Something happened this week at the top layer of the stack: the governance layer. On July 16, twenty-nine countries established the World AI Cooperation Organization (WAICO) in Shanghai. Three weeks earlier, the number of signatories to the Pax Silica Declaration in Washington had risen to twenty-four. Four of the five regions we track are now part of at least one of these two architectures, and at least one state belongs to both.

The Gulf

Abu Dhabi capital took a position in global computing infrastructure. On July 21, MGX and BlackRock’s Global Infrastructure Partners completed their acquisition of Aligned Data Centers and committed an additional $5 billion to support expansion. Aligned’s portfolio includes 51 data center campuses and approximately 6.4 GW of operational and planned capacity.

The central issue here is not where the assets are located, but whose balance sheet they sit on. Most of the campuses are outside the Gulf. This is therefore not an attempt to build sovereign computing capacity. It is an attempt to own parts of the stack located on someone else’s territory. This is precisely where sovereignty and ownership diverge.

Oman became a founding member of WAICO, the only GCC country among the twenty-nine signatories. Qatar and the UAE, meanwhile, are Pax Silica signatories. The Gulf is divided between two governance architectures, and that division runs through a single council.

Central Asia and the Caspian

All five Central Asian states moved toward Beijing this week. What differed was the channel through which they did so. Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan joined WAICO as founding members. Turkmenistan did not send a delegation to Shanghai. Instead, its Ministry of Communications signed a memorandum of understanding with China’s National Development and Reform Commission that includes cooperation on artificial intelligence. The broader package also covered energy, science and technology, education, health, culture, transportation, and logistics, as well as a 2026-2028 scientific and innovation partnership program and a five-year cooperation plan for 2026-2030.

The distinction between a multilateral signature and a bilateral memorandum is not trivial. One constitutes institutional membership; the other provides a ministerial-level framework. The direction, however, is the same.

Kazakhstan belongs to both architectures. Astana was one of the ten partners that signed the Pax Silica Declaration in late June, and on July 16 it became a founding member of WAICO. In Shanghai, Tokayev proposed that Astana host WAICO’s first annual meeting and that the organization establish its Central Asian office in the capital.

During the same visit, more than seventy documents were signed, representing a portfolio valued at over $15 billion. The items directly related to artificial intelligence included a strategic partnership between the Ministry of Artificial Intelligence and Digital Development and Huawei; an agreement on fundamental principles between Kazakhtelecom and HV & Submarine Hengtong Group for the “Data Center Valley” in Ekibastuz; an AI and robotics agreement among the Almaty municipal government, NERO Group, and UBTECH Robotics; and the opening of the IBI Digital Economy Center in Almaty. The package also included a Roadmap for Trade and Economic Cooperation for 2027–2030.

A qualification is necessary: the $15 billion figure represents the total value of a package of contracts and memoranda, not investment that has already been committed or transferred. The Hengtong project, meanwhile, remains at the stage of agreeing on fundamental principles.

Uzbekistan began construction of the $500 million Future Technologies Center in Nurafshan on July 22–23. The center will focus on artificial intelligence, cybersecurity, and digital technologies and is expected to create more than 5,000 skilled jobs. The source is the Presidential Press Service.

As background, and treated separately because it falls outside the reporting window, Mirziyoyev announced an incentive package at TIIF-2026 in mid-June. AI and data center projects in Karakalpakstan’s special zone will receive tax exemptions until 2040, the state will finance essential infrastructure such as roads and water, and electricity will be provided at a subsidized rate of five cents per kilowatt-hour. The stack debate rarely produces a figure this explicit.

Azerbaijan emphasized its role as a western gateway. At the Shusha Global Media Forum on July 13, Aliyev highlighted the submarine fiber-optic cable being developed between Sumgait and Aktau and explicitly linked it to artificial intelligence, data centers, digital services, and data transmission between Europe and Asia. According to Aliyev, the cable will be approximately 380 kilometers long and offer capacity of up to 400 Tbit/s. A feasibility study for an energy cable across the Caspian seabed is also being prepared.

Once completed, these two lines would create a route for Central Asian electricity and data that does not pass through Russia or China. The fiber project is still under development, however, while the energy line remains at the feasibility-study stage. Neither is infrastructure yet; both are promises of infrastructure.

A memorandum of understanding was signed between IDDA and Polygraf AI to establish a regional center of excellence in Baku. The center will operate within Polygraf Labs, a wholly owned subsidiary of the company. According to the company, its mandate is not to resell or transfer American products, but to develop new security capabilities suited to the region’s operational, regulatory, linguistic, and market realities. The company plans to employ more than 200 people in the region, although this is a corporate projection rather than a binding commitment. The initiative is described as Polygraf AI’s second sovereign AI operation after Japan. It aligns with Azerbaijan’s 2025-2028 Artificial Intelligence Strategy and its 2026-2028 action plan for accelerating digital development. The parties also plan to support the expansion of Azerbaijani digital solutions into African markets.

Africa

On July 21, the IMF directly linked the continent’s potential AI gains to electricity and connectivity. According to its assessment, artificial intelligence could increase output in sub-Saharan Africa by approximately 4 percent over the next decade, but only if countries substantially improve electricity access, internet infrastructure, and digital skills. Without these improvements, the gain could fall to as little as 0.2 percent. In 2024, only about 38 percent of Africans used the internet, while roughly half of sub-Saharan Africa lacked reliable access to electricity.

A twentyfold difference. And the variable determining that difference is not in the model, but in the grid.

South Africa and Senegal are among WAICO’s founding members. Cameroon was also among the signatories. The continent secured an unusually strong level of representation in the new governance institution.

South Africa is simultaneously positioned within two ecosystems. Beginning on July 15, Minister Nzimande made an official visit to China, where discussions focused on expanding cooperation in artificial intelligence, science, technology, and innovation. At the same time, Pretoria continues to develop cloud and AI relationships with Western companies. This is a choice not to choose.

The continent’s position this week can therefore be summarized as follows: it gained a voice at the governance layer, while its constraints at the physical layer remained in place.

The Western Balkans

Serbia became WAICO’s only Balkan founding member. Alongside Russia and Belarus, it is also one of the organization’s three European founding members.

The rest of the region is either embedded in other architectures or in none of them. Greece joined Pax Silica in June. Croatia, Slovenia, Bulgaria, Romania, and Greece are bound to the EU Artificial Intelligence Act and the EuroHPC system. No comparable announcement of institutional alignment was identified during this reporting window for Albania, Bosnia and Herzegovina, Kosovo, Montenegro, or North Macedonia.

Unlike Central Asia, the Balkans did not move as a bloc. Serbia acted alone.

Türkiye

The foundation for the TÜRKSAT Gölbaşı Data Center was laid on July 22. The project, launched at a ceremony attended by Minister Uraloğlu, will cover 35,300 square meters and become one of Türkiye’s largest data centers. Plans include six system halls, two high-performance rooms containing twenty cabinets each, 6,300 square meters of white space, and office facilities for 200 people. The installed power capacity will be 33 MVA. TÜRKSAT’s existing data center capacity will increase threefold in the first phase and by more than eightfold in the second. The project targets LEED Gold certification for energy efficiency and compliance with TIER III and TSE 50600 data center standards.

The center is designed to meet the data and cloud-service needs of public institutions, particularly the e-Government Gateway, using domestically developed software. According to Uraloğlu, it will also provide infrastructure for services such as artificial intelligence, deep learning, and data mining.

The phased capacity plan is noteworthy: first a threefold increase, followed by an eightfold increase. It is an approach that explicitly distinguishes between the final capacity being announced and the capacity that will initially enter service.

Two governance architectures took shape simultaneously this week, and the regions we follow positioned themselves within them. One qualification remains necessary: WAICO is a signature, and Pax Silica is a declaration. Neither yet generates a single megawatt, lays a kilometer of fiber, or operates a GPU.

The IMF’s figures for Africa demonstrate this most clearly. What determines the difference between 4 percent and 0.2 percent is not which organization a country joins, but whether the electricity arrives.

A state can enter two governance architectures in the same week, but it remains connected to a single grid.

Subscribe now

This week, the lowest layer of the stack, who can access which chips, once again determined the highest layer. Washington moved the UAE into the most trusted tier of its export-control system, making the question of who can access sovereign AI compute capacity, and under what conditions, no longer merely commercial but directly geopolitical. During the same week, the UN Digital Week in Geneva brought actors from five regions to the same table, but each arrived with a different move. Kazakhstan positioned itself as a hub, Uzbekistan presented its strategy, Qatar launched an ethics alliance, Egypt signed an agreement for a regulatory sandbox, and Azerbaijan joined the cybersecurity agenda. This simultaneous activity at the governance layer shows that the infrastructure race has now become intertwined with the race to shape the rules. The sequence remains unchanged: first energy, then connectivity, then compute. What mattered this week, however, was the sovereignty framework within which that sequence was constructed.

Gulf / Middle East: The Trusted Partner Tier

The most consequential move of the week was not an agreement, but a reclassification. Through a final rule that entered into force on 10 July, the Bureau of Industry and Security, under the US Department of Commerce, removed the UAE from the restrictive D:3 and D:4 country groups and transferred it to Group A:5, the most privileged tier under the Export Administration Regulations, which includes NATO members and close allies. The UAE is the first Arab country to receive this status, and the fact that it is the only country in the group that is not a member of the multilateral export-control regimes makes the classification particularly distinctive.

