Present day
Capital as Strategic Asset
The current AI ecosystem is defined by a massive surge of private capital, much of it concentrated in multi-billion-dollar rounds for a small handful of frontier model developers. Increasingly, AI development is viewed not merely as a commercial opportunity but as a strategic asset to be secured in a geopolitical contest between nations — for technological, economic and military supremacy.
The scale of that investment, and its concentration, has become entangled with state efforts to secure an edge in what is widely framed as a race for global influence and hard power.
Present day
The Geopolitical Players
US v. China
Influential US venture capitalists and the current federal administration are pushing a winner-takes-all, tech-accelerationist narrative framed primarily around beating China.2 In Washington the contest is viewed not just as economic competition but as a defining geopolitical struggle: China is seen as the only country capable of marshalling the economic, military and technological resources to threaten US primacy,3 and “winning” — defined in the US as attaining AGI first4 — is considered essential to prosperity, national security, and defending democratic values against a model that may embed censorship and state surveillance into its technology stack.
Chinese players, however, seem less focused on AGI as a goal5 and more on boosting adoption now: making deployment cost-efficient by prioritising model efficiency and developing open-source models,6 and delivering the benefits of AI to the real economy.7 While the US leads on frontier AI as a “brain” — chatbots, microchips, LLMs — China has channelled resources into scaling industrial capacity through embodied AI, integrating it into humanoid robots. China stands to dominate the transition of AI into industrial capacity; US strengths lie in AI as a service.
In the US, the threat of losing to China is continually weaponised8 by business leaders, including VCs, to lobby for deregulation.9
The middle powers and sovereignty
The competitive mindset extends beyond the US and China. In the EU there has been an emphasis on tech sovereignty, sparking a rush to develop domestic AI industries. The European Commission’s technological sovereignty package10 aims to help Europe “become a leader in AI” — strengthening semiconductor capacity, accelerating data centre and cloud buildout, and developing homegrown models. For middle powers, domestic capability is seen as vital to avoiding over-dependence on foreign powers; as the US becomes an increasingly unreliable ally for Europe, a native tech stack is treated as necessary for national security and economic competitiveness.
The Global South
While the Global South presents a significant investable opportunity for AI diffusion, there are concerns over whether developing regions will share in the benefits.11 The vast majority of investment is concentrated in the North,13 and a lack of digital infrastructure limits access. The labour of AI development is also largely extractive: workers are employed for low-wage, traumatising gig work — content moderation and data labelling — to build the very systems that could automate their jobs.
So far the investment gap has been largely filled by China. Through its Belt and Road and Digital Silk Road initiatives, Chinese state and private firms have built more than $22 billion in digital infrastructure across 106 countries.12 Given China’s focus on cheaper, more efficient models, it is also better positioned to export and subsidise AI diffusion in emerging markets. Neglecting the Global South could represent a strategic miscalculation for the US, if China’s comparative willingness to invest14 strengthens its geopolitical influence.
Mid 2026
A Race to the Bottom?
The prevailing belief for both global powers is that AI represents a novel general-purpose technology, much like electricity, and that the country achieving dominance will harness profound systemic advantages — translating directly into hard power: industrial capacity, drug discovery, energy technologies, and military capabilities like new missiles, autonomous drones and defence systems.
Experts warn15 that accelerating competitive dynamics sideline domestic regulation, human rights-based frameworks, and commitments to international cooperation. It doesn’t help that there is a lack of internal expertise and guidance available to investors and founders to help them navigate the complex societal implications of the technology they are building.16
This could create severe political, social and economic risks, both domestically and globally: job displacement and rising inequality (see Scenario II on labour and productivity), environmental degradation (see Scenario I on energy and water), cybersecurity and bioterrorism risks, and the automation of warfare (see Scenario III on defence and dual-use).
Late 2026
Three Regimes, Diverging
All of these tensions unfold against an increasingly fractured global governance backdrop.17 We are entering an era of strategic fragmentation,18 where major jurisdictions are implementing different, and often incompatible, rules.19
Has enacted the EU AI Act,20 a comprehensive framework prioritising fundamental rights, transparency and safety, backed by the threat of significant enforcement actions and massive fines.
Largely pursuing a deregulatory agenda,21 seeking to curb state-level AI laws — though a patchwork of state regulation has nevertheless continued to emerge — in order to encourage innovation, attract investment and maximise speed.
Forging its own path by seeking to become a global leader in AI governance,12 establishing a global AI cooperation body to set standards. Domestically it tightly regulates models to reflect Communist Party values, embedding censorship and surveillance.
