bigger than the Swiss market
[Source: Bloomberg, as of 31 January 2025]
These companies have been as successful as they have because of network effects and economies of scale. Let me touch on these two economic phenomena:
- Network effects: If everyone else is on it, you don’t go anywhere else! The success of networks like Facebook, Instagram and LinkedIn made it impossible for rivals to compete, with the likes of Myspace dwindling away.
- Scale economics: These platforms are very expensive to build, but once you reach scale, they are extremely cash generative and require very little extra capital expenditure (capex). In other words, they have increasing returns to scale. Elon Musk’s Starlink accounts for 60% of all the satellites that are currently orbiting the planet.2 Once they are launched there is very little maintenance capex and every new contract is pure cashflow, to either be reinvested or distributed!
In industries from PC software (Microsoft), internet search (Alphabet), online retail (Amazon) and social media (Meta) to high-end consumer hardware (Apple), most of the profits and market share have gone to one company.
Will this pattern be repeated in AI, with a few winners taking all the spoils? Or will everyone win? It feels like we may be at a fork in the road.
Alphabet CEO Sundar Pichai suggested the former looked more likely when he said last summer that: “the risk of underinvesting is dramatically greater than the risk of overinvesting.”3 This has led to a huge capex boom as tech companies build out their AI capabilities.
Investment commitments announced by the top four technology companies in recent weeks suggest capex is set to rise by roughly a third to about $330bn this year. Microsoft has said it will spend around $80bn, Meta as much as $65bn and Amazon more than the $75bn it spent in 2024. Alphabet said earlier this month it would spend around $75bn, well above earlier estimates. Due to their huge scale in cloud computing and now AI, these companies are often known as the “hyperscalers.”
Will AI be different?
The big question is whether the hyperscalers will benefit from network effects or platform economics in AI to the extent we have seen elsewhere.
In the case of the former, just because I use ChatGPT, Claude, Mistral or DeepSeek, does that mean that you should? I am yet to be convinced.
There is also a big question mark around future capital intensity. Will AI be “capital heavy” or “capital light” for the key players once they have reached scale? If it continues to be capital heavy, long-term returns to the hyperscalers may not be as high as enjoyed in the past or hoped for in the future.
New AI models have sparked further questions. The Chinese model DeepSeek, which is more efficient than US rivals, was the first serious challenge to Nvidia’s dominance of the AI chip market. It was followed in early February by Le Chat, developed by the French company Mistral, which is 13x faster than OpenAI’s GPT-4 and is completely open source. While GPT-4 cost over $100 million to develop, Mistral built Le Chat for $22 million.
Both DeepSeek and Le Chat use chips from manufacturers other than Nvidia, suggesting that other companies may be more of a competitive threat than previously assumed. The growth and margin assumptions for Nvidia – which are based on the idea of it being a “winner takes all” company - could be too high. Some commentators, including Louis-Vincent Gave at Gavekal, have suggested that DeepSeek is also a “Sputnik moment”4 for the hyperscalers, which have built their AI strategy around Nvidia’s high powered chips.
Nvidia stock has retreated in recent weeks, but the company is hardly on its knees. Its proprietary coding language, Cuda, is still the industry standard. And just because the DeepSeek and Mistral models are more efficient, that does not mean that leaner AI models will not benefit from the higher computational power offered by Nvidia’s best chips.
What does it mean for markets?
If AI turns out to be another “winner takes all” technology, we will probably continue to see very concentrated performance in equity markets.
Even so, we may not get people mentioning the “Magnificent 7” quite as frequently as they do today. I always thought the Magnificent Seven was an unwise choice of term, as in the 1960 film only two the seven survived. On further reflection, perhaps it is an accurate moniker, as seldom are the winners of the next leap in tech the same as the previous ones! Do you remember Nokia, Ericsson, Psion, Blackberry, Intel and Cisco to name but a few? In their defence, the network benefits of the “Mag 7” may provide a measure of protection compared to those mentioned above as they were all hardware companies.