The practical implication is as follows: approved UAE entities can now acquire advanced AI chips and servers under the Strategic Trade Authorization exception without applying for individual export licences. This is an extension of the US-UAE Artificial Intelligence Cooperation Framework signed in May 2025. The first two automatically approved entities are G42 and Core42, while MGX applications will receive priority review. UAE subsidiaries of OpenAI, Microsoft, Google, Oracle, Amazon, Apple, Meta and xAI are also covered.

It is important to note that the framework is not a one-sided concession. The UAE reaffirmed its commitment to make matching investments in the construction of US AI infrastructure. In other words, this is a two-way equation. The most striking provision of the rule from a sovereignty perspective is that G42 and Core42 must reapply to retain their approved status if they do not “become US companies” within 270 days of 10 July. The price of access is institutional affiliation itself.

BIS officially justified the decision on the basis of the UAE’s regional security partnership and its commitment to preventing the diversion of sensitive technologies. UAE Minister of State Saeed Al Hajeri described the decision as a “historic achievement” and as recognition of the country’s export-control framework. Some analysts, meanwhile, have linked the decision to Emirati lobbying in Washington and to the country’s regional security support. Whichever interpretation is adopted, the mechanism is clear: access to compute capacity is now a matter of trust classification, and entry into that tier is determined not by energy or capital, but by alignment.

During the same week, three additional signals emerged from different layers of the Gulf stack. In Riyadh, HUMAIN announced a strategic partnership with the Canadian sovereign AI company Cohere. HUMAIN will allocate at least 50 MW of AI compute capacity to support Cohere’s next-generation foundation models. This represents Cohere’s first major compute deployment outside North America and is intended to support the development of sovereign models for the Saudi market, with data processed inside the Kingdom. The infrastructure is expected to become operational in the fourth quarter of 2027, meaning that this is a future commitment, not existing capacity.

The agreement coincided with Canadian Prime Minister Mark Carney’s visit to Jeddah. During the visit, the two countries signed 13 commercial agreements and memoranda worth more than CAD 1 billion as part of a diversification package centred primarily on mining. These included a memorandum on AI Investment and Skills Development. This memorandum is not itself a capital or compute agreement, but rather a cooperation framework focused on joint investment opportunities and skills development. Read together with the HUMAIN-Cohere agreement, however, the two developments constitute parallel Canada-Saudi AI stories from the same week: one focused on institutional compute, the other on a state-level framework. Their common ground is a search for alternatives beyond US intermediation.

At the application layer, meanwhile, Egypt’s Ministry of Economy introduced the beta version of an agentic AI tool called INSAIGHTS, integrated into the Data Saudi platform. The tool scans more than 7,500 economic and social indicators and converts user queries into real-time analysis. It should be noted that the tool remains in beta. Nevertheless, the use of agentic AI to open up state data demonstrates that the Gulf’s activity this week was not limited to compute and institutional frameworks, but also extended to the public-service layer.

Central Asia: A Showcase Built in Geneva

This week, the region’s centre of gravity shifted towards multilateral diplomacy. The UN Digital Week held in Geneva from 6 to 10 July, encompassing the UN Global Dialogue on AI Governance, the AI for Good Summit and the WSIS Forum, provided Kazakhstan and Uzbekistan with a shared stage.

Kazakhstan opened a national pavilion at Palexpo under the slogan “Kazakhstan: Eurasia’s AI and Digital Hub,” led by Deputy Prime Minister and Minister of Artificial Intelligence and Digital Development Zhaslan Madiyev. The pavilion was divided into three thematic sections: Innovation Cluster, GovTech and Data Center Valley. Madiyev also became one of the founding members of the AI for Good Global Commission established under the ITU and co-chaired by Rwandan President Paul Kagame and Salesforce CEO Marc Benioff, a body bringing together more than 40 state and corporate leaders. This was part of Kazakhstan’s effort to present its narrative of moving from an “e-government model to an AI-native nation” on a global platform. Madiyev also met with António Guterres.

At the same Geneva moment, Uzbekistan’s Minister of Digital Technologies, Sherzod Shermatov, presented approximately 100 priority AI projects being implemented across healthcare, education, agriculture, energy, finance, industry and e-government. The fact that the two countries appeared at the same table but arrived with different moves is significant: Kazakhstan is marketing itself as a hub and securing a commission seat, while Uzbekistan is showcasing its strategy. Their circumstantial similarity does not erase their structural differences.

Outside Geneva, the region’s infrastructure and sovereignty tensions also produced tangible examples this week. On 3 July, Kazakhstan signed a trilateral memorandum with Citi and Freedom Holding Corp announcing its intention to establish a GPU-based data centre called the “Freedom AI Super Cluster.” Positioned as a “compute centre” for the public sector, business and science, the project was announced in the context of a Foreign Investors’ Council meeting chaired by President Tokayev. It does not replace the Data Center Valley in Ekibastuz, but supplements it. In other words, it is a separate compute-sovereignty agreement. Under another memorandum, Kazakhstan’s Ministry of Transport signed a preliminary agreement, or term sheet, with the UAE-based AI company Presight to explore the development of an AI-powered transport system. This represents the transport-layer expression of the same hedging line extending from the Gulf into Central Asia.

In Uzbekistan, the tension between sovereign and imported stacks became visible within a single event. On 1 July, the Chinese AI giant iFLYTEK formally launched its Central Asia Open Platform in Tashkent, bringing its Spark BDM stack into the region. Shermatov delivered the opening address. At the same event, Kazakhstan’s ISSAI institute presented its open-source sovereign models developed for low-resource languages, including Qolda-AVL, the first open-source trimodal model for Kazakh combining speech, vision and language. The imported application layer and the sovereign-model effort stood side by side in the same room. In addition, an AI cluster equipped with a GPU cluster and specialised laboratories was inaugurated at the Tashkent University of Information Technologies, marking a concrete step towards building domestic capacity at the talent and compute layers.

Africa: A Continent at the Governance Layer

Africa’s story this week came less from infrastructure and more from governance and was once again largely constructed in Geneva.

The most notable development came from Nigeria. The country ranked first in Africa in the 2026 Global Index on Responsible AI, published by the Cape Town-based independent think tank Global Center on AI Governance. With a score of 45.93, Nigeria placed 38th globally, rising 42 positions from 80th place in 2024, when it scored 7.21. The index is one of the first layered assessments covering 135 countries, while Africa’s regional average stands at only around 22. Nigeria’s rise therefore represents an exceptional performance against the continent’s generally low baseline. Minister of Communications, Innovation and Digital Economy Bosun Tijani described the result as confirmation of the country’s strategy to become “the voice shaping AI governance in the Global South,” while the ranking itself is independently produced. Egypt ranked second in Africa.

Egypt’s principal moves this week, however, came through two separate agreements signed at WSIS Geneva. The first was an agreement between the government and UNDP intended to strengthen responsible AI governance, institutional capacity and the establishment of the country’s future AI Sandbox, a supervised environment in which AI systems can be tested before being deployed at scale. It should be emphasised that the sandbox is yet to be established and is not an existing facility. The second was a memorandum signed by Egypt and Rwanda on responsible AI development, digital public services, innovation and digital skills, formalising discussions held between the two ministers several days earlier. Both countries framed the agreement in terms of “an African approach to AI” and digital sovereignty. Rwanda is establishing a National AI Agency, while Egypt is implementing the second phase of its national AI strategy.

Across the continent, a common tension is becoming visible: policy commitments are multiplying rapidly, while implementation and physical infrastructure continue to lag behind. GIRAI’s global finding that most frameworks are non-binding and that evidence of implementation falls to 45 per cent in the Global South also points to the central challenge facing the governance agreements signed by African states this week. While the energy and compute layers remain weak, claims to governance leadership must progress in parallel with the underlying layers if they are to be sustainable.

Western Balkans: The Regulatory Vacuum

This week, the region’s signal came not from an investment but from an absence. A region-wide study published on 2 July found that no Western Balkan country has enacted legislation regulating the use of AI, and that most institutions lack clear rules on whether public officials may use AI tools in their work.

Serbia provides the clearest example of the gap between appearance and safeguards. The country has deployed AI chatbots in its Tax Administration, Business Registers Agency and e-Uprava portal, yet still lacks legislation governing the development and use of AI. The existing framework consists only of the 2025–2030 strategy and ethical guidelines. The study states that the Tax Administration’s virtual assistant alone cost approximately EUR 196,000 in 2026. This figure is based on a single-source public procurement analysis and should therefore be treated as an attribution to that source pending independent verification. The region has not yet closed the gap between the speed with which it is introducing AI into public services and the speed with which it is developing the legal frameworks needed to govern that use.

The structural gap remains unchanged: the EU AI Act was designed for member states with functioning enforcement bodies, while no consistent pre-accession alignment model exists for candidate countries. This coordination and capacity deficit continues to be the principal constraint on AI governance in the region.

Türkiye: The Plan Moves into Implementation

Two concrete and distinct developments emerged from Türkiye this week, each corresponding to a different layer of the national AI strategy.

