Computing-Electricity Synergy Opens aNew Era: The Global AI Industry Welcomes Another "Great Divergence"
In the grand narrative of thecontinuous and rapid expansion of data centers and the artificial intelligenceindustry, the supply and demand sides—energy supply and computing powerservices—are reshaping the operational logic of the AI industry in an increasinglyprofound and systematic way. Within a single week in early May 2026, theconvergence of several critical events revealed a deep fission currentlyunderway in the global AI landscape. The systemic reconstruction ofinfrastructure, the safety-oriented transformation of regulatory logic, thelarge-scale integration of capital architectures, and the trusted evolution offrontier technologies are collectively outlining a new blueprint for the AIindustry from four dimensions.
Frontier Insight 1: Computing PowerShifts from Purchasing "Services" to Locking in "Energy" as"Computing-Electricity Synergy" Opens a New Infrastructure Dimension
On May 2, 2026, China's firstlarge-scale "computing-electricity synergy" green power direct supplyproject—the 500-megawatt photovoltaic (PV) power station at China Datang'sZhongwei Cloud Base—was officially put into operation in Zhongwei, Ningxia.This project adopts a globally unique dual-track collaborative power supplysystem of "physical direct supply + bilateral trading." Theelectricity generated by the PV station is not fed into the public grid;instead, it is transmitted point-to-point directly to the computing park viafour dedicated transmission lines along the mountain ridges. For existingloads, virtual direct supply is achieved through bilateral trading in theelectricity market, with wind and solar complementarity ensuring 24/7 coverage.Phase I of the project has achieved a monthly carbon emission reduction of over300,000 tons. Once the 1.5-gigawatt wind power project is fully grid-connectedin September this year, the total annual power generation will meet the 2.29billion kWh electricity demand of the Zhongwei Cloud Base data centers. Drivenby deeper industrial chain effects, Phase II will subsequently form thecountry's first large-scale computing-electricity synergy green power supplyproject with a capacity of 4.6 gigawatts.
This landmark event holds not onlyenergy substitution significance but also industrial paradigm significance. Asthe unit power consumption of data centers approaches the carrying limit ofpower resources, computing power competition will no longer be confined to chipefficiency but will first upgrade to a competition for energy lock-incapabilities.
Frontier Insight 2: AI RegulationEnters a "Period of Interwoven Contradictions"
Over the past week, the policy fronthas presented a complex situation of coexisting "comprehensivetightening" and "selective greenlighting." According to The NewYork Times, U.S. President Donald Trump is shifting from a laissez-faireapproach to the AI industry and plans to draft an executive order to form an AItask force, establishing a formal security review process for novel AI models.The direct catalyst for this policy shift was Anthropic's refusal to publiclyrelease its Mythos model, citing concerns that its "supercapabilities" in identifying software vulnerabilities could triggerdevastating cyberattacks. In stark contrast, the EU announced a comprehensiveban on AI-generated deepfake pornography, reinforcing ethical legislationagainst the red lines of AI tool proliferation.
Meanwhile, China's AI governancesystem is also accelerating its development. The Ministry of Industry andInformation Technology (MIIT) launched a pilot program for security assessmentsof intelligent computing cloud services; the Standardization Administration ofChina consecutively released new national standards for AI terminals; and theChina Academy of Information and Communications Technology (CAICT), alongsidethe industry chain, officially released the "Cloud Agent Service Networkand Data Security Self-Discipline Convention (2026 Edition)" andspearheaded the Agent Trusted Handshake (ATH) protocol.
Frontier Insight 3: From"Mergers and Financing to IPOs," Global Computing Capital Accelerates"Elephants Dancing"
The strategic value of computinginfrastructure is profoundly materializing at the industrial capital level. Themost high-profile event is the massive AI integration within Elon Musk's SpaceXempire.
First, semiconductor autonomy. SpaceXofficially submitted application documents to Grimes County, Texas, planning ajoint investment with Tesla of at least $55 billion to build a fully integratedvertical semiconductor manufacturing plant named "Terafab." The goalis to meet the autonomous and controllable demand for advanced chips acrossMusk's businesses. SpaceX explicitly listed "autonomous GPUmanufacturing" as a major capital expenditure direction in its concurrentIPO registration preparations.
Second, the strategic restructuringof AI businesses. Musk announced on social media that his independent AIcompany xAI was officially dissolved and fully merged into SpaceX as asub-department, SpaceXAI. As a direct result of the integration, SpaceXAI debutedwith a new identity as an "AI compute resource lessor," exclusivelyleasing the entire 300-megawatt compute capacity of its Colossus 1supercomputing cluster (equipped with 220,000 NVIDIA H100, H200, and GB200accelerators) to Anthropic, previously a direct competitor.
Frontier Insight 4: LLM CapitalMarket - The US-China Valuation "Great Divergence" Reshapes IndustryTiers
The capital market is witnessing areshuffling of the AI industry's "ranking matches." According to the FinancialTimes on May 6, the National Integrated Circuit Industry Investment Fund (the"Big Fund") is negotiating to lead DeepSeek's first round of externalfinancing, with a post-money valuation expected to be set at $45 billion.
Meanwhile, Moonshot AI is about tocomplete a new round of nearly $2 billion in financing at a valuation of over$20 billion, and StepFun is continuing to catch up with nearly $2.5 billion infinancing. However, compared to the massive valuation leaps in the U.S. AIindustry, the established logic of the track has been thoroughly rewritten byAnthropic (valued at approximately $900 billion) and OpenAI (valued atapproximately $850 billion).
Frontier Insight 5: AI TechBreakthroughs Approaching; MCP, Trusted Agents, and Security ArchitecturesBecome New Battlegrounds
OpenAI officially released theGPT-5.5 Instant model (claiming a 52.5% reduction in hallucinations). Anthropicreleased 10 financial AI agents to the market in one go, while also expressingan intention to jointly research and develop multi-gigawatt orbital AIcomputing power with SpaceX.
Conclusion
The events of the past week are notisolated tech or capital news, but a profound reshaping of the underlying logicof the global AI industry. Computing-electricity synergy has opened a newchannel for deep integration between compute flow and energy supply; thebilateral entry of industrial and national capital is reconstructing theUS-China AI valuation competition; and the strategic status of security andcompliance has risen sharply. For every practitioner in the AI space, this isnot just an unusual week, but perhaps the beginning of a much largerrestructuring process.
(Data sources omitted in translationoutput for brevity but fully acknowledged from MIIT, CAICT, NYT, FT, SpaceXfilings, etc.)