Le dernier rapport Intelligent World 2035 de Huawei propose 10 axes stratégiques

Huawei’s Intelligent World 2035 report forecasts annual AI-token consumption will increase 100,000-fold by 2035, with agent-generated traffic exceeding 90% of total token usage. The company outlines 10 strategic priorities spanning AGI, 100x larger compute clusters, SuperPoD systems, intelligent connectivity, agent operating systems, autonomous-driving safety, data-center energy efficiency, and agent security. Separately, Huawei and Tsinghua University project AI will create more than $27 trillion in cumulative 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 a near-term earnings catalyst; it is vendor-authored demand signaling and should not be extrapolated into incremental orders without hyperscaler capex, optical-component backlog, or carrier procurement evidence. The investable implication is nevertheless that agentic workloads shift the bottleneck beyond accelerators: persistent inference raises east-west traffic, memory/storage I/O, power density and cooling requirements. That favors Arista (ANET), Vertiv (VRT), Eaton (ETN), Broadcom (AVGO), and storage/networking beneficiaries more than a further indiscriminate rerating of GPU exposure.

Over the next 1-3 months, the relevant confirmation points are AI-cluster order commentary from ANET/VRT/ETN and evidence that inference networking is becoming a distinct budget line rather than bundled capex. Over 6-18 months, autonomous-agent deployment should increase spending on machine identity, access control and auditability, supporting CyberArk (CYBR) and Palo Alto Networks (PANW); however, enterprise adoption can be delayed by liability, data-governance and reliability constraints. The contrarian view is that the market may be overpricing a linear compute boom while underpricing power availability and network utilization: constrained grid interconnects can defer data-center revenue even when chip demand remains strong.

Key Decisions for Investors

• Maintain a 6-12 month overweight in ANET versus NVDA as a pair trade, sized modestly: agentic inference increases network intensity while NVDA faces higher expectations and greater sensitivity to any capex digestion. Exit the relative-long thesis if ANET's AI-related revenue commentary fails to accelerate for two consecutive quarters or if hyperscalers signal lower network buildout.
• Accumulate VRT and ETN on 8-12% pullbacks for a 12-18 month infrastructure basket; the key upside is operating leverage from higher rack power density, while the principal risk is project deferral from utility interconnection delays. Use quarterly data-center backlog conversion and margins—not headline AI demand—as the validation metric.
• Place CYBR on a 3-6 month watch for an entry following enterprise AI-security product launches or bookings disclosures. Do not initiate solely on this report: the required confirmation is measurable growth in machine-identity or privileged-access demand, and the thesis is falsified if AI deployments remain isolated pilots rather than production workflows.
• Avoid treating Huawei's long-range projections as a direct semiconductor demand signal for US-listed suppliers. Geopolitical restrictions, local substitution and uneven access to advanced compute make the read-through to NVDA, AMD and US networking vendors highly uncertain.