Features writer
Jeffrey Gogo
Features writer
Share
Fact Checked by
Elena Bozhkova
Features Lead
Key Takeaways:
- AI data centers to become just as important as oil reserves for governments.
- For operators, AI spending will move from a cost to a “yield engine.”
- Stablecoins, tokenized Treasuries won’t replace existing banking system but will handle a large share of cross-border and institutional flows.
Governments are expected to start treating artificial intelligence (AI) data centers and energy-backed computing power as strategic infrastructure in 2026, similar to how oil reserves are managed, according to analysts.
It comes as competition over the resources needed to run large-scale AI systems intensifies. Demand for compute from companies like OpenAI is outpacing supply, and energy constraints are starting to limit growth.
“Governments won’t just subsidize data centers,” Maja Vujinovic, CEO of Digital Assets at FG Nexus, told Cryptonews. “They will begin treating GPU capacity and energy-backed compute as strategic infrastructure, similar to oil reserves.”
Vujinovic said countries with cheap energy, access to advanced chips and stable policies will move first. Some governments are likely to launch “national compute-credits” that allocate access to GPU capacity, she said.
“Corporations will lock in multi-year power deals the way airlines hedge fuel. The geopolitical competition shifts from, ‘Who has the best AI models?’ to ‘Who can secure the most power and compute to run them?’”
Asked which regions will likely become the “OPEC” of this new era, Vujinovic said, “There won’t be a single OPEC-style group per se.” Instead, “power concentrates where chips, energy, and execution speed come together.”
AI-denominated Yield
The U.S is expected to remain dominant in large-scale AI workloads while the UAE and Saudi Arabia are also becoming important, she said, “because they can align capital, power, and permitting faster” than Western economies.
Taiwan and South Korea are critical nodes due to their role in advanced chip manufacturing, though geopolitical uncertainty clouds long-term outlook, Vujinovic added, telling Cryptonews:
“The key difference from oil is speed. Compute capacity can move and scale much faster than physical commodities. So, while concentration is real, it’s more dynamic and competitive than traditional resource cartels.”
The rise of compute as infrastructure will also likely see the emergence of a new financial category in 2026: AI-denominated yield.
Experts say most firms currently treat AI spending largely as a cost. But this year, it will become a so-called “yield engine”.
“The first wave of ‘AI income products’ will appear, revenue-sharing instruments backed by inference usage, fine-tuning demand, or compute resale,” Vujinovic predicts.
“Funds, treasuries, and institutions will treat AI workloads like digital export products with predictable cash flow. AI capacity becomes an investable stream just like data storage or cloud SaaS was a decade ago.”
Standardizing the yield will rely on familiar metrics, including GPU-hours, performance, uptime and reliability, she says.
Revenue can then be reconciled directly against those metrics, supported by tamper-resistant logs, third-party audits and hardware-level verification that workloads ran on approved infrastructure..
“The open question isn’t whether this model works,” Vujinovic said, “but how quickly auditors and buyers treat AI workloads like any other revenue-generating infrastructure asset.”
Ryan Li, CEO of crypto-focused AI Surf, said, in 2026, artificial intelligence will “feel less like a standalone product and more like something embedded directly into the crypto tools and apps people already use.”
“The shift will be from manually prompting AI to having it help complete complex tasks with little human intervention,” Li, whose company recently raised $15 million in a funding round led by Pantera Capital, told Cryptonews.
Interest in domain-specific AI has risen sharply over the past year. Li said while general-purpose artificial intelligence platforms like ChatGPT still lead in name recognition, “more specialized systems such as Claude for coding or Manus for long-tail tasks are gaining traction.”
That’s because “they move beyond text generation and perform specific jobs better,” he detailed. “Specialization matters more in crypto because it’s a noisy and complex environment where real money is at stake.”
“AI tools…automate manual processes in existing workflows and enable entirely new workflows that were not possible before. By the end of 2026, it will be normal for the majority of crypto users to interact with AI in some form, and many will benefit from it without even thinking about it.”
Tokenized Dollars Move Into Traditional Banking Turf
Meanwhile, dollar-pegged stablecoins, like Tether’s USDT and Circle’s USDC, and tokenized U.S. Treasuries are fast becoming a parallel settlement layer for global finance, according to analysts.
They say the products won’t replace the existing banking system, but they “will handle a meaningful share of cross-border and institutional flows at the margin.”
Stablecoins now move more value annually – estimated at $33 trillion in 2025 – than Visa, and tokenized Treasuries have grown into a multi-billion-dollar market in just a few years, according to industry data.
FG Nexus’ Vujinovic expects that, in 2026, the inflection point is likely to be psychological rather than technical.
“For a certain class of CFOs, funds, and fintechs, using on-chain dollars for settlement and liquidity stops being crypto and just becomes normal, faster plumbing.”
Vujinovic said adoption grows once on-chain money settles in minutes and not days, offers real-time visibility and integrates with existing treasury systems through regulated custodians.
“Once funds move in minutes instead of days, reconciliation is automated, and regulated custodians sit underneath, on-chain starts to look operationally safer than traditional correspondent banking,” she said.
Even if stablecoins take on more settlement volume, traditional banking infrastructure will continue to play a central role, analysts say. Networks such as SWIFT will not disappear, but their function may change.
“SWIFT’s value has never really been about moving money,” Vujinovic said, “It’s about coordinating trust between institutions at a global scale.”
As settlement rails diversify, SWIFT’s role will move toward standards, compliance and governance. “It becomes less about being the rail itself and more about being the layer that connects regulated institutions across different rails,” she stated.
“Settlement technology can change quickly. Global trust and compliance infrastructure changes much more slowly, and that’s where incumbents continue to matter.”
Bounded Autonomy for AI Agents
We asked Mau Ledford, cofounder and CEO of Sogni AI, which components of the artificial intelligence stack he expects to be verifiable on-chain this year.
“In 2026, blockchains won’t store AI,” he replied, “they’ll store the receipts. You’ll be able to prove which model ran, on whose compute, and who got paid, while the actual data and outputs stay private and off-chain.”
He added that decentralized GPUs will move from “spare capacity to a real inference layer, especially for creative AI.” However, “they only win if reliability and orchestration feel as smooth as the cloud,” he averred.
Ledford also expects that AI agents will “absolutely” have wallets in 2026. “They [AI agents] will operate with guardrails, like junior employees with spending limits, audit logs, and kill switches,” he told Cryptonews.
His comments were echoed by Ben Goertzel, CEO and chief scientist at SingularityNET, who said autonomous AI agents are already operating as “first-class” economic participants in “small ways.”
“We will have AI agents with wallets buying data, paying for compute, and negotiating simple contracts,” Goertzel, who is also the CEO of Artificial Superintelligence (ASI) Alliance, told Cryptonews, adding:
“We still need better identity, reputation, spending limits, and security so agents do not get hijacked or behave unpredictably. So for 2026 it will be bounded autonomy, not full self-sovereign AI citizens…”
He predicts that crypto-native AI will challenge hyperscalers “by opening access and letting many participants build, run, and own pieces of the AI stack.”
“We are not eliminating centralization completely. We are balancing it by building a plural, decentralized AI ecosystem instead of one owned by a few giants,” Goertzel noted.
