Overview
This Future Insights review examines five technology signals that have emerged since the previous Americas report in this series. The region is entering a more strategic phase of AI development: governance is becoming more security-centered, enterprise adoption is being reframed as work redesign, ransomware is exposing institutional weakness, semiconductor policy is colliding with AI infrastructure economics, and cloud providers are accelerating capital expenditure at a scale that turns data centers into geopolitical infrastructure. The common thread is that technology advantage in the Americas is no longer only about invention. It increasingly depends on whether states, firms, regulators, utilities, and workers can absorb technological change without creating new fragilities.
For South Africa, the Americas matter because they function as both a technology frontier and an early warning system. U.S. AI and semiconductor choices influence global standards, hardware prices, model access, and cybersecurity doctrine. Latin American experience with digital transformation, institutional gaps, workforce anxiety, and cyber risk is especially relevant because many of its constraints resemble those facing middle-income democracies elsewhere. The signals below therefore should not be read as distant regional news. They indicate how countries outside the main technology blocs may need to position themselves as AI infrastructure, cybersecurity, industrial policy, and workforce transformation become more tightly linked.
Five-signal overview
- The United States issued a new executive order that connects frontier AI development to cybersecurity, critical infrastructure defense, and voluntary pre-release government access to advanced models.
- Latin American business leaders, through Mercer's regional forum, reframed AI adoption as a problem of work redesign, reskilling, and organizational governance rather than software procurement alone.
- Kaspersky's 2026 ransomware analysis identified Latin America as the region with the highest share of organizations facing detected ransomware attacks, highlighting the systemic risk of data-centric extortion.
- ITIF warned that broad U.S. semiconductor tariffs could raise costs, slow AI infrastructure investment, and weaken the very technology leadership that tariffs are meant to protect.
- TrendForce revised 2026 capital expenditure forecasts sharply upward for major cloud service providers, with North American AI data center expansion driving a global infrastructure race.
Signal 1: U.S. AI governance is moving toward security-centered state access
What happened
On 2 June 2026, the White House issued the executive order "Promoting Advanced Artificial Intelligence Innovation and Security." The order directs U.S. agencies to prioritize cyber defense for national security systems, Department of War systems, civilian federal systems, state and local authorities, and critical infrastructure operators. It also calls for an AI cybersecurity clearinghouse, a classified benchmarking process for advanced cyber capabilities, and a voluntary framework through which developers could provide the federal government with access to covered frontier models for up to 30 days before wider release to trusted partners. The order explicitly states that this framework should not create mandatory licensing or preclearance for AI model development or release (The White House, 2026).
Why it matters
This matters because the United States is trying to reconcile two objectives that often pull against each other: keeping AI innovation fast and open enough to preserve private-sector dynamism, while bringing frontier models close enough to the state to evaluate cyber risk before models diffuse through critical systems. The signal is less about ordinary AI regulation than about an emerging security architecture around frontier capability. It suggests that AI governance in the Americas may increasingly be built through public-private arrangements, classified benchmarking, national security channels, and cybersecurity institutions rather than through consumer protection law alone.
What it could mean
For the Americas, the order could shift the center of AI governance from ethical principles toward operational security. If the voluntary model works, the United States may create a de facto international template in which frontier AI firms cooperate with governments through early access, vulnerability discovery, and critical infrastructure hardening. If it fails, either because firms resist, the state lacks assessment capacity, or public trust erodes, pressure could grow for more coercive licensing or fragmented state-level regulation. For South Africa, the immediate implication is that global access to frontier AI may increasingly be mediated by security relationships, trust frameworks, and the ability to participate in cyber governance networks.
Possible futures
Possible future A: Voluntary access becomes a durable model for frontier AI assurance
In this future, the U.S. framework evolves into a credible assurance process without becoming a formal licensing regime. Frontier AI firms share selected pre-release models under strict confidentiality, agencies develop classified and unclassified benchmarking capacity, and critical infrastructure operators receive earlier warning about cyber-relevant model behavior. The strategic logic is that the state does not need to control all AI innovation if it can shape the risk boundary around the most consequential systems. For South Africa, this would create both opportunity and dependency. The opportunity is that international partners may eventually receive safer, better-tested models and clearer cyber guidance. The dependency is that assurance standards would be set elsewhere, likely around U.S. national security assumptions. South Africa would need enough local technical capacity to interpret foreign assurance claims, negotiate access conditions, and avoid mistaking U.S. certification for suitability in South African legal, linguistic, public-sector, and infrastructure contexts.