As the regulatory gap widens, regulatory arbitrage is likely to intensify.18 Harmonisation remains a remote chance absent a major catastrophic AI incident that catalyses international cooperation. The rift between the EU and the US could also accelerate Chinese AI development: if transatlantic relations deteriorate, allied nations could break with the US, potentially granting China access to critical technologies.
2027 · Branch point
Where, Then, Do We Go From Here?
Three futures follow. The first continues the current trend into a splinternet. The second is a full race to the bottom, which splits again depending on whether the frontier delivers. The third treats compliance as the route to a global market. Choose one to read first — all three are below.
2027 → 2031
Fragmentation into a Splinternet
Despite intense competitive pressure from the deregulatory agenda of the US, jurisdictions like the European Union press forward with their own AI regulations. As the US government proves increasingly willing to weaponise its tech sector — by limiting access to frontier technologies or cloud infrastructure — regions accelerate their efforts to secure strategic independence.
The global digital landscape fractures into a splinternet, leading to a loss of the economies of scale that traditionally make venture capital returns possible. As the US, EU and China diverge in their policy preferences, companies striving to operate across markets are compelled to build costly, jurisdiction-specific models to navigate fragmented compliance obligations.
Faced with this fragmentation, investors largely choose the path of least resistance — backing startups that can operate in a single regulatory regime rather than ones that need to satisfy incompatible rules across markets, and deprioritising multi-jurisdictional compliance-by-design as a due diligence criterion.
Demand-side pressure fails to correct this, as enterprise procurers have no consistent baseline to hold vendors to. Without credible third-party evaluation bodies or certification providers in permissive environments, buyers lack the information and leverage to distinguish compliant products from non-compliant ones, so no market mechanism forces convergence toward stricter standards. Startups attempting to build for multiple jurisdictions get seen as capital-intensive, slower-return bets that investors avoid — removing the very market actors who might have created a unified market. Startups withdraw from international markets entirely, isolating nations into separate AI islands.18
2032 → 2035
The Global South Is Left Behind
Without a unified global regulatory framework, and without investor pressure to enforce fair compensation or working conditions in portfolio companies’ supply chains, AI companies continue extractive practices in the Global South. Because developing nations often compete to attract foreign investment, they lack the political leverage and fiscal capacity to unilaterally impose strict regulations, tax tech companies, or demand revenue-sharing without the risk that companies simply relocate.
At the same time, the region is largely ignored by investors, who see the deregulated US as offering faster, simpler returns. The Global North reaps the productivity benefits while the Global South is locked out of high-value roles, instead bearing the environmental costs and the brunt of job displacement.
China steps in to fill the gap, expanding soft power by treating AI diffusion as a strategic, state-led initiative — subsidising data centres, fibre optic cables and energy grids alongside the tech. When developing nations adopt Chinese infrastructure, they inherit systems designed with Communist Party characteristics, including the export of “smart city” and “safe city” surveillance technologies that empower local authoritarianism through facial recognition and biometric tracking, increase Beijing’s espionage capabilities, and enforce censorship on sensitive political topics.
Meanwhile no major breakthrough occurs. The loss of economies of scale means AI startups can no longer operate internationally, lowering revenues and R&D; talent and knowledge fragment without global mobility and open collaboration. Whether AGI will be attained remains unresolved — which heightens the stakes, strains state and private patience over unclear ROI, and prolongs the race, with each power still convinced it could be first. This is what can tip the world into the worst case.
2027 → 2030
Global Deregulation
The public narrative is captured by investors and hyperscalers who frame any attempt at establishing guardrails as a threat to national security and economic competitiveness. Driven by a mandate of “winning at all costs” against China, the US federal government completely abandons regulation.
Investor lobbying for deregulation succeeds, and the global AI landscape devolves into a relentless race to the bottom spearheaded by the United States. To enforce this agenda globally, the US weaponises trade — sustaining trade wars and explicitly threatening tariffs against the European Union. The European Union, succumbing to the pressure, rolls back regulatory efforts like the EU AI Act.
From this highly deregulated baseline, the path splits into two sub-scenarios, based on what happens at the AI frontier.
