At MWC Barcelona 2026, Dr Deng Bi, President of Huawei’s Oil and Gas Business Unit, set out the company’s strategy for heavy, resource intensive industries. The claim is straightforward: industrial AI has matured to the point where it delivers measurable financial returns.
The company’s approach rests on an ‘AI + Connectivity’ foundation aimed at core production scenarios, with the company positioning itself as a technology partner to help shorten R&D cycles and solve specific operational problems – improving safety, supporting lower‑carbon operations, and lifting productivity.
Deng argues that, after years of policy‑driven pilots and proof‑of‑concepts, 2026 marks a shift in how industrial AI is adopted, with customers now rolling out systems that have clear commercial value, and with many projects generating benefits that exceed their costs.
The role of AI is changing
Once seen only as a support tool, AI is now moving into complex, core production processes in heavy industry – such as seismic analysis.
For example, as geological targets become more complex, traditional seismic imaging and interpretation are running into efficiency and accuracy limits. Huawei’s AI platform is designed to address these bottlenecks by streamlining data processing and automating key interpretation tasks. In practice, this has cut seismic analysis cycles from months to weeks, significantly improved fault identification, and sharply reduced the time needed to interpret well data – helping exploration teams move faster and make more confident drilling decisions.
For Deng, such projects show that AI has moved beyond peripheral support tasks and is now shaping the core economics of production. This underpins Huawei’s value‑driven construction model: instead of broad, upfront infrastructure spending in the hope that useful applications will emerge, ICT is deployed to serve AI projects that have already proven their business impact. In practice, this means starting from specific, validated scenarios on the factory floor or in the field and letting those requirements shape how and where networks and computing are built.
Q&A with Dr Deng Bi:
For a traditional, asset-heavy industry like O&G, has AI already become a core productive force?
D: In the past, AI was mostly supplementary, performing tasks like simple monitoring or office collaboration. However, we now see AI transitioning from a supporting role to driving core production. In the field of exploration, AI is reconstructing the workflows for seismic data processing. In the production phase, foundation models are optimising process parameters in real time. The hallmark of this shift is that the benefits generated by technology clearly outweigh the investment costs. Moving forward, this trend will continue to push the industry toward deeper ‘systematic intelligence.’
During digital transformation, many energy companies fall into the trap of ‘building for the sake of building’, leading to low infrastructure utilisation. What is Huawei’s advice to avoid this?
D: Our core philosophy has always been ‘driving construction through application.’ Digital transformation is not simply about piling up servers or network equipment; it must be guided by the actual value created by AI to drive infrastructure construction.
The underlying logic is simple: first, identify high-value business scenarios (such as seismic data processing or complex reservoir simulation). By solving these specific pain points, we identify the required computing power, network, and data governance architecture. A digital and intelligent foundation built this way is deeply intertwined with the business and capable of truly supporting the enterprise’s long-term development.
What measurable efficiency gains can Huawei’s ‘AI‑driven core replacement’ deliver in exploration and production?
D: In the exploration field, data growth is exponential, and traditional computing paradigms are indeed hitting bottlenecks. We have introduced a brand-new architectural design in our E&P Computing Power Centre Solution.
Take Full Waveform Inversion (FWI) for seismic data as an example, which is an extremely compute-intensive process. In our collaboration with BGP (Bureau of Geophysical Prospecting) of CNPC, we deeply integrated neural network technology with geophysical technology. By utilising massive seismic exploration data to train an AI foundation model for seismic interpretation to locate oil and gas, processing efficiency was improved by 2 to 3 times, shortening processing times from months to hours.
Furthermore, in specialised algorithm scenarios like PSTM (Pre-Stack Time Migration), performance has significantly improved compared to traditional solutions. This not only means faster speeds but also higher reservoir proven rates, making ‘unclear’ geological strata ‘clear.’
Additionally, Huawei and CNPC and other partners are jointly developing a ‘Smart Drilling System,’ utilising deep learning algorithms to identify lithology in real time. This increased the reservoir drilling encounter rate to 85% and boosted single-well production by 30%, while shortening the drilling cycle by 15% and substantially reducing drilling costs.
Huawei and CNPC’s Changqing Oilfield launched an integrated training and inference architecture. Relying on the Pangu and Kunlun foundation models, they built the industry’s first CV model for hot work. This enhances intelligent and effective safety management for special operations like hot work in smart oilfields, addressing pain points such as oil leaks during operations. Compared to small models, the CV hot work large model improved inspection efficiency by 25% and reduced the false alarm rate by 8%; it shortened the foundation model development cycle and O&M costs by over 18% and reduced the scenario replication cycle by 40%.
Communication in offshore oil and gas fields is highly challenging – what breakthroughs has Huawei made in offshore scenarios?
D: In regions like Latin America and Southeast Asia, we have supported customers in upgrading the networks of their offshore oil platforms. By upgrading the fibre-optic backbone network, we achieved a connection rate of 200Gbps; this not only resolved the backhaul issue for massive sensor data but also built an information highway for offshore platforms.
Meanwhile, in regions like the Middle East, we are supporting O&G companies in building 450MHz industrial 5G private networks. Unlike carrier mobile communication networks, industrial 5G networks are explicitly designed to support mission-critical communications for production operations and advanced industrial internet applications. Specifically, industrial production requires high uplink speeds for massive sensors, compatibility with hundreds of industry protocols, achieving zero packet loss in mobile scenarios, and ensuring the highest levels of security and availability.
As the ecosystem develops and matures, the 450MHz industrial network will cover a wider range of industry scenarios, such as smart mobile inspections, environmental monitoring, and smart public facility management.
Many vendors are promoting foundation models, but O&G operators remain concerned about data security and AI ‘hallucinations’. How do you ensure AI is reliable in industrial settings?
D: First, we focus entirely on industry-specific scenarios. Our models are trained on massive amounts of industrial data and are integrated with industry mechanism models to strictly constrain AI hallucinations. For example, in the fault diagnosis of natural gas compressors, the accuracy of our prediction model already exceeds 90%.
Second, regarding deployment architecture, we support diverse hybrid deployment models where core data can be closed loop locally, ensuring absolute data security. We are building AI to industrial-grade standards, not simply putting a chatbot into a factory.
Compared to oilfield service (OFS) companies, what is Huawei’s competitive advantage in the O&G industry?
D: Our advantage lies in our focus on building the ‘AI + Connectivity’ foundation. OFS companies excel in geology and operational processes, while we excel in transmitting data quickly, computing it accurately, and storing it reliably. This is a highly complementary relationship. For instance, we have collaborated with mainstream OFS companies to complete the adaptation of over 20 mainstream seismic data processing software applications. We act more as the builder of a ‘digital foundation,’ enabling professional O&G applications to run faster and more stably on this foundation through an open ecosystem.