Possible future B: Security cooperation hardens into a tiered AI access order
In this future, voluntary access becomes the beginning of a more stratified global AI regime. U.S. agencies, trusted firms, and close allies gain early visibility into frontier model risks, while countries outside the trusted circle receive delayed, limited, or commercially filtered access. The formal language may remain cooperative, but the practical effect would be a hierarchy of compute, model access, vulnerability intelligence, and incident response. For South Africa, this would be a serious strategic challenge. The country would have to decide whether to align more closely with U.S.-led AI security frameworks, pursue broader multi-alignment through Europe, China, and open-source ecosystems, or invest in regional assurance capacity with African partners. Each route has trade-offs. Alignment can improve access and credibility, but may constrain policy autonomy. Diversification can preserve options, but may complicate cybersecurity, procurement, and compliance. The second-order risk is that South African institutions could become dependent on tools whose risk models are optimized for other states.
Possible future C: Public trust problems force a harder regulatory turn
In this future, voluntary cooperation proves too opaque for legislators, civil society, or state-level authorities. If a major AI-enabled cyber incident occurs, or if the public comes to believe that frontier model developers receive privileged treatment without accountability, the United States could move toward more formal oversight. The result may be a complex mix of federal rules, state litigation, procurement conditions, and sector-specific obligations. For South Africa, this would signal that AI governance cannot rely only on informal collaboration between government and technology vendors. It would strengthen the case for transparent local rules on public-sector AI procurement, model risk assessment, auditability, and data protection. Yet a harder U.S. regulatory turn could also slow model diffusion and raise compliance costs globally. South African firms that integrate U.S.-origin models would need to track changing obligations, while regulators would need to avoid copying rules whose institutional assumptions exceed domestic enforcement capacity.
Signal 2: Latin America is treating AI adoption as work redesign
What happened
BNamericas published a Mercer press release on 5 June 2026 summarizing Mercer's Latin American Forum 2026. The forum argued that AI adoption in Latin America can no longer be treated only as technology investment, because firms must redesign roles, workflows, skills, talent management, benefits, and employee support if AI is to generate value. Mercer reported that 72% of investors believe firms that integrate human capabilities and AI will gain a competitive advantage, while 98% of global executives expect significant organizational design changes over the next two years. The same release noted that 75% of organizations recognize the need to accelerate digitalization, but only 30% consider their digital agility high (BNamericas, 2026).
Why it matters
This matters because it reframes AI productivity as an organizational absorption problem. The core bottleneck is not simply whether firms can buy tools or access models. It is whether they can redesign processes, reskill employees, manage anxiety, measure capability, and align leadership with human resources and technology functions. Latin America is a useful mirror for South Africa because both contexts include advanced corporate sectors alongside uneven digital skills, institutional constraints, inequality, and workforce vulnerability. The signal suggests that middle-income economies may lose AI value not because the technology is unavailable, but because work systems are too rigid to convert model capability into productive capability.
What it could mean
For the Americas, this could mark a shift from AI experimentation toward labor-market restructuring. Firms that treat AI as a narrow automation tool may see limited gains and growing employee resistance. Firms that redesign work around human judgment, task decomposition, reskilling, and governance may create more durable productivity effects. For South Africa, the lesson is direct: AI strategies that focus only on adoption rates, pilots, or tool access will be incomplete. The harder policy question is how to build organizational capability across firms, government departments, universities, municipalities, and state-owned entities that often operate under severe capacity constraints.
Possible futures
Possible future A: AI becomes a disciplined productivity program rather than a technology fad
In this future, leading Latin American firms use the current AI moment to reorganize work with unusual seriousness. They map tasks, redesign roles, build internal academies, revise performance systems, and treat AI as a capability embedded in operating models rather than as a software layer placed on top of existing dysfunction. The most important effect would be cumulative learning: firms would become better at identifying where AI genuinely changes unit economics, where it merely accelerates low-value activity, and where human judgment remains decisive. For South Africa, this would support a more practical adoption agenda. Business schools, sector bodies, and public agencies could focus less on generic AI awareness and more on process redesign, supervisory capability, and skills measurement. The constraint is distribution. Highly capable firms may advance quickly, while smaller firms and public institutions fall behind. Without deliberate diffusion mechanisms, productivity gains could widen the gap between corporate leaders and the broader economy.