2031 → 2035
Two Ways Down
Scenario A · AGI arrives
Capability Without Alignment
Investors continue to channel capital into AI scaling, prioritising speed without regard for safety. Capabilities improve dramatically, spearheaded by the US, but without globally harmonised efforts to govern safety and alignment. The arrival of AGI sets off its own competitive dynamic among buyers: consumers and enterprises face intense pressure to integrate it as fast as possible before competitors do. Systemic threats materialise:
- Cybersecurity risk and terrorism — mass deployment drastically lowers the barrier to inducing catastrophic real-world harms, empowering rogue actors and amateur hackers to bring down critical infrastructure, hospitals or energy grids.
- Autonomous weapons — as relations deteriorate, the superpowers prove unwilling to agree restrictions on military AI. Forecasters say the escalating arms race could prompt NATO states to authorise fully autonomous weapons around 2040.22
- Erosion of shared reality — autonomous agents and deepfakes flood the digital ecosystem, destroying the public’s ability to discern reality or hold democratic elections.
- Exacerbated inequality — mass job displacement sends inequality skyrocketing, increasing social polarisation to an unprecedented level.
- Climate impact — data centre energy demand grows unchecked; consumer electricity bills skyrocket; sociopolitical backlash is fomented.
Although the US has technically won, it destabilises itself in the process. Vulnerability to catastrophic cyberattacks and democratic erosion weaken its ability to project power and maintain alliances — one of its core advantages over China’s authoritarian model. Investors, having lobbied for and profited from the deregulated race, are themselves exposed to the same risks they hadn’t priced in.
Containment proves not to be watertight. Where nuclear weapons require rare materials and large-scale industrial infrastructure, AGI is much easier to leak. Through espionage, defectors and the general diffusion of knowledge, the rest of the world catches up — giving actors across the globe the same capacity for autonomous weapons and cyber attacks. Far from a clean geopolitical victory, AGI opens the US to systemic vulnerabilities and domestic chaos while China rapidly reaches the frontier, acquiring the same catastrophic capabilities.
Scenario B · The bust
Technological Disappointment
Hyperscalers fail to deliver on AGI. Investors’ continued pouring of capital into scaling, past the point where returns can be justified, makes the economic bust severe. Demand fails to act as a stabilising force: having lost confidence through an era of hype and unmet promises, and facing the reality that compute and token costs outweigh the ROI delivered, consumers and enterprise buyers pull back sharply.
Technology company valuations plummet. The economic fallout causes rising unemployment and a sociopolitical crisis. Countries turn inwards, sparking a surge in populism, nationalism and geopolitical isolationism. Newly-built data centres sit unused, representing a huge waste of natural and economic resources.
Even though AGI is never reached, the technology has already inflicted significant damage. The concentration of power and capital into a small number of tech giants has entrenched inequality, pitting a displaced entry-level working class against an ultra-wealthy tech elite, and the proliferation of generative AI has caused a severe loss of shared truth through the balkanisation of media.
China proves more insulated. Its state-directed model manages the correction behind closed doors — scaling back and redirecting investment while the West experiences panic selling — and its pre-existing lead in embodied AI delivers productivity gains independent of AGI. China emerges without the same financial and reputational damage, while isolationism sweeping through Western democracies boosts its global standing. In the Global South, demand for data labellers collapses alongside the industry, while China’s infrastructure remains standing, deepening reliance on Beijing.
2027 → 2030
Compliance as Competitive Advantage
Institutional investors and VCs realise that AI companies must meet the highest global standards to achieve economic integration and access a global total addressable market. Enterprise AI businesses are heavily incentivised to comply with the strictest regulations worldwide, such as the EU AI Act, and investors start focusing on funding the companies that can navigate them.
This shift coincides with the demands of large multinational corporations and financial institutions, which begin refusing to procure “black box” AI models due to their own compliance and liability risks. Commercial demand pushes the market towards transparency and responsible governance, creating a de facto regulatory ceiling regardless of the US federal government’s deregulatory stance.
Startups that embed responsible AI governance — safety, privacy, compliance — become more competitive, more likely to win lucrative enterprise procurement contracts and to sell to cautious foreign governments. Guardrails like the NIST frameworks continue to drive adoption in both government and corporate sectors. Private capital concentrates in responsible AI, and the market turns to focus on AI that is governable and capable of delivering genuine ROI.
2031 → 2033
Political Realignment and Shared Guardrails
As the industry moves toward compliance-by-design, and public concern around AI’s societal impacts grows, governments become more inclined to regulate. The EU AI Act becomes simpler to defend globally when major US enterprises are already meeting its standards. In Washington, the domestic business coalition opposing international coordination shrinks as compliant companies become the market winners.