The first was the beginning of the implementation phase of the Türkiye Artificial Intelligence Action Plan announced by President Erdoğan last month. Under a process coordinated by the Ministry of Industry and Technology and TÜBİTAK, the 2026 TÜBİTAK AI Ecosystem Call was opened. Consortium-based projects in areas including smart manufacturing, agriculture, financial technologies, e-commerce, climate change and education will be eligible for grants of up to TRY 10 million per project, approximately USD 250,000. Applications will remain open until 18 September. The HIT-30 High Technology Investment Programme allocates USD 1.6 billion to AI projects. The plan also envisages cooperation on robotics and autonomous systems with Japan and ten European countries under the CONCERT Japan framework, an entrepreneur-investor hub called Terminal Istanbul, regulatory sandboxes in at least five sectors, a national AI growth fund and a national AI council.

The plan’s most visible target is to mobilise at least USD 10 billion, primarily from the private sector, for data centres, cloud infrastructure and AI capacity. It should be emphasised, however, that this is a mobilisation target, not committed capital. Similarly, the government’s expectation that the plan will generate more than TRY 1 trillion in added value is a projection rather than a confirmed outcome. The concrete and time-specific elements are the value of the TÜBİTAK grants and the 18 September application deadline.

The second development was the NATO Summit held in Ankara on 7–8 July. The summit communiqué used restrained language on AI: the allies are “developing an interoperable transatlantic battle cloud and adopting powerful AI models.” This was a single sentence within a broader capabilities paragraph. The communiqué also included more than USD 50 billion in new procurement, EUR 70 billion for Ukraine in 2026, more than USD 40 billion in counter-drone capacity investments over five years, and a new NATO-Industry Cooperation Strategy. The NATO Defence Industry Forum held on 7 July, when read together with HAVELSAN’s first NATO export involving the ADVENT Combat Management System, complements Türkiye’s position at the defence-software layer. The fact that the summit was hosted in Ankara and coincided with the visibility of Turkish defence technology made this week the clearest demonstration of the shift in value from platforms towards data and software architectures.

The stack is always constructed in the same order: first energy, then connectivity, then compute. But across all five regions this week, the determining factor was not the order itself, but the trust tier and sovereignty framework within which that order was established.

Subscribe now

This week, one pattern stands out across five regions: value in the stack is increasingly established at its lowest layer, energy, while the sovereignty framework governing the upper layers, namely operating systems, platforms, and software, is becoming decisive. Kazakhstan is positioning the same data-centre asset toward both Washington and Beijing; Uzbekistan is giving concrete form to a plan to export electricity as data-centre services; and Türkiye is moving beyond the role of a platform exporter toward a position as a contributor to NATO-compatible combat-data and command architecture. The sequence is the same everywhere: first energy, then connectivity, then compute.

Central Asia: The two-way hedge

The central story of the week is not a choice, but a hedge. Kazakhstan moved closer to the US-led Pax Silica framework during the Washington meetings in late June, extending its engagement with US-led supply-chain and critical-minerals initiatives. The framework covers critical minerals, semiconductors, energy, data centres, and workforce development. It highlights Kazakhstan’s mineral reserves and its position along the Middle Corridor, adding a diplomatic and institutional layer to the country’s broader engagement with US technology and infrastructure partners.

Then, on July 3, the same country presented the same asset, Data Center Valley, to a Chinese audience in Beijing at the Digital Central Asia forum, held within the Global Digital Economy Conference. The director of the CAREC Institute called for joint China-Central Asia investment in green data centres, cloud infrastructure, and high-performance computing while preserving digital sovereignty. A single computing asset presented to Washington’s supply-chain framework and Beijing’s digital-economy forum within one week captures the week’s hedge.

Astana is underpinning this with an institutional base. At the 38th plenary session of the Foreign Investors’ Council, artificial intelligence was positioned as the central axis: Data Center Valley, Central Asia’s first two supercomputers, and the region’s first university dedicated to artificial intelligence. The figures are substantial: 6.5% growth in 2025, GDP exceeding $300 billion, and cumulative net foreign direct investment surpassing $150 billion. Kazakhstan’s AI Law is already in force, while the recently adopted Digital Code is scheduled to enter into force on July 11, 2026. Tokayev’s statement that Data Center Valley has attracted the interest of Amazon and G42 should be read as the president’s own characterization, rather than as a confirmed investment commitment.

At the governance layer, Kazakhstan is also tying its own legislative body to a platform architecture it controls: the e-Parliament platform is being adapted to the new unicameral Kurultai structure; more than two million citizen requests have been processed through the e-Otinish AI platform; and since the beginning of the year, a moratorium has been applied on building government systems outside QazTech. The Kurultai elections on August 23 will provide the political basis for this transition.

Uzbekistan is turning the same logic into an export model. Its target is $5 billion in IT and AI service exports by 2030. According to Minister Shermatov, IT exports have risen from less than $1 million in 2017 to around $1 billion today, while IT Park’s first-quarter 2026 exports reached $191.8 million. Shermatov’s formula is the ministerial version of the closing aphorism: the country can sell its electricity in the form of AI data-centre services. Cheap energy, IT Park tax exemptions, and import-duty exceptions form the incentive side of this equation.

Azerbaijan is building its institutional and data-infrastructure layer through central coordination. The Digital Development Council, established by presidential decree in February 2026, held its first meeting on June 29 under the chairmanship of First Vice President Mehriban Aliyeva. Its 2026 to 2028 Action Plan includes 58 initiatives across digitalization, artificial intelligence, the innovation ecosystem, and cybersecurity. Roadmaps have been prepared with 40 state institutions, 13 of which have been approved or are in the approval process. A National Data Center is being established, government information systems are being migrated to the government cloud, G-Cloud, and the user base of the mygov platform rose from 2.7 million to 3.5 million in three months.

An Azerbaijan Resilience Cluster is also being activated to bring together companies that strengthen technological sovereignty. On the cybersecurity side, the country is establishing a National Security Operations Centre, a National Cyber Incident Response Centre, and a Digital Forensics Centre. The C4IR under the Ministry of Economy, the Fourth Industrial Revolution Centre established in 2021, provides the institutional foundation for this framework and prioritizes artificial intelligence, data analytics, and digital-twin technologies under the 2026 to 2029 Digital Economy Strategy. One line brings the picture together: rather than relying on headline-grabbing agreements, Baku is building its data and governance layers under central coordination.

The Gulf: After MGX

The story is not a new deal, but the clarification of a structure. On July 1, MGX closed its first fund at $49 billion, $4 billion above target and one of the largest sovereign-backed bets on artificial intelligence to date. MGX has built exposure across frontier-model companies, AI infrastructure, and platform assets, reflecting a strategy of investing across multiple layers of the AI stack. Its model is full-stack aggregation, from laboratories and data centres to platform access.

Placed alongside Saudi Arabia’s HUMAIN, this gives the week’s structural distinction. HUMAIN is betting on domestic vertical integration, with approximately $23 billion in agreements involving Nvidia, AMD, AWS, and Qualcomm, alongside a target of building 1.9 GW of capacity by 2030. These represent two theories: collecting stakes across every layer, or vertically building the stack on one’s own territory. The political scrutiny surrounding MGX-linked investments also illustrates how sovereign technology capital can become intertwined with questions of political access and strategic influence. This is an ongoing backdrop rather than a new development.

North Africa: The operating layer

This week, one frame brings the region together: not who pours the concrete, but who controls the operating layer. According to the Africa Center for Strategic Studies, Chinese firms operate in roughly one-third of African ports, and their presence extends from physical infrastructure into software, automation, and AI-enabled port operations. Since the Belt and Road Initiative, Chinese actors have invested around $50 billion in African port infrastructure. Maritime gateways associated with this presence support trade flows estimated at roughly $350 billion annually.

China’s 15th Five-Year Plan combines continued technological self-reliance, including in advanced semiconductors, with a stronger emphasis on broad AI deployment through its AI Plus agenda. The application layer is increasingly being offered to the Global South, while foundational capabilities, including advanced compute, infrastructure, and core technological capacity, remain concentrated in China.

Egypt enters this picture through its own initiative. The Hendy-Ingabire talks are moving toward a joint African framework for responsible AI, with pilots in healthcare, agriculture, local languages, and digital governance; a memorandum of understanding under development; and Egypt’s role in the African Union AI Working Group and the Smart Africa AI Council.

The energy layer gives concrete form to the principle of energy before compute. Scatec’s Obelisk solar-plus-battery project, an investment of approximately $590 million, competes directly with Egypt’s LNG import exposure on price. The company’s own first-quarter figures place annual savings at $150 million to $320 million at current LNG price levels, and that figure naturally moves with gas prices. The central issue is not the number in the headline, but the fact that a solar-battery installation can now compete with imported gas on price.

Western Balkans: Optics and regulation

Europe’s regulatory terrain has shifted. On June 29, the Council gave final approval to the AI Act simplification package, postponing the deadline for high-risk compliance obligations to December 2, 2027. This delay directly affects the region’s timetable for alignment with the EU acquis and reopens the question of how quickly the Western Balkans should design their own AI legislation.

Serbia is the visible example of this gap. At the SNS rally on June 27, robots dressed in folk costumes danced on stage. The distance between leadership optics, including the GPAI presidency, the Ministerial Declaration, and the region’s first AI strategy, and practice without fully developed legal safeguards is increasingly apparent. Serbia has built considerable international AI visibility, but reporting and policy analysis continue to identify a gap between that positioning and a fully developed domestic legal framework. Given the thinness of dedicated technology media in the region, monitoring this gap requires looking directly at institutional sources, including the EU Council, EEAS, and EuroHPC JU.

