Who controls the roads to advanced AI?

Introduction

Advanced AI is often described as a software revolution, but its development increasingly depends on something more physical: access to vast amounts of computing power. Training and running frontier AI models requires specialised chips, large data centres, high-speed networking, energy infrastructure and cloud platforms that can coordinate them at scale. As a result, a small number of cloud providers have become critical intermediaries between AI developers and the computational resources they need.[GOV.UK]GOV.UKA I Foundation Models: initial reviewAI Foundation Models: initial reviewMay 4, 2023 — This initial review will help create an early understanding of the market for foundatio…Published: May 4, 2023

This has turned cloud infrastructure into a central governance question. If AI contributes to scientific acceleration, medical breakthroughs, automation, education and long-term human flourishing, then the firms controlling access to advanced compute may influence who can participate in that future. The issue is not simply market concentration in the ordinary business sense. It is whether a handful of companies effectively control the roads that lead to advanced AI, and what that means for innovation, competition, public research and democratic oversight.

Who controls the roads to advanced AI?

For much of computing history, organisations could buy their own servers and operate them internally. Advanced AI changes that equation.

Training a frontier model can require tens of thousands of high-performance AI accelerators operating together for weeks or months. Building facilities capable of supporting those systems demands enormous investments in power delivery, cooling, networking and engineering. Few organisations can afford to replicate this infrastructure independently.[GOV.UK]GOV.UKannual report and accounts 2024 to 2025Report and Accounts 2024 to 2025Jul 10, 2025 — The CMA will shortly conclude our market investigation into the cloud services market, whi…

As a result, the dominant cloud providers have become the practical gateway to advanced AI development:

Together, these firms consistently account for roughly two-thirds of global cloud infrastructure spending, depending on the measurement period. Multiple industry analyses place their combined market share above 60%, with AWS maintaining the largest share and Microsoft and Google following behind.[AAG IT Services]aag-it.comthe latest cloud computing statisticsMicrosoft Azure had a market share of 22% in Q1 2022, with Google at 10…[CRN]crn.comglobal cloud market share q4 2025 google grows aws lead narrowsGlobal Cloud Market Share Q4 2025; Google Grows, AWS…11 Feb 2026 — Combined, AWS, Microsoft and Google Cloud achieved 68 percent share…

This concentration matters because cloud providers are no longer merely renting servers. They increasingly control:

An AI startup, university lab or public agency may therefore depend on the same company for both general computing infrastructure and advanced AI capabilities.

Why cloud platforms became AI’s backbone

Several technical and economic forces reinforce the importance of cloud providers.

Compute is expensive and difficult to scale

The most obvious factor is cost.

Frontier AI systems require enormous quantities of computation. The infrastructure needed to support them is measured not in millions but often billions of pounds of capital expenditure. Data centres must secure electricity, cooling systems, networking equipment and specialised processors that are themselves produced through highly concentrated supply chains.[GOV.UK]GOV.UKartificial intelligence sector study 2024Intelligence sector study 20243 Sept 2025 — In 2024, there were 51 AI related inward investment projects into the UK, representing more t…

Cloud providers already possess many of these assets. They can spread costs across thousands of customers and continuously reinvest revenue into larger facilities.

This creates a powerful scale advantage. The more customers a cloud platform serves, the easier it becomes to finance the next generation of infrastructure.

Access to AI chips is often indirect

The cloud giants are also major purchasers of advanced AI hardware.

Rather than buying thousands of high-end accelerators directly, many organisations gain access through cloud rental arrangements. Even firms developing their own models frequently rely on cloud providers because obtaining sufficient hardware independently can be difficult or prohibitively expensive.[GOV.UK]GOV.UKservices market investigationThe Competition and Markets Authority (CMA) investigated the supply of public cloud infrastructure services…

This gives cloud providers substantial influence over who receives computing resources and under what terms.

During periods of hardware scarcity, access can become a strategic bottleneck rather than a routine commercial service.