Moving beyond narratives of a winner-takes-all arms race, nations recognise the shared threats of unchecked AI and pivot toward building a technology that is beneficial for all. Open-source AI is treated as a shared resource for humanity, governed by a global alliance that successfully coordinates safety across geopolitical boundaries.
While competition between the US and China persists, both powers recognise that certain risks are too catastrophic to leave ungoverned. Bioweapons and fully autonomous weapons pose existential threats that neither side benefits from unleashing, and the two ratify binding agreements restricting their development and deployment. The race for AI dominance continues — with guardrails that redirect competition towards building trustworthy and safe AI.
2034 → 2035
The Global South as an Investable Opportunity
Smaller, use-case-specific AI models prove easier to govern and make trustworthy than frontier models. This makes the Global South more attractive as an area of investment, given demand for localised applications in the region — driven both by specific use cases and by constraints on compute and connectivity.
Recognising the market gap — deploying AI to democratise local education, optimise agricultural yields, reduce logistics costs for medical deliveries in remote villages — investors allocate capital towards startups dedicated to these regions.23
China continues to invest through the Digital Silk Road, and its advantage in affordable, efficient models means it remains a significant presence in emerging markets. But it faces genuine competition: Western investors, attracted by the commercial opportunity in localised AI, are building an alternative to Chinese infrastructure dependency. Developing nations are no longer choosing between Chinese investment or nothing — they have options, which strengthens their political leverage and creates a more equitable global order.
Side by side
The Futures Compared
The same variables across all four outcomes, including both branches of the deregulated race. Note that in three of the four columns, China’s relative position improves.
| Variable | Splinternet | Deregulation · A | Deregulation · B | Race to the top |
|---|---|---|---|---|
| Regulatory state | EU maintains its own rules; US pursues deregulation; China governs domestically around CCP values while exporting infrastructure; the world fractures into distinct “AI islands” with rising regulatory arbitrage. | US abandons all regulation; the EU rolls back the AI Act under pressure; no international coordination; a complete race to the bottom, with the US enforcing deregulation globally. | EU and NIST guidelines become a global standard. | |
| VC / investor choice | Back startups in less-regulated markets; abandon compliance-by-design as a due diligence criterion; defund multi-jurisdictional startups as slow, capital-intensive bets. | Lobbying successfully dismantles all guardrails globally including the EU AI Act; capital pours into scaling with no safety or compliance filters; “winning at all costs” dominates. | LPs and VCs recognise compliance unlocks global TAM; fund startups meeting the highest standards; enterprise procurement drives commercial demand for trustworthy, transparent AI. | |
| AGI attainment | Not reached — fragmentation reduces economies of scale; talent siloed across jurisdictions; R&D slows; unclear ROI strains state and investor patience. | Reached by the US — but without safety or alignment frameworks, enabling catastrophic and large-scale systemic risks to materialise. | Not reached — but prior proliferation of generative AI has already caused significant societal damage before the bust. | Pursuit of frontier AGI becomes less central; focus shifts to governable AI delivering genuine ROI; if advanced AI is achieved, it is developed within internationally coordinated safety frameworks. |
| US v. China | US leads on frontier AI-as-service; China leads on embodied AI and model efficiency; competition intensifies with no decisive winner; middle powers accelerate domestic stacks. | US technically “wins” but destabilises itself — cyberattacks, democratic erosion and upheaval weaken power projection; AGI diffuses via espionage and defectors; China catches up, acquiring the same catastrophic capabilities. | The US suffers major public financial fallout and reputational damage; China’s state-directed model insulates it from the worst of the crash; its lead in embodied AI delivers real productivity gains independent of AGI. | Competition continues within agreed guardrails; both powers ratify binding restrictions on autonomous weapons and AI-enabled bioweapons; renewed diplomatic collaboration replaces winner-takes-all framing. |
| Global South | Neglected by Western investors; extractive labour practices continue; China fills the gap via the Digital Silk Road, exporting surveillance infrastructure; the climate burden falls disproportionately on developing regions. | Data labelling exploited at scale during the race; upon AGI attainment jobs are automated or reshored to the North, collapsing digital-sector income; climate costs hit hardest despite emissions originating in the North. | Demand for data labellers collapses alongside the industry, displacing workers with no replacement income; China’s infrastructure remains standing, deepening dependency on Beijing. | Reduced AGI fixation frees capital for localised, use-case-specific AI; the region is treated as an investable market rather than an extractable resource; supply chain labour standards enforced. |