Possible future B: AI adoption increases worker anxiety without delivering proportional gains
In this future, organizations introduce AI tools quickly but redesign work poorly. Employees experience monitoring, role ambiguity, and fears of replacement, while managers struggle to convert experimentation into measurable performance. The result is a paradox: firms appear technologically modern, but trust declines and productivity gains remain modest. This trajectory is plausible because AI adoption often proceeds through fragmented departmental initiatives rather than integrated organizational change. For South Africa, the warning is acute. A labor market already marked by unemployment, inequality, and skills scarcity could absorb AI as another source of insecurity if firms and government do not build credible reskilling pathways. The second-order effect could be political resistance to AI in sectors where employees perceive automation as extraction rather than augmentation. Avoiding this future requires transparent workforce planning, shared productivity gains, and institutions that can distinguish legitimate job redesign from disguised labor shedding.
Possible future C: Human resources becomes a strategic technology function
In this future, the AI transition elevates human resources from an administrative function to a core strategic actor. HR teams gain responsibility for skills intelligence, redeployment pathways, employee wellbeing, organizational design, and the measurement of human-AI performance. This would change the internal politics of firms: technology teams could no longer deploy AI without workforce governance, and executives would have to treat trust, feedback, and capability development as infrastructure. For South Africa, this model could be especially valuable in large employers, banks, telecoms, mining firms, retailers, universities, and public agencies where AI adoption will affect complex workforces. The challenge is that HR functions themselves may lack analytical capacity and strategic authority. If South Africa wants AI to support inclusive productivity, it may need a new class of professionals who combine labor economics, organizational psychology, data governance, and technology implementation. That is a deeper institutional project than buying licenses or running short courses.
Signal 3: Latin America's ransomware exposure is becoming a systemic development risk
What happened
On 12 May 2026, Kaspersky published its International Anti-Ransomware Day analysis of ransomware trends and tactics. Based on Kaspersky Security Network data, the company reported that Latin America had the highest share of organizations with detected ransomware attacks in 2025 at 8.13%, followed by Asia-Pacific, Africa, the Middle East, the Commonwealth of Independent States, and Europe. The report highlighted the rise of encryption-less extortion, the use of Telegram channels and dark web forums for distributing compromised data and credentials, the growth of initial access brokers, and the emergence of ransomware families experimenting with post-quantum cryptography (Kaspersky, 2026).
Why it matters
This matters because ransomware is no longer only an IT incident. It is becoming a development, sovereignty, and trust problem. When attackers focus on stealing and leaking data rather than merely encrypting files, the damage spreads across financial systems, public services, identity infrastructure, healthcare, education, and citizen confidence. Latin America's exposure shows how digital transformation without adequate security investment can convert connectivity into systemic vulnerability. For South Africa, the parallel is important: public institutions, municipalities, hospitals, universities, and firms are digitizing while many still face budget constraints, skills shortages, legacy systems, and uneven cyber hygiene.
What it could mean
For the Americas, the ransomware signal could push cybersecurity from a discretionary technology spend into a core resilience requirement. The region may see stronger reporting rules, more cyber insurance scrutiny, higher demand for managed detection and response, and deeper regional cooperation. But the burden may fall unevenly: large banks and multinationals can invest, while municipalities, small firms, and public agencies remain exposed. For South Africa, the implication is that AI, cloud, and digital public services cannot be separated from cyber resilience. Digital inclusion that lacks security may generate new channels for harm.
Possible futures
Possible future A: Cybersecurity becomes a condition for credible digital development
In this future, Latin America's ransomware exposure forces governments, banks, telecoms, hospitals, ports, and utilities to treat cybersecurity as critical infrastructure. Investment shifts from reactive incident response toward identity security, backup architecture, threat intelligence, vulnerability management, and regional information sharing. The strategic benefit would be that digital transformation becomes more credible: citizens and firms can trust online services because resilience is built into them. For South Africa, this future would strengthen the case for embedding cyber requirements into public procurement, municipal modernization, health digitization, smart infrastructure, and financial inclusion programs. The challenge is affordability. Security capability is expensive and scarce. If requirements become too sophisticated for smaller institutions, compliance may become performative. South Africa would need tiered standards, shared services, and public-interest cyber capacity that protects weaker institutions rather than merely auditing their failure.