AI services are increasingly bundled

The cloud market is evolving from infrastructure rental towards integrated AI ecosystems.[fortunebusinessinsights.com]fortunebusinessinsights.comcloud computing market 102697Cloud Computing Market Size, Share & Growth Report, 203420 Apr 2026 — The global cloud computing market size was valued at USD 781.27 bil…

Instead of simply providing compute, cloud providers increasingly offer:

For customers, these bundles can be convenient. For competition authorities, they raise concerns about dependency and lock-in.

A customer that trains models, stores data, builds workflows and deploys products through one provider may face substantial costs if it later wants to switch.

The rise of vertically integrated AI ecosystems

One reason regulators pay attention to cloud infrastructure is that concentration does not stop at cloud services.

Many leading AI firms are increasingly linked to major cloud providers through investments, partnerships and distribution agreements.

Microsoft’s relationship with OpenAI became one of the most prominent examples. Azure provides critical infrastructure for OpenAI services while OpenAI products help drive demand for Azure. Similar dynamics emerged around Amazon’s multibillion-dollar partnership with Anthropic, which included commitments to use AWS infrastructure for major workloads. The UK’s Competition and Markets Authority (CMA) investigated aspects of these arrangements because of concerns about their potential impact on competition.[The Guardian]theguardian.comThe CMA aims to determine whether the partnership constitutes a "relevant merger situation" that could warrant an in-depth review. Announ…

These relationships are not necessarily anti-competitive. They can provide AI developers with funding, engineering support and access to infrastructure that would otherwise be difficult to obtain.

But they also create a more vertically integrated ecosystem in which cloud providers may simultaneously be:

• Infrastructure suppliers.
• Investors.
• Model distributors.
• Platform operators.
• Competitors.

The concern is not that any single relationship automatically creates monopoly power. It is that repeated partnerships could gradually make a small number of infrastructure firms unavoidable intermediaries throughout the AI supply chain.

How bottlenecks affect researchers, startups and public institutions

The effects of infrastructure concentration differ depending on who needs access.

Universities and public research

Historically, universities helped drive major advances in computer science because researchers could access computing resources that, while expensive, were not entirely beyond institutional budgets.

Advanced AI changes the scale of the challenge.

Many academic groups cannot easily match the computational budgets available to large technology firms. This can limit their ability to reproduce results, conduct independent safety research or compete at the frontier of model development.[GOV.UK]GOV.UKcma ai strategic updateAI strategic updateApr 29, 2024 — This document provides a strategic update on the CMA's approach to AI. This is set out through the foll…

The result may be a shift in where cutting-edge knowledge is produced. Research that once occurred primarily in universities increasingly moves into private organisations with access to large-scale infrastructure.

Startups

Cloud infrastructure simultaneously lowers and raises barriers to entry.

On one hand, startups no longer need to build data centres. They can rent sophisticated infrastructure on demand and reach global markets quickly.

On the other hand, the largest AI systems require spending levels that remain difficult for smaller firms to sustain. Startups may become dependent on cloud credits, preferential partnerships or access agreements with much larger companies.[Business Insider]businessinsider.comTraditionally, AWS was a go-to platform for startups due to its scalable compute and storage offerings. However, the rise of generative A…

This can create a landscape in which many innovative firms exist, but a smaller number of infrastructure owners still shape the overall market.

Governments and public services

Governments increasingly use cloud platforms for digital services, data storage and AI deployment.

That creates practical benefits. Major cloud providers offer reliability, security expertise and global infrastructure that many public bodies would struggle to build themselves.

However, dependence on a small number of providers can also create strategic vulnerabilities.

A government deploying AI in healthcare, education, scientific research or public administration may discover that crucial capabilities depend on commercial infrastructure controlled by foreign firms. Questions of sovereignty, resilience and bargaining power become more significant as AI becomes embedded in state functions.[OECD]oecd.orgcomponent 6Market features in AI infrastructure: Competition in artificial…14 Nov 2025 — In the context of AI infrastructure, such competitio…

Does concentration speed progress or slow it?

One of the most important disagreements concerns whether concentration is mainly a problem or partly a source of progress.

Supporters of large-scale infrastructure argue that frontier AI would advance more slowly without hyperscale cloud providers.