| Investor consequence | Short-term returns viable in deregulated markets, but the addressable market shrinks as divergence blocks international expansion; growing unease that investments may not deliver on the priced-in timeline. | Short-term profits substantial, but investors are exposed to the same cybersecurity, instability and societal risks they helped create; regulatory blowback erodes long-term portfolio value. | Tech valuations collapse; widespread economic fallout and unemployment. | Trustworthy, compliant AI wins lucrative enterprise and government contracts globally; investors backing responsible startups access larger, more stable international markets. |
| Societal outcome | Rising geopolitical tensions; no power achieves decisive leadership; global inequality grows; China’s influence expands. | Skyrocketing inequality; collapse of shared information reality; democratic erosion; autonomous weapons proliferation; catastrophic cyberattacks; runaway climate damage; global destabilisation. | Global recession; democratic backsliding via nationalism and populism; entrenched inequality; China emerges as the relative geopolitical winner as Western democracies turn inward. | AI benefits distributed more equitably across regions; democratic institutions reinforced; geopolitical risks reduced via binding agreements; a more collaborative global order emerges. |
Whichever path
Takeaways for Investors
The future direction of AI’s impact is not inevitable. The decision tree remains broad and, at many critical points, is shaped by the decisions of innovators and capital allocators. There are opportunities to invest in AI that stabilises our political environment, and that could usher in a more equitable global order:
The infrastructure required to make AI trustworthy — startups developing data provenance tools, deepfake detection, algorithmic auditing, and cybersecurity solutions.
Some of the world’s most significant under-tapped consumer markets. Fund startups built in and for these populations — neo-banks using voice-based AI to serve populations with low literacy rates, or localised SLMs trained on local language data addressing agricultural, educational and healthcare needs.
The capital and energy requirements of large LLMs — coupled with growing pushback against data centres and recognition of the benefits of fine-tuned, specific models — are creating a lucrative vacuum for cleaner alternatives.
Prioritise companies building augmentative AI tools designed to enhance worker productivity, in education and healthcare for example, rather than pure automation plays that rely on mass layoffs and are likely to trigger intense backlash.
Practical actions to take
- Test for regulatory reach. Assess a company’s ability to comply with the EU AI Act and NIST guidelines. Startups attempting to skirt regulation will likely hit a growth ceiling when enterprise and international buyers demand transparency.
- Build compliance into diligence. Integrate safety, privacy and compliance-by-design frameworks into core due diligence24 and annual portfolio monitoring.
- Audit the data supply chain. Require portfolio companies to map and audit their data supply chains, ensuring fair pay and safe conditions for Global South digital workers.
- Build the shared standards. Partner with philanthropic networks and civil society to share due diligence guides and impact frameworks, and build peer networks that establish industry-wide standards for responsible AI.
References
Sources
All sources below are retained from the original Reframe Venture analysis.
- Venture Capital Journal (2025), Funding for AI dominated in VC in 2025.
- MIT Technology Review (2025), There can be no winners in a US–China AI arms race.
- Carnegie Endowment, Peril and promise in the US–China AI race.
- Foreign Affairs, China and the real artificial intelligence race.
- High Capacity, Does China care about AGI?
- Brookings, Competing AI strategies for the US and China.
- Reuters (2026), China vows to accelerate technological self-reliance in AI push.
- Financial Times, On weaponising the China threat.
- Fortune (2023), Marc Andreessen on AI and competition with China.
- European Commission (2026), Strengthening Europe’s tech sovereignty.
- OMFIF (2024), How the Global South may pay the cost of AI development.
- War on the Rocks, China’s AI governance offensive threatens US tech leadership.
- Center for Data Innovation (2024), AI and the Global South (PDF).
- Financial Times, On China’s willingness to invest in emerging markets.
- ODI, The Paris AI Summit: is geopolitical rivalry derailing AI governance?
- Reframe Venture (2026), Responsible AI: The VC Perspective (PDF).
- BISI, Global fragmentation of AI governance.
- Oxford Business Law Blog (2025), AI regulation: politics, fragmentation and regulatory capture.
- Yale Review of International Studies, The geopolitics of AI regulation.
- European Parliament, EU AI Act: first regulation on artificial intelligence.
- The White House (2025), Eliminating state law obstruction of national artificial intelligence policy.
- Forecasting Research Institute, LEAP report, wave 5.
- Brookings, AI in the Global South: opportunities and challenges towards more inclusive governance.
- Reframe Venture, AI due diligence for VCs.