In the global race for artificial intelligence, the conversation usually revolves around chips, models, data, and engineering talent. But when it comes to building real AI infrastructure, the first question investors ask is far more practical: can a country provide enough reliable electricity — and can that reliability be secured under a long-term contract?
On August 24, at the Silk Road Finance & Technology Forum in Tashkent, Bobur Khodjaev, Head of the Department for Financial Technologies, Digitalization and Artificial Intelligence at the Presidential Administration of Uzbekistan, addressed this issue directly. Companies considering the construction of data centers in the country, he said, identify the reliability of power supply as one of their first concerns. According to Khodjaev, renewable generation already accounts for more than 30% of the country’s power mix, while the government is deliberately focusing on Karakalpakstan as a location for data centers because of its strong solar and wind potential.
This is no longer merely a technical issue for the energy sector. Electricity is becoming one of the key resources in global technological competition. It is no longer enough for countries to offer tax incentives, skilled labor, or access to markets. They must also be able to guarantee large volumes of power at a predictable price, with sufficient reserve capacity and uninterrupted 24/7 availability.
Megawatts Are Becoming Part of Investment Policy
The following day, on August 25, Uzbekistan commissioned three new power plants and three battery energy storage systems (BESS) with a combined capacity of 2,423 MW. The projects are valued at approximately $1.7 billion. Although their launch was timed to coincide with the 35th anniversary of Uzbekistan’s independence, their economic significance extends well beyond the ceremony itself.
The largest facility is a 1,573 MW combined-cycle gas power plant in the Bayaut district of the Syrdarya region. This is conventional gas-fired baseload generation rather than a renewable project. Yet facilities of this kind are precisely what can cover nighttime and peak demand — something without which a data center cannot operate reliably. In the Navoi region, a 300 MW solar power plant was launched alongside a 150 MW storage facility. In Karakalpakstan, a 200 MW wind farm was commissioned together with a 100 MW storage system. Another 100 MW BESS facility was launched in the Namangan region.
These projects are not being developed solely because of AI. Electricity consumption is rising alongside industrial activity and household demand. Uzbekistan’s total installed power capacity has reached 25.8 GW, of which 8 GW, or 31%, comes from solar, wind, and hydropower. In 2026, the country plans to increase electricity generation to 90 billion kWh — 40% more than in 2020.
Alongside traditional consumers, however, a new competitor for electricity is emerging: digital infrastructure. President Shavkat Mirziyoyev has separately instructed the government to expand computing capacity and increase the number of applied AI projects. The authorities had previously reported more than 200 such initiatives across different sectors of the economy.
It is important not to conflate two very different challenges. National AI projects in education, healthcare, or industry require entirely different amounts of electricity from a cluster designed to train large-scale AI models. It is the second type of consumer that changes the investment logic: it is not simply buying “market access” but guaranteed megawatts.
Uzbekistan Is Selling More Than Incentives
The shift in this model is particularly visible in Karakalpakstan. A special regime has been introduced there for artificial intelligence and data center projects. Companies investing more than $100 million are offered IT Park incentives, exemptions from customs duties on imported equipment, and a special electricity tariff of 5 cents per kWh. By comparison, the standard tariff for legal entities had previously been quoted at around 8.3 cents per kWh. The government is also covering part of the cost of road, electricity, and water infrastructure.
This is a different kind of investment proposition: not an isolated tax benefit, but a package combining land, power, infrastructure, and a special regulatory regime. Investors assess the entire package — grid access, the price and reliability of electricity, water for cooling, telecommunications, the ability to reserve capacity, and the speed of regulatory approvals.
Investor interest is already translating into concrete agreements, although their status needs to be described accurately. In Karakalpakstan, a framework investment agreement worth $130.9 million has been signed for an AI data center involving Taiwan’s Keycore Technology. It is an investment commitment and framework, not an automatically completed or commissioned facility. Other projects are developing in parallel: Saudi Arabia’s DataVolt is building the 12 MW TAS-1 data center in Tashkent, with the first phase expected to be completed by the end of 2026; broader discussions on financing digital and energy infrastructure are also under way with Korean institutions, including Korea Eximbank.
The issue of power supply reliability raised by international companies is therefore anything but theoretical. Uzbekistan is already competing for capital for which access to megawatts is a genuine barrier to entry.
What Investors Are Actually Buying
New generation capacity used to be justified primarily by the needs of households, industry, and national energy security. Technological competitiveness must now be added to that list.
A data center can be located in many different jurisdictions. Investors therefore look beyond the regulatory regime and tax environment. They want to know whether the grid can withstand constant demand, whether spare capacity exists not simply “somewhere in the country” but at the specific substation serving the site, whether a long-term power purchase agreement can be concluded, where the electricity will come from, and whether sufficient water is available for cooling.
The rapid spread of generative AI is only intensifying these requirements. Every new model and every new computing cluster creates demand that does not disappear at night and cannot tolerate hours-long outages. In this context, a tariff of 5 cents per kWh is only one line in a financial model. Without service-level guarantees, reserve capacity, and transparency over the source of electricity, cheap power remains little more than a marketing figure.
That is why the simultaneous commissioning of gas-fired generation, renewables, and battery storage makes more sense than relying on solar power alone. BESS does not replace power generation. It smooths demand peaks and gives the system additional time and flexibility to respond. For AI workloads, this is critical.
Central Asia Has an Opportunity — but Not an Automatic Advantage
The region has something many mature markets increasingly lack: abundant solar and wind potential, available land, and competitive energy costs. If these advantages are combined with modern grids, storage systems, fiber connectivity, and predictable rules for investors, part of the world’s digital infrastructure could move here rather than being concentrated solely in Northern Europe or the Middle East.
This could gradually change the role of a region that is still often perceived globally as either a supplier of raw materials or a transit corridor. Data centers and computing clusters belong to a higher-value-added segment of the economy. But potential does not translate into competitive advantage on its own.
Cheap electricity offers limited value if it cannot be guaranteed around the clock. Karakalpakstan offers wind, sun, and space, but it is also a region facing considerable pressure on its water resources. Data centers consume not only electricity but also water for cooling — unless operators invest in more expensive dry-cooling or liquid-cooling systems. The special electricity tariff also needs to be explained in full: who compensates for the difference, how long the rate will remain in place, and whether the subsidy is ultimately transferred to other consumers. Otherwise, an incentive may look attractive in a presentation but prove fragile during due diligence
Grid infrastructure is another issue. New gigawatts of generation have limited value if transmission remains the bottleneck. Investors do not buy a country’s total installed capacity — they buy electricity that can actually be delivered to a specific site.
A Test, Not a Slogan
In the coming years, the public debate around artificial intelligence will continue to focus on chips, models, data, and talent. But for a government seeking to attract this industry, the test is both simpler and more demanding.
Uzbekistan has already assembled many of the elements of such an investment proposition: new generation capacity, battery storage, a special regulatory regime, and a clear political priority for AI infrastructure. This is a more substantial step than most countries in the region had taken only a few years ago. The next question is not how many megawatts are announced, but whether those megawatts can be converted into contracts that international investors are prepared to sign.
The competition for AI investment is increasingly becoming a competition for guaranteed megawatts. Not promised megawatts — delivered ones.
Alona Lebedieva, owner of the Aurum Group industrial and investment group