Building advanced data centres requires extraordinary capital investment. Large firms can finance projects that universities, governments and startups might struggle to fund. Shared infrastructure may also reduce duplication and allow innovations to spread more quickly across industries.[OECD]oecd.orgMeasuring domestic public cloud compute availability for…by V Lehdonvirta · 2025 · Cited by 1 — This Working Paper develops a meth…

From this perspective, concentration is partly a consequence of technological reality. If advanced AI requires immense resources, then some degree of scale may be unavoidable.

Critics do not necessarily dispute the value of scale. Their concern is that infrastructure power can become self-reinforcing.

If the largest providers control compute, attract the most customers, secure the largest AI partnerships and earn the highest revenues, they may gain advantages that become increasingly difficult for rivals to challenge. Innovation can continue while competitive pressure weakens.[Open Loop]openloop.orgcompetition in ai foundation models and the cma ai principlesOpen LoopCompetition in AI Foundation Models and the CMA AI PrinciplesFeb 12, 2025 — This report presents the findings and recommendation…[GOV.UK]GOV.UKA I Foundation Models: initial reviewAI Foundation Models: initial reviewMay 4, 2023 — This initial review will help create an early understanding of the market for foundatio…Published: May 4, 2023

The debate is therefore less about whether cloud giants create value and more about whether the benefits of scale can coexist with meaningful competition and broad access.

Governance choices that could widen access

The central policy question is not whether cloud infrastructure should exist. It is how access to that infrastructure should be governed.

Portability and switching

Competition authorities have focused on making it easier for customers to move between providers.

Technical standards, data portability requirements and interoperability measures could reduce dependence on a single platform. If organisations can switch providers more easily, cloud firms may face stronger competitive pressure on pricing and service quality.[OECD]oecd.orgCompetition in the Provision of Cloud Computing ServicesIn June 2025, the OECD held a roundtable to discuss recent developments in cl…Published: June 2025

Public-interest compute

Some researchers advocate public or publicly supported compute resources.

The idea is not necessarily to replace commercial clouds but to ensure that universities, independent researchers and public-interest projects can access advanced computing resources without relying entirely on private gatekeepers.

Such proposals resemble earlier investments in scientific infrastructure, where governments funded facilities too expensive for individual institutions to build alone.

Scrutiny of partnerships and acquisitions

Regulators increasingly examine relationships between cloud providers and frontier AI firms.

The concern is that investments, exclusive agreements or preferential access arrangements could entrench infrastructure power before markets fully mature. The CMA’s reviews of foundation model markets and cloud infrastructure reflect this broader interest in preventing early bottlenecks from becoming permanent.[Practical Law]uk.practicallaw.thomsonreuters.compublishes update paper in initial review of AI foundation…On 11 April 2024, the CMA published an update paper in its initial review of…Published: April 2024

Geographic diversification

Another concern is geographic concentration.

Advanced cloud infrastructure is unevenly distributed around the world. Countries without significant domestic compute capacity may become dependent on infrastructure located elsewhere. OECD work has highlighted the importance of understanding where cloud compute is physically available and who controls it.[OECD]oecd.orgThe Organisation for Economic Co-operation and…The OECD designs international standards and guidelines for development co-operat…

If advanced AI becomes a key driver of scientific and economic progress, unequal access to compute could reinforce global inequalities rather than reduce them.

Why this matters for an AI-enabled human bloom

The optimistic case for AI is not merely that companies become more productive. It is that intelligence itself becomes more abundant.

In that vision, advanced AI could accelerate medicine, scientific discovery, education, engineering, climate mitigation and many other activities that expand human flourishing. The long-term promise of an AI-enabled civilisation depends on the widespread availability of those capabilities rather than their confinement to a small number of organisations.

Cloud infrastructure sits near the centre of that question.

If access to advanced computation remains broad, competitive and reasonably open, cloud platforms could function like enabling infrastructure: the roads, ports and power grids of an age shaped by machine intelligence. Researchers, startups, governments and citizens would all have opportunities to build on increasingly powerful systems.

If access becomes heavily concentrated, the same infrastructure could act as a choke point. Decisions made by a small number of companies about pricing, availability, partnerships and acceptable use might shape who can participate in the next phase of technological progress.