Türkiye: Toward the software layer

Türkiye’s connectivity layer is continuing to deepen through fibre expansion and the transition to 5G, strengthening the infrastructure base beneath more software-intensive defence and digital capabilities.

In the defence-industrial sphere, value is shifting from platforms to software. HAVELSAN delivered the ADVENT naval combat management system to Romania, marking its first export of the system to a NATO member. ADVENT is a software-intensive command-and-control system, and an AI-enabled version remains under development. This does not by itself establish Türkiye as a central contributor to NATO’s future combat-data architecture. It does, however, provide a concrete example of Türkiye exporting software-intensive, interoperable command-and-control capability into a NATO ally’s inventory.

The Alliance’s digital transformation, including multi-domain operations, AI analytics, cloud infrastructure, and cyber capability, is increasingly moving interoperability toward an architecture that is digital, network-connected, cyber-secure, and software-defined.

The NATO Summit in Ankara on July 7 and 8, together with the wider calendar of the year, including NATO EDGE-26 in İzmir in November, provides the institutional foundation for this positioning. The shift in the value layer from platforms toward data and software architecture is where infrastructural realism is most clearly visible this week.

The stack is always built in the same order: first energy, then connectivity, then compute. Across all five regions this week, value is being determined by who stands where in that sequence.

Subscribe now

Gulf/Middle East: A Fifty-Billion-Dollar Stack

The week’s most consequential move came from Abu Dhabi. MGX, the state-backed AI investment platform chaired by Sheikh Tahnoon bin Zayed, the UAE’s National Security Adviser, closed a fund of approximately $50 billion, reported by Bloomberg at $49 billion. The capital was raised from regional sovereign wealth funds, global pension funds, and major institutional investors. Bloomberg emphasized that this marked a historic first for Abu Dhabi in raising capital from external partners at this scale: the emirate has turned not only to its own sovereign balance sheets, but also to third-party capital.

More important than the figure is the architecture. MGX’s capital is divided into three pools: AI infrastructure, including data centers and connectivity; semiconductors; and core AI technologies, including models, robotics, and data. This structure reflects not passive fund management, but an assertion of sovereignty across the full stack. MGX already holds stakes in OpenAI, xAI, and Anthropic, and is participating in the AI Infrastructure Partnership alongside BlackRock and Microsoft. The $49 billion is therefore dry powder added to a portfolio already embedded in the global AI stack.

Gulf AI infrastructure also faced direct physical security risks this year. On March 1, AWS announced that two of its facilities in the UAE had been directly struck and that a nearby attack in Bahrain had affected infrastructure, disrupting regional cloud services. Iranian claims of attacks targeting Oracle-linked infrastructure in Dubai were rejected by UAE officials. These developments revealed that the Gulf’s AI expansion faces not only energy and supply-chain risks, but also kinetic threats to concentrated cloud and compute assets. In an analysis published this week, the IISS identified undersea fiber-cable chokepoints and compute concentration as independent categories of security vulnerability. In this context, MGX’s move signals that Abu Dhabi views regional risk not as a reason to retreat, but as a policy challenge to be managed.

On the Riyadh front, a six-month extension to the memorandum of understanding concerning the proposed joint venture between stc and HUMAIN was disclosed to Tadawul on June 18. The proposed structure envisages 1GW of data-center capacity, with an initial capacity of 250MW. The extension reflects the fact that regulatory, operational, and commercial requirements have not yet been completed. The need for additional time for a project of this scale is a measurable indication of the gap between stated targets and operational reality.

Central Asia

On June 26, Prime Minister Bektenov told parliament that the first data center in Ekibastuz is expected to begin commercial operations in 2027, while a second facility of the same capacity is expected to come online in 2028. This is the most concrete official figure made public since the Firebird–Nvidia framework agreements signed on June 15-16, and it is critical for interpreting the deal correctly. The agreement package is valued at up to $10 billion in total; the first phase is worth approximately $5 billion and covers the planned initial 125MW facility. The content of the second phase has not yet been fully defined. The widely circulated 1GW figure in international reporting does not refer to contracted capacity, but rather to the project’s long-term target and the upper limit of the Ekibastuz site’s potential power capacity.

The strategic logic of the agreement is coherent: Ekibastuz’s low-cost coal power, US-aligned chip supply through Nvidia, and the institutional backbone provided by the Ministry of Artificial Intelligence and Digital Development, established in 2025. Kazakhtelecom has been designated as the operator, while Firebird is the development partner. The country declared 2026 the Year of Artificial Intelligence and Digitalization, while the Digital Qazaqstan strategy provides the policy framework. Kazakhstan’s accumulation in AI infrastructure is also real: the Alem.Cloud supercomputer cluster entered service in July 2025 with 2 exaflops of capacity, while the AI-Farabium cluster began operations in October with 1,500 petaflops. Data Center Valley is being positioned as the next major leap in scale built on this accumulated base.

The central question will be answered in 2027: can the framework agreement be transformed, on schedule, into an operational facility with secured power, GPUs, and access to international customers? Whether Data Center Valley is a functioning hub or merely a headline will be determined by the operational performance of Phase One in 2027.

North Africa: Egyptian Coordination, Morocco Regionalizes

Egypt’s June 21 directive on a three-ministry data-center and cloud-computing strategy consolidates several policy moves that accumulated throughout June under a single framework. On June 13, Communications Minister Raafat Hindi told the Senate’s Defense and National Security Committee that artificial intelligence had become a strategic enabler of national security and economic development, and announced that the National AI Council had been expanded to cover quantum computing and emerging technologies. During the same week, Hassan Allam Digital Infrastructure received a license from the telecommunications regulator and launched its approximately $400 million first-phase data-center investment. The June 21 directive, meanwhile, brings together the ministries of Communications, Electricity, and Investment for the first time under a joint framework, tasking them with producing an investment map based on power capacity, renewable energy, and connectivity infrastructure. Hindi’s framing makes Egypt’s intent clear: data centers are a cornerstone of the digital sovereignty agenda, while Egypt’s geographical position, hosting multiple undersea cables connecting Europe, Africa, and Asia, provides structural leverage.

Karnak, the sovereign Arabic large language model developed to operate on Egyptian infrastructure, was presented during the June 13 session not as a research output but as an existing infrastructure component. This framing is significant: both Egypt’s Karnak and Saudi Arabia’s Allam are now being described in official policy discourse as elements of national digital infrastructure. The transition from tool to stack is the clearest indication of the region’s evolving discourse.

In Morocco, the Jazari Institute moved to Merzouga in the Drâa-Tafilalet region from June 16–20 through its Rally AI Future Lab initiative. Approximately 1,000 researchers, engineers, students, entrepreneurs, and developers participated in the five-day intensive program. An agreement was also signed to establish a YouCode coding school in Errachidia. The geographical choice, a rural and agricultural region rather than the capital Rabat or Casablanca, represents an attempt to materialize the “AI Made in Morocco” strategy’s claim of reaching all twelve regions on the ground. Al Jazeera’s June 26 analysis, published against the backdrop of the African Union ministerial meeting held in Tangier in April, questioned who would control Africa’s AI infrastructure. It provides the appropriate analytical framework for this week’s Egyptian and Moroccan initiatives: across North Africa, the question is no longer whether AI capacity will be built, but under which governance conditions, under whose control, and whether this infrastructure will carry genuine sovereignty beyond its name.

Western Balkans: Legislation Without a Stack

POINT 14, held in Sarajevo on June 11-12, provides the analytical anchor for this week’s regional assessment. Researchers from the universities of Durrës, Sarajevo, and Graz documented the pattern predicted by infrastructural realism: although Western Balkan governments adopt EU digital legislation under external pressure or through imitation, they lack the institutional capacity necessary to sustain long-term implementation. Chinese technology presence is strongest in Serbia, followed by Montenegro and Bosnia and Herzegovina; Albania and North Macedonia have more deliberately limited this presence. Kosovo and North Macedonia face high donor dependence; Montenegro struggles with capacity constraints; Bosnia with fragmented governance; and Serbia continues to seek a balance between the EU, China, and Russia. The most critical structural gap is this: EU digital rules were designed for member states with effective enforcement bodies, while no coherent pre-accession compliance model exists for candidate countries.

During this period, I did not encounter a region-wide AI infrastructure announcement that would fundamentally alter the overall assessment of the region. However, there is a development from May that should be noted: the first AI working group meeting of the Western Balkans Six, held in Podgorica by the Regional Cooperation Council and UNDP, marks the formal beginning of a regional AI policy-coordination process under the 2025-2028 Common Regional Market Action Plan. It is a modest but structurally meaningful step: the region is beginning to develop a common policy language for AI governance. Yet the gap between this coordination layer and the institutional capacity and physical infrastructure required for credible implementation remains the fundamental constraint, and the POINT 14 researchers documented this clearly.

Türkiye: The Plan Meets the Calendar

The week’s only concrete development emerged from the ecosystem layer of the June 13 Action Plan. Eleven Turkish companies participated in GITEX AI Europe, held in Berlin on June 30-July 1, under a joint Türkiye Pavilion coordinated by the Presidency’s Investment and Finance Office. Participating companies, including B2Metric, Buluttan, and İTÜ ARI Teknokent, are tangible examples of the applied AI layer that the Action Plan’s “produce” pillar seeks to accelerate.