Possible future B: Data-centric extortion undermines trust in digital identity and public platforms
In this future, ransomware groups focus less on business interruption and more on the reputational, regulatory, and political leverage created by stolen data. Identity records, health information, payment credentials, school systems, and government files become bargaining chips. The damage is not only financial; it changes how citizens behave. People may resist digital identity systems, avoid online services, or distrust public platforms if they believe the state cannot protect sensitive data. For South Africa, this is a crucial warning because digital public infrastructure is often promoted as a route to inclusion and administrative efficiency. If security is weak, the same infrastructure can become a concentration of risk. The trade-off is subtle: centralized systems can improve service delivery and reduce fraud, but they also create high-value targets. South Africa's digital identity and public-service reforms therefore need privacy-by-design, segmented architecture, incident transparency, and political accountability before large-scale integration proceeds too far.
Possible future C: Ransomware professionalization creates a persistent tax on middle-income economies
In this future, initial access brokers, data leak markets, and automated intrusion tools turn ransomware into a recurring economic tax on digitally modernizing economies. Attackers do not need to defeat the strongest institutions; they can exploit weaker suppliers, local governments, schools, clinics, and small firms that connect into broader systems. The result is a drag on productivity and trust: budgets are diverted to recovery, insurance becomes expensive, and digital projects slow because leaders fear liability. For South Africa, the lesson is that cyber resilience cannot be left to each institution individually. The country may need sector-level security operations, shared threat intelligence, national incident exercises, and support for small suppliers in critical value chains. Otherwise, the weakest nodes will define the risk of the whole ecosystem. This future also implies that cybersecurity skills policy is economic policy. Without enough trained analysts, engineers, auditors, and incident responders, South Africa's AI and cloud ambitions will rest on brittle foundations.
Signal 4: U.S. semiconductor tariffs could collide with AI infrastructure economics
What happened
On 4 June 2026, the Information Technology and Innovation Foundation published an analysis arguing that broad U.S. Section 232 semiconductor tariffs could weaken U.S. growth and technological leadership. ITIF noted that the Trump administration began implementing new semiconductor, semiconductor machinery, and downstream technology tariffs in January 2026, initially imposing a 25% tariff on a narrow subset of semiconductors. The analysis argued that a blanket 25% semiconductor tariff could reduce U.S. GDP growth by more than $58 billion in the first year and by $1.6 trillion over ten years, raise consumer prices, and increase costs for AI data center construction and operation at a time when semiconductors are foundational to AI infrastructure (Ostertag & Long, 2026).
Why it matters
This matters because AI competitiveness depends on supply chains as much as algorithms. Semiconductors are embedded in cloud infrastructure, edge devices, vehicles, defense systems, industrial automation, and consumer electronics. Tariffs intended to strengthen national security may raise the cost of the compute base that supports national security, productivity, and innovation. The signal reveals a broader policy dilemma in the Americas: reshoring and strategic autonomy are politically attractive, but blunt cost-raising instruments can damage the ecosystems they are meant to protect. For South Africa, chip prices and AI infrastructure costs are not remote concerns; they shape local cloud pricing, device affordability, industrial modernization, and public-sector digitization.
What it could mean
For the Americas, semiconductor policy may become a test of whether industrial strategy can distinguish between capacity building and cost inflation. Targeted subsidies, workforce development, allied supply chains, and permitting reform may strengthen resilience. Broad tariffs may instead redistribute costs across firms and consumers before domestic capacity is ready. For South Africa, the key implication is vulnerability to policy choices made elsewhere. As a semiconductor importer with limited advanced manufacturing capacity, South Africa is exposed to global pricing, export controls, tariff pass-through, and cloud infrastructure decisions.