A scenario where we all win, instead of the winner taking all, could be a very positive economic outcome. Lower cost AI could raise productivity and profitability, but estimates do vary! For the next decade:
- Joseph Briggs, Senior Economist at Goldman Sachs, forecasts an increase of 9.0% in productivity and 6.1% in GDP5
- Daron Acemoglu, Professor of Economics at MIT, forecasts an increase of 0.5% in productivity and 0.9% in GDP6
Despite the uncertainty, a scenario where the benefits of AI are more broadly distributed could lead to a very different performance from the stock market, with companies outside of the big tech names performing much more strongly than in recent years.
By way of example, cars and airplanes have clearly been truly transformational to economic productivity since their invention and commercialisation (I acknowledge their environmental impact but overlook it here for the sake of brevity). Given their high capital intensity and the cyclicality of their cashflows, however, very few auto manufacturers and even fewer airlines have proven to be good long-term investments. They consistently trade at valuations far below those of Nvidia and other stocks caught up in the AI excitement.
So are the AI leaders in a “winners’ curse” situation reminiscent of the telcos in 2000, financial stocks in 2008 or perhaps railroads in the late 19th century?
Only time will tell.
Acknowledgements
As the great singer Tom Lehrer once wrote:
I am never forget the day I first meet the great Lobachevsky
In one word he told me secret of success in mathematics
Plagiarize!
Plagiarize!
Let no one else's work evade your eyes
Remember why the good Lord made your eyes
So don't shade your eyes
But plagiarize, plagiarize, plagiarize
Only be sure always to call it please "Research"
This article was written having read widely on this topic over the past few weeks. I would like to acknowledge some excellent work from, amongst others, Deutsche Bank, Goldman Sachs, Morgan Stanley, JPMorgan, Panmure Liberum, Louis-Vincent Gave and Charles Gave at Gavekal, John Authers at Bloomberg and Robert Armstrong at the Financial Times.
Any reference to sectors/countries/stocks/securities are for illustrative purposes only and not a recommendation to buy or sell. This article may include forward-looking statements based upon our current opinions, expectations and projections. We undertake no obligation to update or revise any forward-looking statements.
1 Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, Tesla
3 Alphabet Second Quarter 2024 Earnings
4 The launch of Sputnik in 1957 shocked a complacent US to realise that it needed to increase investment in education and research and development to avoid falling behind.
5 Generative AI could raise global GDP by 7%, Joseph Briggs, Goldman Sachs, April 2023
6 The Simple Macroeconomics of AI, Daron Acemoglu, MIT, May 2024
Important information
This communication is marketing material. The views and opinions contained herein are those of the author(s) on this page, and may not necessarily represent views expressed or reflected in other Schroders communications, strategies or funds. This material is intended to be for information purposes only and is not intended as promotional material in any respect. The material is not intended as an offer or solicitation for the purchase or sale of any financial instrument. It is not intended to provide and should not be relied on for accounting, legal or tax advice, or investment recommendations. Reliance should not be placed on the views and information in this document when taking individual investment and/or strategic decisions. Past performance is not a reliable indicator of future results. The value of an investment can go down as well as up and is not guaranteed. All investments involve risks including the risk of possible loss of principal. Information herein is believed to be reliable but Schroders does not warrant its completeness or accuracy. Some information quoted was obtained from external sources we consider to be reliable. No responsibility can be accepted for errors of fact obtained from third parties, and this data may change with market conditions. This does not exclude any duty or liability that Schroders has to its customers under any regulatory system. Regions/ sectors shown for illustrative purposes only and should not be viewed as a recommendation to buy/sell. The opinions in this material include some forecasted views. We believe we are basing our expectations and beliefs on reasonable assumptions within the bounds of what we currently know. However, there is no guarantee than any forecasts or opinions will be realised. These views and opinions may change. The content is issued by Schroder Investment Management Limited, 1 London Wall Place, London EC2Y 5AU. Registered No. 1893220 England. Authorised and regulated by the Financial Conduct Authority.
Authors
Topics