The more significant signal is that GITEX AI Türkiye has been confirmed for September 9–10 at the Istanbul Expo Center, jointly hosted by the Investment Office and the Ministry of Industry and Technology. Its organization in Istanbul not as a sub-section of a general technology fair, but as an independent GITEX AI event, puts the Action Plan’s vision of an “international showcase and investment diplomacy city” onto the calendar.

Subscribe now

Gulf / Middle East: The UAE Consolidates Its Stack

On 14 June, Sheikh Mohammed bin Rashid Al Maktoum approved the establishment of the Artificial Intelligence and Data Authority, bringing under one roof a structure previously dispersed across three separate institutions. The new authority incorporates the AI Office, the digital government sector within the telecommunications regulator, and the UAE Data Office. It reports directly to the Cabinet and operates under the leadership of Minister of State for Artificial Intelligence Omar Sultan Al Olama. Its remit spans a broad range of functions, from developing the national AI strategy and governing data to autonomous government services and international partnerships.

Institutionally, this is a consequential move. The UAE has spent years building a layered AI ecosystem: Stargate, G42, MGX, and HUMAIN-linked compute partnerships. Yet its governance architecture remained distributed across several bodies. On 14 June, the control layer was brought into alignment with the infrastructure layer. Whoever governs the data platform governs the stack.

A few days earlier, the Kuwait Investment Authority joined KKR, NVIDIA, and Vistra to establish Helix Digital Infrastructure, with more than $10 billion in committed capital to develop data centres, power generation, transmission, and connectivity infrastructure for hyperscale operators. The venture is led by former AWS CEO Adam Selipsky. This is KIA’s third major AI-infrastructure commitment within the past year, following the AIP coalition and the Brookfield AI Infrastructure Fund. The pattern is not accidental: Gulf sovereign wealth funds are systematically positioning themselves across the entire AI stack, from power generation to fibre infrastructure.

These moves are unfolding against the backdrop of the ongoing war in the Middle East. An analysis published by the Middle East Council on Global Affairs on 11 June documented how AI-generated fake videos, including footage purporting to show smoke rising from a skyscraper in Bahrain and a deepfake depicting the sinking of a US aircraft carrier, have been deployed as instruments of information warfare during the conflict. Its conclusion was clear: investment in sovereign digital infrastructure can no longer be separated from military and political resilience. The UAE’s governance consolidation and Kuwait’s infrastructure push are not merely economic decisions; they are responses to a threat environment that has turned control over the AI stack into a national-security issue.

Central Asia: Two Countries, One Week, Competing Models

On 15 June, Kazakhstan signed a $10 billion package with Firebird and NVIDIA to build a Data Centre Valley in Ekibastuz. The plan envisages a cluster of 100,000 NVIDIA GPUs, with Firebird’s co-founder arguing that the investment could place Kazakhstan among the world’s top ten AI powers by 2027. The agreement came in the sixth month of Kazakhstan’s Year of AI and Digitalisation, following the entry into force of Central Asia’s first comprehensive AI law and the creation of a dedicated Ministry of Artificial Intelligence and Digital Development. Kazakhstan has built a plausible institutional framework; however, the real test of 2026 will be whether the National AI Platform becomes genuinely usable for actors beyond the state and whether compute commitments move from announcement to implementation.

The following day, Uzbekistan opened the fifth Tashkent International Investment Forum, held from 16 to 18 June. Minister of Digital Technologies Sherzod Shermatov summarised the country’s infrastructure thesis in one sentence: “AI requires computing power. Computing power requires energy.” The forum brought together 8,300 participants from 100 countries, companies managing $42 trillion in assets, and 193 US firms, the largest American business delegation in the history of bilateral relations. AI and digital public administration were among the principal investment themes, alongside critical minerals and renewable energy.

The comparison is revealing. Kazakhstan is advancing through the state layer: sovereign legislation, national platforms, and direct GPU procurement with US partners. Uzbekistan is advancing through the market layer: attracting foreign capital, building its energy base, and positioning itself as a long-term investment destination. Both are competing to define what Central Asia’s AI ecosystem will look like and neither has yet demonstrated that infrastructure investment has translated into sovereign capacity.

North Africa: Morocco’s Distributed Bet

On 16 June, Morocco launched the Rally AI Future Lab in Merzouga, a five-day intensive programme organised by the Al Jazari Institute under the patronage of King Mohammed VI. Bringing together one thousand researchers, engineers, students, and developers, the event forms part of Morocco’s “Indigenous AI” strategy and its objective of extending the Al Jazari network across all 12 administrative regions of the country. The geographical objective is explicit: the government aims to make AI literacy and capacity accessible in the mountains as much as in Casablanca.

This reflects a deliberate architectural choice. Rather than concentrating AI development in a single national centre, Morocco is building a distributed network of centres of excellence, with Al Jazari Root in Rabat positioned as the coordinating node. The laboratory in Merzouga, in the Drâa-Tafilalet region, one of the country’s most remote areas, signals whom the strategy is designed to serve. It also aligns with the positioning articulated in April by Minister of Digital Transition Amal El Fallah Seghrouchni within the EU–Morocco “third voice” framework: Morocco is presenting AI governance as a question of democratic infrastructure, distinct from both the US-led innovation-first model and China’s state-directed model.

Western Balkans: The Framework Question

On 16 June, UNESCO published an account of a regional conference on AI in education held in Sarajevo from 7 to 10 May. More than 80 teachers and education practitioners from Bosnia and Herzegovina, Croatia, Serbia, and Slovenia participated in the event, which addressed ethical AI use, data protection, learner well-being, and classroom implementation. Its composition reflects a structural reality of AI adoption in the Western Balkans: the defining frameworks are arriving from outside, through international institutions, before domestic regulatory and educational infrastructures are fully in place.

This is not accidental. The region is caught between EU accession ambitions, US influence, and continued engagement with Chinese technology providers, with Serbia maintaining the region’s strongest technological ties with China. The Balkans are living less with an AI architecture of their own choosing than with one that has been selected for them: alignment with the EU AI Act is imposed through the accession process; UNESCO ethical frameworks take root through multilateral programmes; and Chinese infrastructure is already layered into the region through earlier telecommunications and surveillance investments. The Sarajevo conference is a small but concrete illustration of how this process unfolds at the classroom level. The central question, who writes the rules, remains structurally unanswered.

Türkiye: Sovereignty as Strategy

On 13 June, President Erdoğan unveiled Türkiye’s AI Action Plan 2026–2030 at Tersane Istanbul, a site whose roots extend back to the reign of Ottoman Sultan Mehmed II. The choice of venue was deliberate. So was the framing: Erdoğan described digital sovereignty and military capability as inseparable, and structured the plan around four pillars: discover, benefit, produce and govern. The government said it would mobilise at least $10 billion, primarily from the private sector, for data centres, cloud computing, and AI infrastructure, while allocating at least 2% of public investment programmes to AI projects.

What distinguishes this announcement from earlier Turkish AI strategies is the explicit emergence of a domestic large language model ecosystem. Erdoğan highlighted Bilge, TÜBİTAK’s developing domestic language model, as an important indicator of progress toward digital sovereignty. He also referred separately to a large language model being developed by the T3 Foundation and Baykar, and to HAVELSAN’s nine-billion-parameter large language model on its MAIN platform. The framing matters: these are not initiatives operating at the margins of state strategy. The defence industry, public research institutions, and national foundations are institutional actors positioned within the production layer of a state-directed AI stack.

The fundamental risk is not merely technological dependency. It is the possibility that such dependency becomes algorithmic and then epistemic dependency, that the systems through which Türkiye’s institutions operate may embed a worldview Ankara neither designed nor controls. Bilge, Baykar, Main: these are not merely technical assets. They are Türkiye’s attempt to answer its own questions before someone else does.

Three regions. One pattern. This week, the geopolitics of artificial intelligence infrastructure moved simultaneously across three different geographies: the Caspian basin, where American companies are planting their institutional flags; Sub-Saharan Africa, where the gap between digital ambition and real compute sovereignty has become a structural condition; and the Western Balkans, where the EU is quietly rebuilding the region’s digital architecture before membership is complete.

The mechanisms differ. But the underlying logic is the same: whoever controls the infrastructure determines the outcome.

OpenAI’s Caspian Tour

This week’s most important story of AI diplomacy took place in Astana on June 3. An OpenAI delegation met with Kazakhstan’s Deputy Prime Minister and Minister of Artificial Intelligence and Digital Development, Jaslan Madiyev. The meeting focused on cooperation in education, the expansion of ChatGPT Edu, and “future areas of partnership,” a diplomatic formulation that effectively signals market positioning.

“Whoever controls the infrastructure determines the outcome.”

OpenAI did not bring only tools to the meeting. It brought data. According to the analysis presented, Kazakhstan has entered the global top 30 in ChatGPT usage; Kyrgyzstan ranks 52nd and Uzbekistan 58th in consumer use per capita. As of March 2026, 23 percent of messages by Kazakh users were work-related. The demographic distribution clearly shows where dependency is beginning to take root: 72 percent of all requests come from users aged 18–34.

Just days earlier, Azerbaijan’s Ministry of Science and Education signed a memorandum of understanding with OpenAI to develop an adaptive learning platform for more than 500,000 students. Azerbaijan’s parliament also framed the new agreements and AI working-group activity as an opportunity to attract more AI technologies and innovation from the United States. This should be read against the backdrop of reports that Azerbaijan–US trade volume increased 2.2 times in the first quarter of 2026, reaching $296 million.