Possible futures
Possible future A: Tariff policy is narrowed and industrial strategy becomes more targeted
In this future, U.S. policymakers respond to economic criticism by limiting broad tariffs and focusing on instruments that expand capacity without sharply raising input costs. These could include fab incentives, workforce pipelines, advanced packaging support, allied supply-chain agreements, and tariff offsets for firms investing in AI infrastructure. The strategic logic would be to strengthen domestic production while preserving the affordability of downstream innovation. For South Africa, this would be the least disruptive path. Cloud, AI, and device costs would still reflect global demand, but they would be less distorted by broad tariff escalation. It would also provide a useful policy lesson: industrial strategy works best when it identifies bottlenecks precisely rather than treating all imports as weakness. South Africa could apply that lesson to its own digital industrial policy by supporting niche capabilities, testing facilities, energy resilience, and procurement demand without pretending it can localize entire semiconductor supply chains.
Possible future B: Cost escalation slows AI diffusion outside frontier firms
In this future, semiconductor tariffs and component shortages raise the cost of AI infrastructure enough that the largest firms continue investing, but smaller firms, public institutions, and emerging-market users face higher prices. The result would be a more concentrated AI economy. Hyperscalers with balance-sheet strength absorb costs and pass them through gradually, while smaller cloud providers, startups, universities, and governments delay projects or rely more heavily on externally hosted services. For South Africa, this is a serious risk. Higher compute prices could widen the gap between large corporations that can buy AI capability and public-interest institutions that need AI for education, health, language access, agriculture, and public administration. The second-order effect would be dependency: if local actors cannot afford experimentation, they become consumers of finished systems rather than participants in adaptation. South Africa would need compute-sharing models, public research infrastructure, and regional procurement strategies to preserve agency under cost pressure.
Possible future C: Semiconductor nationalism fragments technology supply chains
In this future, tariffs become part of a broader fragmentation of semiconductor and AI supply chains. The United States, China, Europe, and selected allies develop overlapping but increasingly separate regimes for chips, cloud, export controls, standards, and security assurances. Firms operating across the Americas would face more complex compliance, while countries outside the main blocs would struggle to maintain access to affordable, interoperable technologies. For South Africa, this would make technology diplomacy more consequential. The country would need to balance relationships with multiple suppliers while protecting its own security, competition policy, and industrial ambitions. Fragmentation could create opportunities for non-aligned markets to attract assembly, testing, data-center, or services investment, but only if they offer reliable power, legal certainty, skills, and cyber resilience. The risk is that South Africa becomes a price-taker in a divided system, forced to accept higher costs and lower bargaining power because it lacks scale and coordinated regional strategy.
Signal 5: North American AI data center spending is becoming a global infrastructure race
What happened
On 6 May 2026, TrendForce reported that strong AI demand had led several major North American cloud service providers to raise 2026 capital expenditure guidance. TrendForce revised its forecast for combined 2026 capital expenditure by the world's top nine cloud service providers to approximately US$830 billion, with annual growth raised from 61% to 79%. The report said Microsoft, Google, Meta, and AWS had increased or were expected to exceed large capex outlooks, and that investment is concentrating in high-performance GPU clusters, in-house ASIC development, and next-generation data centers for high-power-density computing. TrendForce also projected total installed data center power capacity at about 155 GW in 2026, up roughly 29% year on year, and expected AI servers to surpass general-purpose servers in electricity consumption during 2026 (TrendForce, 2026).
Why it matters
This matters because AI infrastructure is becoming one of the defining capital allocation stories of the decade. Data centers are no longer neutral background facilities. They are strategic assets that combine compute, chips, energy, cooling, land, fiber, water, finance, and geopolitical trust. North American spending at this scale can accelerate model capability and cloud service availability, but it can also intensify energy constraints, supplier bottlenecks, and market concentration. For South Africa, the signal is important because local AI adoption will depend on infrastructure economics largely shaped by hyperscalers and global component markets.
What it could mean
For the Americas, the infrastructure race could reinforce U.S. cloud dominance while creating new opportunities for Canada, Brazil, Mexico, and other markets that can offer energy, land, regulatory stability, and demand. Yet it may also produce backlash if communities see data centers competing for electricity or water without clear local benefits. For South Africa, the question is how to participate in AI infrastructure value chains without overextending scarce energy resources or becoming entirely dependent on offshore compute. Local and regional data-center strategy will need to be tied to grid planning, renewable energy, data governance, skills, and public-interest compute access.