The future of AI abundance is therefore not only a question of algorithms. It is also a question of infrastructure governance. Who controls the compute, the chips and the cloud platforms may help determine whether advanced AI becomes a broadly shared civilisational resource or a capability mediated through a small set of commercial gatekeepers.[OECD]oecd.orgcomponent 5The impact of AI adoption on market dynamics14 Nov 2025 — This paper examines how the adoption of artificial intelligence (AI), particula…[GOV.UK]GOV.UKannual report and accounts 2024 to 2025Report and Accounts 2024 to 2025Jul 10, 2025 — The CMA will shortly conclude our market investigation into the cloud services market, whi…

Amazon book picks

Further Reading

Books and field guides related to Who controls the roads to advanced AI?. Use these as the next step if you want deeper reading beyond the article.

Book

Chip War

By Chris Miller

Explains how chips and infrastructure shape control over advanced technology.

See on Amazon

Book

The Coming Wave

By Mustafa Suleyman

Covers concentration and control of powerful AI infrastructure.

See on Amazon

Book

The Master Switch

By Tim Wu

Shows how control over communications infrastructure becomes market power.

See on Amazon

Book

Power and Progress

By Daron Acemoglu, Simon Johnson

Frames access to AI infrastructure as a distributional and governance problem.

See on Amazon

Browse more on Amazon:Chip WarThe Coming WaveThe Master Switch

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Shop location

UsingUSA USA USAUKCanadaAustraliaIreland

Example eBay listing

Server Data Center Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: data center poster

Browse similar oneBay.co.uk

Example eBay listing

Data Center Garden Work Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: data center poster

Browse similar oneBay.co.uk

Example eBay listing

An Advanced Futuristic Data Center Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: data center poster

Browse similar oneBay.co.uk

Example eBay listing

Data Center Interior Bristling With Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: data center poster

Browse similar oneBay.co.uk

Browse more oneBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

1. Source: GOV.UK
Title: A I Foundation Models: initial review
Link:https://www.gov.uk/cma-cases/ai-foundation-models-initial-review

Published: May 4, 2023

2. Source: assets.publishing.service.gov.uk
Title: AI Foundation Models technical update report
Link:https://assets.publishing.service.gov.uk/media/661e5a4c7469198185bd3d62/AI_Foundation_Models_technical_update_report.pdf

3. Source: aag-it.com
Title: the latest cloud computing statistics
Link:https://aag-it.com/the-latest-cloud-computing-statistics/

4. Source: crn.com
Title: global cloud market share q4 2025 google grows aws lead narrows
Link:https://www.crn.com/news/cloud/2026/global-cloud-market-share-q4-2025-google-grows-aws-lead-narrows

5. Source: oecd.org
Title: component 6
Link:https://www.oecd.org/en/publications/competition-in-artificial-intelligence-infrastructure_623d1874-en/full-report/component-6.html

6. Source: oecd.org
Link:https://www.oecd.org/en/publications/measuring-domestic-public-cloud-compute-availability-for-artificial-intelligence_8602a322-en.html

7. Source: GOV.UK
Title: annual report and accounts 2024 to 2025
Link:https://www.gov.uk/government/publications/cma-annual-report-and-accounts-2024-to-2025/annual-report-and-accounts-2024-to-2025

8. Source: oecd.org
Link:https://www.oecd.org/en/events/2025/06/competition-in-the-provision-of-cloud-computing-services.html

Published: June 2025

9. Source: crn.com
Title: cloud market share q1 2026 aws microsoft google battling in ai era
Link:https://www.crn.com/news/cloud/2026/cloud-market-share-q1-2026-aws-microsoft-google-battling-in-ai-era

Published: May 2026

10. Source: oecd.org
Link:https://www.oecd.org/en.html

11. Source: oecd.org
Title: component 5
Link:https://www.oecd.org/en/publications/artificial-intelligence-and-competitive-dynamics-in-downstream-markets_ccf0624a-en/full-report/component-5.html

12. Source: oecd.org
Title: competition in artificial intelligence infrastructure
Link:https://www.oecd.org/en/events/2025/12/competition-in-artificial-intelligence-infrastructure.html