Two countries. One week. The same company. OpenAI is pursuing a coherent Caspian strategy: deploying educational infrastructure as a first step toward creating institutional dependency before a regulatory framework has even formed. The game is simple: become the default standard before becoming the subject of policy.

This matters on two analytical layers. The first is the infrastructure layer: ChatGPT Edu licenses create institutional dependencies that are difficult to reverse. When 165,000 teachers in Kazakhstan integrate their lesson flows into a single American platform, switching costs become a question of sovereignty. The second is the epistemological layer: large language models do not merely process language; they also carry a worldview. The question of which system processes a generation’s educational queries is not a pedagogical question. It is a geopolitical one.

“The question of which system processes a generation’s educational queries is not a pedagogical question. It is a geopolitical one.”

The Common Turkic Language Model project adopted at the Turkic Summit in May was precisely a recognition of this reality. The adoption of AlemGPT as the anchor of a regional AI platform, the joint CubeSat satellite project, and Tokayev’s proposal for a cybersecurity council were not disconnected technical initiatives. They were a coherent attempt to define a layer of the AI stack, infrastructure, compute, and language model, that could be positioned outside both American and Chinese architectures. The Turkic language model project, in particular, is a direct counter-move to OpenAI’s ChatGPT Edu deployments in the same geography: if the system processing your students’ queries also shapes what counts as knowledge in your language, building your own model is not a cultural project; it is a sovereignty project.

But the Summit left one question unresolved: does the OTS actually have the institutional capacity to implement this? AlemGPT is a Kazakh product developed with national government support. A genuinely multilingual Turkic model, covering Kazakh, Uzbek, Kyrgyz, Azerbaijani, Turkmen, and Uyghur, requires data infrastructure, compute resources, and cross-border governance arrangements that no OTS member possesses alone. Kazakhstan has enough power to anchor the process, but anchoring is not the same as co-building. OpenAI’s arrival in Astana coincided precisely with this gap, offering the institutional depth that the OTS Turkic language model has not yet filled.

Before all of this, GITEX AI Central Asia and Caucasus, the region’s largest technology gathering to date, took place in Almaty on May 4-5: more than 600 participating companies from 60 countries and over 10,000 professionals. Around the same period, the AI Leaders 2026 graduation ceremony at Alem.AI in Astana brought together more than 800 executives from over 150 organizations across nine countries; it marked the conclusion of an eight-week program run in partnership with Stanford HAI and OpenAI Academy. The institutional architecture of the Caspian AI ecosystem is being built rapidly and its main contractors are American institutions, with Kazakhstan as the central pillar.

Russia did not remain in the background this week. On June 6, Uzbekistan and Russia agreed at the St. Petersburg International Economic Forum to prepare and sign a bilateral roadmap in digital technologies, artificial intelligence, and telecommunications. Uzbekistan’s Minister of Digital Technologies, Sherzod Shermatov, also presented his country’s AI strategy at a session that included Tajikistan, Turkmenistan, Kyrgyzstan, and Belarus. Tashkent’s clearest position in the region is this: opening the door to Silicon Valley investment while deepening digital dependency with Moscow.

Turkmenistan also deserves a note this week. On June 5, the Communications Minister announced that an AI Competence Center could be established under Türkmentelecom, a small but notable signal from the region’s most closed digital environment. When combined with the late-May memorandum of understanding with Bitdeer Technologies, the UNDP-supported national AI strategy process, and Chinese partnership discourse circulating through Turkmen official channels, it becomes clear that Ashgabat is moving faster than its reputation suggests. What it is building, and whose architecture it is using, are questions worth watching.

Africa: A Structural Condition

This week’s Africa headline was startup funding, which reached $1.3 billion as of June 3 — a pace exceeding the record first half of 2025. This jump can largely be attributed to a single deal: Spiro’s $215 million equity round, closed on June 1 with backing from Denmark’s Impact Fund and institutional investors from Europe and Africa.

Spiro is Africa’s largest electric vehicle operator — 100,000 motorcycles across seven markets, 2,500 battery-swapping stations, and now enough funding to enter the DRC and Ethiopia. But defining it simply as “an African EV company” understates the structural reality. Spiro is building a distributed, IoT-connected energy storage network in urban Africa. The stations are solar-powered. Once it reaches sufficient scale, this network becomes an infrastructure that governments will want to regulate, partner with, or eventually acquire. The entry of $215 million in Danish pension capital into this field is not a speculative bet; it is an institutional signal. It is also an investment in Africa’s energy sovereignty: a bet that battery-swapping infrastructure distributed across East and West Africa can simultaneously reduce dependence on imported fossil fuels and centralized grid power.

The rest of the week was less celebratory. The most striking dimension of African AI news was job losses. According to TechCabal’s 2026 tracker, layoffs attributed to AI have exceeded 1,000 across the continent, significantly higher than the same period last year. The justification has shifted: companies no longer cite “market correction” or “macroeconomic pressure.” They point directly to artificial intelligence. Jumia, the continent’s largest e-commerce platform, has announced plans to cut at least 200 more jobs as it integrates AI across operations from logistics and finance to marketing and customer support, aiming to reach profitability in the final quarter of 2026, after already reducing its workforce from 4,318 people in 2022 to fewer than 2,000. Nigeria-based crypto startup Zap Africa laid off 44 percent of its employees after deploying an AI tool to handle first-line customer service.

The labor question is where Africa’s AI story diverges most sharply from narratives in other regions. In Kazakhstan, AI is framed as a workforce upskilling mechanism. In Serbia, it appears as a techno-economic leapfrog opportunity. In Nigeria and Kenya, however, the reality on the ground looks different: efficiency gains accrue to companies operating at continental scale, while displac

ement concentrates in customer-facing and back-office roles held by young urban workers. Africa has the youngest population in the world. It is also currently the continent with the least compute sovereignty.

This needs to be stated clearly: Africa controls less than one percent of global compute power. By mid-2025, Africa had roughly 223 data centers across 38 countries, but its share of global data-center capacity remained below 1 percent. The Arizton market research report published on June 3 projects that the Middle East and Africa data center construction market will reach $15.97 billion by 2031, expanding at a compound annual growth rate of over 25 percent; Nigeria stands out as the fastest-growing investment market. But who is building these centers? Equinix. Digital Realty. Microsoft. Amazon Web Services. Cassava Technologies, expanding with Nvidia support. Infrastructure is being built rapidly, but with American capital and under American legal frameworks.

This is the CLOUD Act problem, the issue repeatedly surfaced in TechCabal’s analysis: even data physically stored in a facility in Lagos operated by a US company remains subject to American legal jurisdiction. The continent is building digital infrastructure, but the sovereignty of that infrastructure is a different question from its geography.

South Africa’s policy situation captures this gap. The draft National AI Policy was withdrawn at the end of April after it emerged that the document contained fake citations generated by AI, a government AI policy document contaminated by the very technology it sought to regulate. A revised draft is scheduled for January 2027. In the meantime, AI adoption in South Africa reached 23.1 percent of the population in the first quarter of 2026; the data center market is expected to grow from $70 million to $572 million by 2031. Infrastructure is moving forward. Governance is waiting.

The Salesforce Agentforce World Tour held in Johannesburg on June 3 revealed a concrete dimension of the skills gap: South African businesses are adopting AI much faster than the education system can adapt. Qualification frameworks operate on five-year cycles; AI skill requirements change within months. This is not only Africa’s problem, but in South Africa, where youth unemployment exceeds 60 percent, it carries particular weight.

The Balkans: The EU Rewires the Region

This week’s Western Balkans story revolved around a single event: the EU–Western Balkans Summit held in Tivat, Montenegro, on June 5. And one critical threshold the day before: the EU Council’s June 4 decision to formally open “Roam Like at Home” negotiations with the six Western Balkan partners.

It is worth beginning with the roaming decision, because it instructively reveals how EU integration works in practice. The “Roam Like at Home” area currently covers the EU, the European Economic Area, and, from January 1, 2026, Moldova and Ukraine. Its extension to the Western Balkans builds on the 2022 Tirana Declaration and the 2023 agreement among 38 European and Balkan telecommunications operators. The goal is the complete removal of additional roaming charges by 2028. Once agreements are finalized and each partner fully aligns with EU roaming rules, seamless connectivity becomes a de facto reality, years before any of these countries formally joins the EU.

This is not a minor development. Harmonization of telecommunications regulation means that the mobile data infrastructure of six non-EU countries will operate within the legal framework of EU rules; this carries serious implications for the legal regimes governing data governance, data sovereignty, and cross-border data flows. The EU is extending the digital single market outward before membership takes place. When accession happens, if it happens, much of the infrastructure layer will already be attached to European frameworks.

The Tivat Summit confirmed this direction. European Council President António Costa described the Western Balkans as an “integral part” of the European Union. The summit agreed to launch new cooperation between the region and the EU Cybersecurity Agency, ENISA, to strengthen resilience against cyber and hybrid threats. The framework also included the European Democracy Shield and the Rapid Alert System against disinformation. The cybersecurity offer is not merely technical cooperation; it is an alignment mechanism against what the EU describes as “foreign information manipulation and interference.”

