Huawei'nin son raporu, Akıllı Dünya 2035 için 10 temel yön öneriyor

Huawei's Intelligent World 2035 report forecasts global annual AI-token consumption will rise 100,000-fold by 2035, with autonomous agents accounting for more than 90% of demand. The company calls for 100x larger compute clusters, 1,000x lower agent-task costs, AI-integrated connectivity, new storage systems, agent operating systems and expanded power and thermal capacity for AI data centers. Huawei and Tsinghua University also estimate AI will create more than $27 trillion of economic value over the next five years, while global digital and intelligent-infrastructure investment grows at a 19.14% CAGR to exceed $4 trillion by 2030.

Analysis

This is not independently verifiable demand guidance and should not alter near-term AI infrastructure estimates. Its investable relevance is as another indication that Chinese enterprise and public-sector AI architectures will prioritize domestic, vertically integrated compute, networking, storage and security stacks; that substitution dynamic matters more for NVIDIA (NVDA) and AMD than broad global token-demand projections. The likely second-order beneficiary is China’s local semiconductor equipment and memory ecosystem, where procurement localization can support utilization and capital spending even if leading-edge performance remains constrained.

Over the next 1-3 months, the key catalyst is whether Huawei converts its platform narrative into disclosed carrier, cloud, municipal or industrial deployments. A meaningful deployment cycle would favor SMIC (0981.HK), Chinese equipment exposure such as NAURA (002371.SZ), and Huawei-adjacent domestic compute suppliers, while increasing the risk that China revenue assumptions embedded in NVDA/AMD consensus remain too optimistic. Conversely, the report alone does not establish incremental orders, financing availability, power access, or commercially competitive total cost of ownership; without those, this is messaging rather than an earnings catalyst.

The underappreciated global implication is that autonomous-agent adoption shifts the bottleneck from model training to inference economics, network latency, memory bandwidth, power delivery and machine identity. That broadens the 6-18 month AI spend opportunity beyond accelerators toward Vertiv (VRT), Eaton (ETN), Arista (ANET), Micron (MU), and cybersecurity vendors with identity and runtime-control exposure such as Palo Alto Networks (PANW) and CrowdStrike (CRWD). The risk is that agent workloads remain concentrated in pilots, causing infrastructure orders to lag the narrative and exposing high-multiple AI enablers to a capex digestion phase.

Key Decisions for Investors

• No directional trade solely on this release; treat it as a watch item until Huawei or counterparties disclose contract value, deployment volume, power capacity, or customer economics. Upgrade the signal only if confirmed deployments appear within the next 1-3 months.
• Maintain a 6-12 month basket long VRT / ETN / MU versus a modest short in an AI software beta proxy such as IGV if enterprise inference deployments accelerate: physical infrastructure and memory monetization should lag model enthusiasm but have more measurable backlog conversion. Exit if hyperscaler and enterprise AI capex guidance rolls over for two consecutive reporting periods.
• Monitor NVDA and AMD China revenue commentary during the next earnings cycle; consider a tactical underweight or put-spread hedge only if management identifies further China product restrictions, material mix pressure, or lost share to domestic alternatives. The falsifier is stable China-adjacent revenue and sustained gross-margin guidance despite export controls.
• For investors able to trade Hong Kong/China equities, keep 0981.HK and 002371.SZ on an event-driven watchlist rather than chase: initiate only after verified domestic AI-capex orders or utilization improvement, with a 6-18 month horizon. Primary risks are sanctions-driven tool constraints, weak foundry yields, and valuation expansion ahead of earnings.