AI and the elephant in the room: market concentration

Where are we in this technological cycle?

As with the rollout of other groundbreaking technology, the introduction of AI can be tracked across three broad phases, each posing their own hurdles for growth potential:

• Build-out: this is where we are now and is currently developing at pace, with a ramp-up of capital expenditure to build infrastructure, resulting in hardware providers seeing exceptional revenue growth.
  
  
• Adoption: As solutions mature, adoption rates (of which there remains a high degree of uncertainty) are expected to pick up. Whilst appetite and interest appear strong, surveys indicate we are still in the early stages of this phase at the business level, even if personal adoption rates are far higher.
  
  
• Productivity: The scale of productivity gains that can be made remain ambiguous. Depending on assumptions around the number of functions impacted and savings made, economists’ estimates range from 0.1% productivity boost to 1.5% per annum; it may take time for evidence to come through in economic data, although rising margins of early adopters may provide a marker. 

When visualising these phases, parallels can be drawn to credit card adoption during the 1950s to 1970s, or the internet’s emergence in the 1990s. At first, commercial businesses were slow to trust systems that felt alien. Mainstream adoption wasn’t driven entirely on the tech’s merit alone; but because the surrounding infrastructure matured, with confidence levels following suit.

AI finds itself in a similar position now. It’s capability is plain for all to see, but the workflows, governance, and skills required to embed it meaningfully in the economy are still playing catch up. We’re moving past the novelty stage, but are still far from the point where AI becomes a fully integrated default layer of private and public sector infrastructure. This points to a gradual evolution, which will be far from frictionless as new challenges emerge.

Additional factors shaping future growth

Geopolitics will play an outsized role

AI dominance has become a hotly contested square on a fast-moving geopolitical chessboard. When the US restricted sales of advanced Nvidia AI chips to China in 2022, the intention was to slow progress. Instead, it accelerated China’s push to develop domestic alternatives, exposing just how tightly global innovation is tied to US technology. While the US is likely to retain an advantage in cutting-edge chips, China has rapidly closed the gap in software, with the launch of DeepSeek R1 in early 2025 deliv­ering performance comparable to ChatGPT at lower cost.

China has also made clear that AI isn’t just about code and cloud servers. Its recent military parades showcased advanced robotics and autonomous systems. Indeed, since 2017, China has deployed more industrial robots than the US, Japan, South Korea, and Germany combined, accounting for 54% of the global total by 2024. Sustained efforts in R&D – Chinese patent filings exceeding 1 million in 2024 – have translated into tangible gains, with labour produc­tivity growth averaging 6% since 2021 compared to 1% in the US.

However, when technological lead­ership is at stake, the US is never likely to take a backseat and the government has shown a willingness to intervene in this critical sector. The AI ecosystem is set to benefit from significant state backing through subsidies, long-term contracts, and strategic partnerships. Recent collaborations between OpenAI, Google, and Palantir, to embed AI agents into US defence departments highlights how closely public policy and private innovation are now linked. In the global AI arms race, government support may prove just as important as computing power.