For Serbia, this week’s summit took place within a complicated position. Serbia is the region’s strongest technology success story, ICT services exports reached a record €4.552 billion in 2025, a tenfold increase compared with 2012, making it the country’s largest net export sector. The SAIFA project, launched on May 14 at the Palace of Science in Belgrade with €3.8 million in EU funding, integrates Serbia into the EuroHPC AI Factory Antenna network; it creates direct links with the Pharos AI Factory in Greece and the IT4LIA AI Factory in Italy. For a non-EU country, this is meaningful digital infrastructure integration before political accession.

Still, Serbia’s geopolitical position remains genuinely complex. The Recursive’s ICT ecosystem analysis published three weeks ago named the tension directly: alongside European AI integration, China’s investment presence is expanding. GITEX AI Serbia is coming to Belgrade in 2027. Expo 2027 will be held in Belgrade. Astana hosted GITEX AI Central Asia this month. Both cities are experiencing the same structural pressure, strategically positioned between European integration frameworks and China’s digital infrastructure ambitions, with American AI companies forming a third front.

North Macedonia’s Vezilka AI Antenna Factory is this week’s most analytically interesting small story. Again funded by the EuroHPC Joint Undertaking, and again connected to Greece’s Pharos AI Factory. The project’s first stated objective: building a Macedonian language model. A non-EU country with a population of two million, a contested name, and complex relations with its neighbors has selected language model sovereignty as its first priority in AI infrastructure. This is a precise and deliberate political decision about what AI is for.

Albania sits at the opposite end. It officially joined the EuroHPC Joint Undertaking in June 2025. It did not receive any AI Factory Antenna funding in the first wave. Tiran Times’ February assessment made the sharpest observation: a “widening gap between political spectacle and policy implementation.” Albania’s E-Albania portal provides more than 95 percent of government services online, a genuinely impressive achievement in digital governance. But the gap between e-government delivery and AI infrastructure integration is exactly where EU-adjacent digital sovereignty is either built or not built.

The Pattern

Three regions. Different levels of development, different relationships with great powers, different domestic political conditions. But everywhere, the same structural question surfaces: who is building the infrastructure, under whose legal framework, and who defines what the system will know?

OpenAI is in Astana and Baku. American capital is in Lagos and Cape Town. EU architecture is in Belgrade and Skopje, and now extends into the telecommunications infrastructure of six non-EU countries. Russian digital ministry officials are in St. Petersburg with the Uzbek delegation. Bitdeer Technologies is in Ashgabat. Nvidia’s GPU clusters are in Johannesburg.

Infrastructure is being built. Sovereignty questions are being deferred. For the vast majority of countries covered in this newsletter, smaller states, states not yet anchored in any major alliance, states with limited compute capacity and legal leverage, deferral is not neutral. It accumulates into structural dependency.

“Infrastructure is being built. Sovereignty questions are being deferred.”

The Macedonian language model is this week’s clearest counterexample. A small country, limited resources, complex political position. The first priority in AI infrastructure: building in your own language, coding your own knowledge base, and constructing what no one else will build for you.

This is infrastructural realism in its simplest form.

Sources for this week: The Astana Times, Trend.Az, AzerNews, Times of Central Asia, Qazinform, CaspianPost, Eurasianet, HPC Serbia, Western Balkans Info Hub, The Recursive, Council of the EU, EEAS, TechCabal, TechCabal Insights, Techpoint Africa, Bloomberg, Arizton.

Personal views. Not institutional positions.

Subscribe now

MIDDLE EAST

According to Microsoft’s Global AI Diffusion Report published in May 2026, the Middle East is home to some of the world’s most extreme contrasts in AI adoption. The UAE ranks first globally in per-capita generative AI usage, with 70.1% of its working-age population actively using generative AI tools. It also recorded a remarkable 6.1 percentage-point increase compared to the previous quarter, a growth rate unmatched by any other economy globally. Qatar (41.8%), Saudi Arabia (29.4%), Jordan (29.7%), and Lebanon (27.3%) also demonstrate relatively strong performance, highlighting the overall momentum of the Gulf and Levant economies in embracing AI.

By contrast, Iraq (12.5%) and Iran (12.6%) remain significantly below the global average. Iran’s figure is particularly noteworthy. The war that erupted in March 2026, followed by the Strait of Hormuz crisis, placed additional strain on the country’s already fragile digital infrastructure. International sanctions, restricted access to global technology platforms, and rising energy costs continue to severely constrain AI adoption. In Iraq, the underlying causes are different, but the outcome is similar. Decades of instability have left behind major infrastructure deficiencies, low internet penetration, and limited institutional capacity for digital transformation, all of which continue to hold the country back.

The 57-point gap between the UAE and Iraq is more than a statistical curiosity. The emergence of two radically different AI futures within the same geographic region could have significant medium-term implications. Economies with high levels of AI adoption are accumulating advantages in productivity, public service delivery, and economic diversification, while countries with lower adoption rates risk remaining outside this cycle of technological transformation. As the technological divide between the Gulf states and conflict-affected countries deepens, regional power asymmetries are increasingly acquiring an AI dimension, one that is likely to become an important factor shaping the geopolitical landscape of the Middle East over the coming decade.

Saudi Arabia closes Hajj 2026 with its most AI-intensive operation yet

The 2026 Hajj season, held between May 25-30, concluded on May 31 with Saudi Arabia announcing the successful management of 1.7 million pilgrims without any major health incidents. At the same time, it marked a historic milestone in the large-scale deployment of artificial intelligence. Saudi Arabia’s Data and Artificial Intelligence Authority (SDAIA) implemented AI-powered crowd analytics systems, drones, and facial recognition technologies centered around the Unified Security Operations Center in Mecca. Drones were used both for aerial monitoring and for enforcing permit regulations under the slogan “No Hajj Without a Permit.” Meanwhile, Ericsson and Mobily celebrated 20 years of Hajj connectivity by expanding 5G coverage and integrating AI into network optimization and real-time performance monitoring across the Mashair region.

This year’s Hajj effectively served as a live stress test of state-level AI deployment under real-world conditions involving 1.7 million people, an environment no controlled experiment could replicate. The lessons learned by SDAIA during this season can be expected to directly inform future iterations of the system.

On the investment front, Abu Dhabi-based MGX continued expanding its global AI portfolio. The company participated in the $2.1 billion Series B funding round of the British AI drug-discovery company Isomorphic Labs, alongside Alphabet, Google Ventures, and Temasek. MGX has now invested in OpenAI, xAI, Anthropic, and Isomorphic Labs, effectively mapping much of today’s frontier AI ecosystem.

TÜRKİYE

Microsoft’s Q1 2026 data paints a noteworthy picture for Türkiye. An estimated 17.4% of the country’s working-age population actively uses generative AI tools, and with a 30.3% growth rate since June 2025, Türkiye ranks seventh globally among the fastest-growing AI adoption markets. This places the country in the same category as some of the most closely watched digital transformation success stories of recent years, including Japan, South Korea, and Thailand.

This trend is far from accidental. Türkiye’s 2026 Presidential Annual Program positions artificial intelligence not merely as an innovation agenda but as an instrument for strengthening state capacity. The program emphasizes machine learning-based risk analysis in tax and customs administration, decision-support systems in healthcare, and predictive analytics for agricultural planning. Such an approach offers a practical example of how a middle power can leverage AI in a global environment increasingly dominated by major technological powers. At the same time, the growing capacity of institutions such as TÜBİTAK and KOSGEB to support AI-focused research, development, and innovation is helping sustain this momentum.

When combined with ongoing infrastructure investments and an evolving policy framework, this acceleration in AI adoption is gradually positioning Türkiye as a more influential actor in regional AI dynamics. For a country situated at the intersection of the Balkans, Central Asia, the Middle East, and North Africa, this is not merely a digital transformation, it is also related to strategic positioning.

CENTRAL ASIA

Two races, two strategies

This week, Central Asia’s two most active digital economies advanced in notably different directions, offering an instructive comparison.

Kazakhstan signed a memorandum of understanding with a U.S.-linked international consortium to build one of the region’s largest data centers. The Tier III–IV facility could reach a capacity of 200 MW, with total investment projected at up to $1.5 billion. To secure independent electricity supply, a 250 MW natural gas power plant is also included in the project. The initiative reflects Kazakhstan’s ambition to be positioned as a regional computing hub. However, energy constraints remain a significant bottleneck. Kazakhstan aims to double its electricity generation capacity within a decade, but implementation challenges persist, with several contracts initially awarded to Russian firms later transferred to Chinese companies.

Uzbekistan, by contrast, is pursuing a different model. Rather than building sovereign AI infrastructure from scratch, it is focusing on integrating AI into existing systems through strategic partnerships. At the GSMA M360 Eurasia Forum in Samarkand on May 20-21, Minister of Digital Technologies Sherzod Shermatov outlined Uzbekistan’s strategy of becoming Central Asia’s leading IT hub through infrastructure development, tax incentives for technology companies, and accelerated AI adoption. At the same time, Uzbekneftegaz signed agreements with international partners including SOCAR and SLB to introduce AI-powered seismic interpretation and geological modeling tools into its operations. Instead of waiting to develop a national AI model, Uzbekistan is embedding AI directly into one of its most strategic industries. Two countries, two philosophies of AI sovereignty. Both worth watching closely.