Published: December 2025

13. Source: oecd.org
Title: component 5
Link:https://www.oecd.org/en/publications/competition-in-artificial-intelligence-infrastructure_623d1874-en/full-report/component-5.html

14. Source: GOV.UK
Title: artificial intelligence sector study 2024
Link:https://www.gov.uk/government/publications/artificial-intelligence-sector-study-2024/artificial-intelligence-sector-study-2024

15. Source: GOV.UK
Link:https://www.gov.uk/cma-cases/cloud-services-market-investigation

16. Source: GOV.UK
Title: cma ai strategic update
Link:https://www.gov.uk/government/publications/cma-ai-strategic-update/cma-ai-strategic-update

17. Source: state.gov
Link:https://www.state.gov/the-organization-for-economic-co-operation-and-development-oecd

18. Source: aws.amazon.com
Title: global infrastructure
Link:https://aws.amazon.com/about-aws/global-infrastructure/

19. Source: theguardian.com
Link:https://www.theguardian.com/technology/article/2024/aug/08/anthropic-ai-startup-uk-regulator-cma-amazon-investment

20. Source: uk.practicallaw.thomsonreuters.com
Link:https://uk.practicallaw.thomsonreuters.com/w-042-9673?contextData=%28sc.Default%29&originationContext=knowHow&transitionType=KnowHowItem

Published: April 2024

21. Source: businessinsider.com
Link:https://www.businessinsider.com/amazon-ai-startups-delaying-aws-spending

22. Source: openloop.org
Title: competition in ai foundation models and the cma ai principles
Link:https://openloop.org/reports/2025/02/competition-in-ai-foundation-models-and-the-cma-ai-principles.pdf

23. Source: Wikipedia
Link:https://en.wikipedia.org/wiki/OECD

24. Source: linkedin.com
Link:https://www.linkedin.com/company/organisation-eco-cooperation-development-organisation-cooperation-developpement-eco

25. Source: facebook.com
Link:https://www.facebook.com/theOECD/

26. Source: youtube.com
Link:https://www.youtube.com/oecd

27. Source: fortunebusinessinsights.com
Title: cloud computing market 102697
Link:https://www.fortunebusinessinsights.com/cloud-computing-market-102697

Additional References

28. Source: linkedin.com
Link:https://www.linkedin.com/posts/mark-haranas-98758667_global-cloud-market-share-q3-2025-aws-lowers-activity-7392270591216529409-EtSD

29. Source: instagram.com
Link:https://www.instagram.com/the_oecd/?hl=en

30. Source: nortonrosefulbright.com
Link:https://www.nortonrosefulbright.com/en/knowledge/publications/8ca8f277/the-uk-cmas-ai-foundation-models-initial-report

31. Source: windowscentral.com
Link:https://www.windowscentral.com/microsoft/as-microsoft-posts-usd75b-in-azure-revenue-uks-cma-says-competition-is-not-working-well

32. Source: linkedin.com
Link:https://www.linkedin.com/posts/charles-henry-monchau-cfa-cmt-caia-4003096_cloud-market-share-q1-2025-aws-29-activity-7325278131760603136-5DCH

Published: May 5, 2025

33. Source: linkedin.com
Link:https://www.linkedin.com/posts/kadir-tas-a1106a22_competition-in-artificial-intelligence-infrastructure-activity-7422611242634125313-Y6Pp

34. Source: thetimes.co.uk
Link:https://www.thetimes.co.uk/article/back-off-regulators-big-tech-buying-into-ai-isnt-a-bad-thing-nzd2klrqj

35. Source: linkedin.com
Link:https://www.linkedin.com/posts/swastis_cloudcomputing-digitaltransformation-aws-activity-7401926607164071937-iHk4

36. Source: bruegel.org
Title: why artificial intelligence creating fundamental challenges competition policy
Link:https://www.bruegel.org/policy-brief/why-artificial-intelligence-creating-fundamental-challenges-competition-policy

37. Source: drcf.org.uk
Title: harnessing innovation and growth opportunities from ai foundation models
Link:https://www.drcf.org.uk/publications/blogs/harnessing-innovation-and-growth-opportunities-from-ai-foundation-models