Possible futures
Possible future A: Hyperscale investment lowers AI access costs but deepens platform dependence
In this future, the massive capex cycle creates enough compute capacity to reduce some bottlenecks and make advanced AI services more widely available through cloud platforms. South African firms, universities, and public agencies benefit from improved tools, better latency in some regions, and faster access to managed AI services. The strategic downside is that dependence on a small number of hyperscalers deepens. Pricing, model availability, data residency options, and compliance features are set primarily by firms whose infrastructure priorities are shaped by North American demand and shareholder expectations. For South Africa, the policy challenge would be to use hyperscale platforms pragmatically while preserving bargaining power. That may require multi-cloud procurement, open standards, local skills capable of migration and audit, and public-sector rules that prevent critical services from becoming locked into opaque vendor architectures. The benefit is access; the risk is structural dependency.
Possible future B: Energy and cooling constraints reshape the geography of compute
In this future, power capacity, grid interconnection, water availability, and cooling technology become as important as chip supply in determining where AI infrastructure is built. North American markets with constrained grids face delays or rising costs, while regions with renewable energy, cold climates, or coordinated permitting attract new investment. This could create opportunities for parts of the Americas, but it also offers a lesson for South Africa. Data centers cannot be treated as ordinary real estate projects. In a power-constrained economy, every major compute facility must be assessed against grid stability, local economic benefit, water stress, emissions, and opportunity cost. South Africa could still host valuable AI infrastructure, especially if tied to renewable generation and regional demand, but the country should avoid a race to attract energy-intensive facilities without clear developmental returns. Compute strategy has to be integrated with energy strategy, not added after the fact.
Possible future C: AI infrastructure concentration triggers a public-interest compute agenda
In this future, governments, universities, and civil society react to hyperscale concentration by building public-interest compute arrangements. These may include national research clouds, subsidized compute for startups and universities, sovereign data spaces, regional AI infrastructure funds, and procurement rules that reserve capacity for public goods. The Americas could see this pressure first in debates over whether AI infrastructure serves broad productivity or mainly strengthens already dominant platforms. For South Africa, this trajectory would be highly relevant. The country needs AI capacity for local languages, education, public health, climate adaptation, mining safety, agriculture, and state capability, but many of those uses may not be commercially prioritized by hyperscalers. A public-interest compute agenda would not mean rejecting private cloud. It would mean deliberately securing capacity, datasets, and skills for socially valuable uses that markets underprovide. The difficulty is governance: public compute can become wasteful if politicized, but transformative if anchored in transparent demand and technical excellence.
Conclusion
The Americas are showing how the AI era is becoming an institutional stress test. The United States is pulling frontier AI closer to national security and cyber defense. Latin American firms are discovering that AI value depends on redesigning work, not merely acquiring tools. Ransomware is exposing the fragility of digital transformation when security investment lags. Semiconductor tariffs reveal the tension between strategic autonomy and the cost structure of innovation. North American data-center spending shows that AI is now built from physical infrastructure, not only code.
For South Africa, the central lesson is that technology strategy must be systemic. AI policy, cybersecurity, skills, energy, data governance, cloud procurement, and industrial policy cannot be managed as separate files. The Americas offer both models and warnings: security frameworks can improve resilience but create access hierarchies; AI adoption can boost productivity but intensify worker anxiety; infrastructure investment can widen access but deepen platform dependence. South Africa's task is to learn from these signals early enough to build institutions that can absorb technology on its own terms.
References
BNamericas. (2026, June 5). Latin America must redesign work to turn AI into a competitive advantage, according to Mercer.
Kaspersky. (2026, May 12). International Anti-Ransomware Day-2026: Kaspersky shares insights into ransomware trends and tactics.
Ostertag, M., & Long, T. (2026, June 4). Section 232 semiconductor tariffs could undermine US economic growth. Information Technology and Innovation Foundation.
The White House. (2026, June 2). Promoting Advanced Artificial Intelligence Innovation and Security.
TrendForce. (2026, May 6). North American AI data center expansion drives 2026 CapEx of top nine CSPs to US$830 billion.
Publication links (website version)
- BNamericas – Latin America must redesign work to turn AI into a competitive advantage, according to Mercer
- Kaspersky – International Anti-Ransomware Day-2026: Kaspersky shares insights into ransomware trends and tactics
- ITIF – Section 232 semiconductor tariffs could undermine US economic growth
- The White House – Promoting Advanced Artificial Intelligence Innovation and Security
- TrendForce – North American AI data center expansion drives 2026 CapEx of top nine CSPs to US$830 billion