Three Central Asian Countries Enter the Global Fast-Growth Rankings

Microsoft’s Q1 2026 data presents a striking picture for Central Asia. Kazakhstan leads the region with an AI adoption rate of 15.9% and ranks 10th globally among the fastest-growing AI markets, recording 25.6% growth since June 2025. Even more noteworthy is the presence of Kyrgyzstan and Uzbekistan in the same global ranking, placing 11th and 12th respectively, both with growth rates of 25.5%. The fact that three Central Asian countries appear simultaneously among the world’s fastest-growing AI adopters suggests that the region’s momentum is not an isolated phenomenon but rather an indication of a broader structural transformation.

At the same time, the figures reveal a significant internal hierarchy. Kazakhstan stands at 15.9%, while Kyrgyzstan reaches 9.5% and Uzbekistan 7.2%, highlighting substantial disparities within the region itself. Tajikistan and Turkmenistan remain at just 6.1%, a figure that represents a considerable structural lag in a world where the global average has reached 17.8%. Although growth rates are broadly similar, the starting points are not, meaning that the gap is not narrowing but widening.

This regional divide also raises important questions about who is actually benefiting from AI investments and digital infrastructure projects. While Kazakhstan is building large-scale data centers and positioning itself as a regional computing hub, Tajikistan continues to struggle with reliable access to electricity. As Microsoft’s report emphasizes, access to electricity, internet connectivity, and digital skills remain the fundamental determinants of AI adoption. Consequently, the future of AI in Central Asia may depend less on headline growth rates and more on whether the region can address these underlying developmental disparities.

NORTH AFRICA

Ambition meets adoption gap

North Africa’s Q1 2026 figures suggest that the region’s ambitious AI strategies have yet to fully translate into widespread adoption. Libya leads the region with an AI adoption rate of 15.0%, followed closely by Egypt (14.8%), Tunisia (13.5%), Algeria (13.2%), and Morocco (11.7%). All five countries remain either slightly below or roughly in line with the Global South average of 15.4%. Growth rates have also been relatively modest. Morocco and Tunisia recorded quarterly increases of 0.8%, Egypt 1.4%, and Algeria 1.2%, all below the global quarterly growth average of 1.5%.

This picture is not surprising, but it raises an important question. Governments across the region have announced national AI strategies, sovereign AI initiatives, and major infrastructure investments, yet the impact of these efforts has not fully materialized in adoption figures. Morocco’s JAZARI ROOT Institute, Egypt’s Karnak model, and Algeria’s supercomputing center in Oran are either still under development or have only recently become operational. Their effects on AI adoption are therefore unlikely to be fully visible before the end of 2026.

The more critical question is whether North Africa can convert policy momentum into adoption momentum. The structural barriers identified by Microsoft’s data suggest that this challenge is less about AI policy itself and more about the foundations required to support it. With electricity access at 88.9%, internet connectivity at 65.7%, and digital skills at 48.2%, all broadly aligned with Global South averages, the region’s AI future may depend less on additional strategy documents and more on sustained investments in education, connectivity, and digital inclusion. Without addressing these underlying constraints, the gap between ambitious policy objectives and real-world adoption is likely to persist.

BALKANS

Innovation gaps mapped, and a regional summit wraps

The ITIF Balkan Subnational Innovation Competitiveness Index, published May 26, delivered a clear-eyed assessment: Western Balkan countries risk falling behind as AI accelerates global economic divergence, particularly if they continue to underinvest in innovation. The index’s most striking finding is territorial — capital cities are pulling ahead while large portions of each country remain disconnected from the digital economy. On the positive side, FDI into the Western Balkans averaged 6.4% of GDP between 2020 and 2023, quadruple the EU average — capital is arriving, even if innovation capacity has not yet caught up with it.

The week closed with the Southeast Europe AI Summit wrapping in Novi Sad on May 28–29, bringing together research, enterprise, venture capital, and public institutions to build the connections needed to drive regional AI innovation. With GITEX confirmed for Belgrade in 2027, and Serbia’s ICT exports having grown tenfold since 2012, the foundations are real. The ITIF data is simply a reminder of how uneven the ground beneath them still is.

EU Membership as a Driver of AI Adoption

Microsoft’s data offers perhaps the most striking structural insight into AI adoption across the Balkans: there is a strong and direct correlation between EU membership and AI adoption rates. Bulgaria (29.7%) and Croatia (26.1%), both EU member states, outperform Serbia (24.1%), Bosnia and Herzegovina (22.1%), Albania (18.5%), and Romania (17.5%) by a noticeable margin. This gap is not coincidental. Access to mechanisms associated with EU membership, including the Digital Europe Programme, European Digital Innovation Hubs, and the EU AI Factories initiative, directly strengthens national capacities for AI adoption and diffusion.

Another noteworthy finding is the performance of Bosnia and Herzegovina. With an adoption rate of 22.1%, Bosnia ranks just behind Serbia and stands 3.6 percentage points ahead of Albania despite not benefiting from the same level of institutional integration with the European Union. When compared with the ITIF Balkan Innovation Competitiveness Index released during the same week, an interesting contradiction emerges. Bosnia appears as one of the region’s most fragile innovation ecosystems, yet this vulnerability is not fully reflected in its AI adoption figures. This divergence suggests that AI adoption can advance independently of broader innovation capacity, at least in the short term. Over the longer run, however, sustainable growth is likely to depend on both evolving together.

Albania, meanwhile, records an adoption rate of 18.5%, reinforcing the regional disparities identified by both ITIF and Microsoft. The dynamism of the capital city has yet to translate into comparable levels of digital transformation across the country as a whole. The Albanian case highlights a broader challenge facing the Balkans: the uneven geographic distribution of technological capabilities, where national averages often conceal significant internal disparities in access, skills, and digital infrastructure.

That's your Monday Wrap for June 1. Back next Monday with the week's freshest signals.

Subscribe now

On May 15, leaders of the Turkic world will gather in the city of Turkistan, Kazakhstan. The agenda includes artificial intelligence, digital transformation, and innovation. Most Western analysts will not even be aware that this summit is taking place.

But they should be.

All eyes are currently on the Gulf. Iranian drones struck data centers in the UAE. Submarine cables were marked as target coordinates. For the first time in history, the Strait of Hormuz became effectively closed not only to commercial shipping but also to data traffic. This situation has simultaneously called into question the three assumptions underpinning trillion-dollar AI infrastructure investment strategies: cheap energy, political stability, and the U.S. security umbrella.

Risk is being repriced. And when risk is repriced, investors search for alternatives.

That alternative is the Middle Corridor.

From Trade Route to Data Route

The Middle Corridor, connecting Türkiye, Azerbaijan, Georgia, the Caspian Sea, and Central Asia, has nearly doubled its cargo volume since 2022. Azerbaijani President Ilham Aliyev proudly emphasized this figure at last year’s summit of the Organization of Turkic States. But what remained unsaid is this: the Middle Corridor is no longer merely a trade route, it has the potential to become a data route.

Ninety-seven percent of the world’s internet traffic passes through submarine fiber-optic cables. A critical portion of these cables runs through the Red Sea and the Strait of Hormuz, both now active conflict zones. The land-based alternative route, extending across Türkiye, Azerbaijan, Georgia, and Central Asia, partially exists. But it still lacks the capacity necessary to function as a genuine alternative.

And this gap represents precisely the opportunity now in the hands of the Turkic states.

The question the Turkistan Summit must answer is simple: does the OTS recognize this?

The OTS’s Digital Dilemma

The organization’s efforts regarding physical connectivity should not be underestimated. The Baku–Tbilisi–Kars railway, Trans-Caspian routes, and the negotiations surrounding the Zangezur Corridor are all tangible infrastructure investments. Kazakhstan has proposed the establishment of a shared AI-supported digital logistics center that would integrate national infrastructure plans. The Turkic Investment Fund aims to reach a capital base of $1 billion by the end of 2026.

But logistics and data are not the same thing.

Transporting containers is one thing; routing global internet traffic is another. The Gulf crisis revealed just how dependent global data flows are on narrow maritime chokepoints. A genuine overland alternative requires fiber infrastructure, data center capacity, and shared technical standards. None of these yet sufficiently exist.

Closing this gap is both an infrastructure challenge and a matter of political will.

What to Watch on May 15

First: pay attention to concrete digital infrastructure commitments. Not strategy documents, but actual projects, cables, data centers, and connectivity infrastructure. Declarations are easy; laying fiber is hard.

Second: watch Azerbaijan. As the current chair of the OTS, Baku sits at the western end of the Middle Corridor. President Ilham Aliyev has positioned Azerbaijan as the logistics hub of the Turkic world. But does Baku also see itself as a data transit hub? The messaging emerging from Turkistan will clarify this.

Third: pay close attention to the language of AI sovereignty. The lesson learned by the Gulf was this: when you anchor your AI infrastructure to a great power, you also inherit that power’s enemies. The UAE’s AI company G42 found itself on the IRGC’s target list because it aligned with Washington. Most OTS members pursue strategic balancing policies. That gives them a different option, but only if they consciously embrace it as such.

The global map of AI infrastructure is being redrawn in real time. The Gulf was expected to become the world’s third major center of computational capacity after the United States and China. That story is not over, but it has become far more complicated. The corridors connecting data centers to users, the cables carrying global internet traffic, the physical layer beneath artificial intelligence, all are now exposed to the same geopolitical pressures that have always shaped oil, gas, and trade routes.

The Turkic world stands directly atop these alternatives.

May 15 will reveal whether its leaders truly understand what they hold in their hands.