Huawei has moved beyond catching up to Western AI infrastructure and is now shaping the international standards that define how the world's largest AI data centers will operate. At its flagship HUAWEI CONNECT 2026 conference on Wednesday, the Chinese semiconductor company announced it would deliver its next-generation Ascend 960DT training chip in the first quarter of 2027, three full quarters ahead of its previously published roadmap. More significantly, Huawei revealed that it had already initiated and won approval for an international optical networking standard that will govern how future AI supercomputers transmit data.

What Makes Huawei's Move So Significant?

The timing reveals a strategic shift in how technology leadership works. Huawei initiated the Near-Package Optics (NPO) standards project with the Optical Internetworking Forum (OIF), a 170-member international consortium that has governed optical networking interoperability since 1998, back in May 2026. The OIF approved the project at its Q2 plenary meeting in Malta, formally launching a new 12.8-terabit-per-second NPO standards initiative. Only after securing this international standard did Huawei announce the Ascend 960 SuperPoD, the first product built on that standard architecture.

This sequence matters because it locks in a structural advantage. Every competing vendor that adopts the OIF 12.8 Tb/s NPO standard will be implementing, in some form, a design Huawei brought to the standards body first. As one analysis of the announcement noted, this is how platform advantages get locked in, particularly for a company operating under U.S. sanctions and without access to the world's most advanced fabrication equipment.

How Does Near-Package Optics Actually Work?

Near-Package Optics is a class of optical interconnect architecture that sits between traditional pluggable modules and the most integrated form, called Co-Packaged Optics (CPO). In a conventional AI data center, high-speed 800-gigabit optical modules insert into front-panel switch slots, and the signal travels across circuit board traces, sometimes 20 to 30 centimeters, to reach the accelerator chip. To compensate for signal degradation over that distance, digital signal processors and clock data recovery circuits sit in the path, retiming and re-amplifying the signal. Those retimer chips consume significant power and add measurable latency.

NPO relocates the optical engine from the front panel to within roughly 5 centimeters of the accelerator chip itself, mounted directly on the circuit board substrate near the chip area. Because the electrical path is now centimeters rather than tens of centimeters, the retimer chips can be eliminated, cutting power consumption and latency together. The trade-off compared to CPO is serviceability: NPO optical engines remain detachable and can be replaced without replacing the chip, while CPO integrations are permanent.

What Are the Real-World Efficiency Gains?

The specific arithmetic of Huawei's Hi-ONE optical module deployment for the Ascend 960 SuperPoD illustrates why this matters at scale. Replacing 48,000 conventional 800-gigabit optical modules with 5,500 Hi-ONE units cuts power consumption by over 550 kilowatts per SuperPoD, according to Huawei's figures. A 550-kilowatt reduction per SuperPoD is not a minor detail; it represents the difference between a 5-megawatt and a 4.45-megawatt data center power budget per cluster. At current U.S. commercial electricity prices of roughly $0.10 per kilowatt-hour, that translates to real operational savings at scale.

The Hi-ONE module achieves 7.2 terabits per second of single-engine transmission capacity, with a built-in light source that Huawei claims is an industry first for a mass-produced NPO product. Co-Packaged Optics, the technology that would eventually supersede NPO by integrating optics directly onto the chip package, is not expected to enter production at scale until at least 2028, according to industry forecasts. This timing gives Huawei a window to establish its NPO architecture as the dominant intermediate standard.

How Does This Fit Into Huawei's Broader AI Chip Strategy?

• Performance Roadmap: The Ascend 960 will deliver roughly double the compute of its 950-series predecessor, continuing what Huawei calls the Tau Scaling Law, a framework proposed earlier this year by He Tingbo, president of Huawei's HiSilicon semiconductor subsidiary. Under that trajectory, the Ascend 970 is planned for 2028 and the Ascend 980 for 2029, locking in one generation per year.
• System Scale: The Ascend 960 SuperPoD unveiled Wednesday connects 4,096 Ascend 960 accelerator cards into a single unified computing node, delivering what Huawei claims is up to 8 exaflops of FP8 compute and 1 petabyte of high-bandwidth memory capacity per SuperPoD.
• Market Timing: The TrendForce-projected Near-Package Optics and Co-Packaged Optics market size is expected to grow from approximately $100 million in 2025 to over $39 billion by 2030, and Huawei has positioned its Hi-ONE product at the entry point of that growth curve with an international standards project behind it.

Wang Tao, also known by his English name David Wang and Huawei's rotating chairman, opened the three-day event at the Shanghai World Expo Exhibition and Convention Center with a keynote titled "Intelligent Future: Building the Silicon-Based Soil for the Intelligent World." He announced that the Ascend 960DT, the training-optimized variant, is now scheduled for commercial availability in Q1 2027, with a second model, the Ascend 960PR optimized for prefill and recommendation inference workloads, following in Q3 2027.

What's the Catch With These Performance Claims?

Every performance figure Huawei announced comes from Huawei's own engineering team. No independent benchmarking organization has validated the doubling claim for the Ascend 960 series. The prior generation, the Ascend 950DT, was described by DeepSeek founder Liang Wenfeng in a July 2026 investor call transcript as trailing Nvidia's current hardware by roughly a four-to-one GPU equivalence ratio and approximately two years in development. That gap is real and should be the baseline assumption until independent auditors measure the Ascend 960 in production.

The international standards initiative, however, is not a performance claim. It is a structural fact. More than 40 companies are now participating in the NPO standard's development, including Broadcom, Alibaba Cloud, Tencent, Baidu, and CAICT, China's state-run information and communications technology research institute. On September 10, the International Photonics and Electronics Committee co-hosted an NPO Standards Industry Summit with OIF at the Shenzhen World Exhibition and Convention Center, in which all these participants reached NPO standards consensus.

What Huawei has accomplished is a shift in how infrastructure leadership is defined. Rather than simply building faster chips, it has positioned itself as the architect of the standards that all future competitors must follow. For a company operating under U.S. export controls and without access to the most advanced fabrication nodes, that is a more durable form of competitive advantage than raw performance metrics alone.

" }