- Huawei has unveiled what it calls the industry’s first 7.2-terabit near-packaged optical engine
- The product follows an emerging specification—but interoperability has not yet been demonstrated
- Huawei’s global reach could allow it to create a market without waiting for the rest of the industry
If a tree falls in a forest and no one is there to hear it, does it make a sound?
And if Huawei builds the world’s first 7.2-terabit near-packaged optical engine but nobody else has produced anything for it to interoperate with, is it interoperable?
Technically, no. Commercially? It may not really matter.
Huawei has unveiled what it describes as the industry’s first 7.2T near-packaged optics, or NPO, engine. It is designed to move enormous volumes of data between the processors inside AI infrastructure while reducing the power and delay that were part and parcel of conventional pluggable optical modules.
The company says the engine supports 36 channels operating at 200 gigabits per second, producing 7.2 terabits of total capacity. That's a lot.
Huawei also claims approximately 66% lower power consumption and 90 percent lower latency than conventional pluggable optics. That's a little.
But the more important story is where Huawei has put the optics. AI processors need to exchange data at mind-boggling speeds. As clusters become larger, the electrical connections carrying that data across circuit boards become shorter, hotter and more power-hungry. NPO moves the optical engine away from the front of the switch and much closer to the processor handling the traffic.
In plain English: electricity does less of the traveling, and light does more. That's important because the next AI bottleneck may not be the availability of processors; it may be the power required to connect them.
Huawei calls NPO “the most mature and pragmatic technology path for AI infrastructure,” arguing that it offers a lower manufacturing threshold and an easier route to mass production than fully co-packaged optics, where the optical components are integrated even more tightly with the switching silicon.
That is the technology story. The standards story is a bit more complicated.
Standard-ish
Huawei’s 7.2T design is associated with work taking place within the International Photonics and Electronics Committee, or IPEC. But 7.2T remains an emerging industry specification rather than a universally adopted, completed standard supported by a mature range of interchangeable products.
Other groups are pursuing related approaches. The Optical Internetworking Forum is developing electrical interfaces that support NPO and co-packaged optics, while the Open CPX MSA is working on common specifications for high-capacity optical engines.
Ciena told FNTV that it is “a founding member of the Open CPX MSA,” which is developing standards for CPO and NPO applications. Ciena’s own Vesta 200 optical engine delivers 6.4T of capacity and, according to the company, can reduce power consumption by as much as 70 percent.
Huawei is first in this case, and its achievement demonstrates its technical prowess. Then again, after increasing its R&D spending to CNY 192.3 billion—approximately $27.8 billion—in 2025, equivalent to 22 percent of its annual revenue, was that ever really in doubt?
Huawei also acknowledges that vendors are taking different technical approaches. During its announcement roundtable, a company representative told FNTV: “This kind of dispute or disagreement is very common in standardization.”
The spokesperson added: “I firmly believe that we will find an appropriate approach … to achieve interoperability.”
The operative word is achieve.
Huawei may have built its product around an emerging industry specification. But compliance with a specification and demonstrated interoperability are not the same thing. Until another vendor produces a compatible 7.2T implementation and the two are tested together, interoperability remains an objective rather than an established fact.
That is not a criticism of the engineering. Somebody has to be in the vanguard.
And when the company going first is Huawei, the absence of an immediate interoperability partner may be almost moot.
A market of one
Huawei is the largest optical-networking vendor in the world.
According to Dell’Oro Group, Huawei held approximately 33% of the global optical-transport equipment market in 2024, compared with 19% for second-place Ciena. Huawei also says it operates in more than 170 countries and regions.
No other optical vendor combines that market share with comparable geographic penetration.
These two facts change the meaning of “ecosystem.”
For a smaller supplier, launching a new optical architecture without multiple confirmed interoperability partners could be commercially dangerous. Customers might hesitate to adopt a technology that could leave them dependent upon one vendor.
Huawei already has an enormous installed base, extensive manufacturing capacity and customer relationships spanning most of the world. It is large enough that when it moves a technology forward, it can create a de facto industry standard and a substantial global market by itself.
The absence of interoperability partners therefore does not make the 7.2T announcement irrelevant. It makes the standards battle more consequential.
If Huawei deploys the technology at scale, particularly across China, Asia, the Middle East, Africa and Latin America, other suppliers may eventually have to decide whether to interoperate with Huawei’s implementation, advance a competing specification or risk surrendering a large part of the addressable market.
Standards are supposed to create markets. Occasionally, markets create the standards.
Geopolitics enters the tech chat
And in the topsy-turvy 21st century, interoperability is no longer a purely technical concern; geopolitics has entered the tech chat.
The Trump administration is drafting a measure that could prohibit imports of new Chinese optical transceiver models into the United States. The Federal Communications Commission says the proposal is intended to prevent potentially insecure Chinese components from becoming embedded in American AI data centers.
At a “how technology works” level, that doesn’t make any sense at all—but more importantly, such restrictions would also carry costs.
Chinese manufacturers supply a large share of the global optical-transceiver market, and replacing that production capacity would take time. Reuters reported that a ban could increase costs for American cloud providers and force them to transition to suppliers including Coherent and Lumentum. The proposed measure could still be modified or abandoned.
It is also unclear whether Huawei’s NPO engine would be covered.
Huawei’s placement on the U.S. entity naughty list principally restricts the export, re-export and transfer of controlled American technology to Huawei. It is not, by itself, a comprehensive prohibition on Americans importing every Huawei product.
When Huawei was asked whether the product could be affected, its spokesperson declined to comment.
The commercial implications are nevertheless clear. If Washington closes the American market to new Chinese optical components, Huawei will continue developing and selling them elsewhere. Its presence in more than 170 countries and regions gives it a global commercialization path that does not depend upon U.S. adoption.
The result could be two increasingly separate optical ecosystems: one organized around American security requirements and allied suppliers, and another built around Chinese scale, manufacturing and standards influence.
That would make interoperability harder for everyone. But the most immediate damage could be to the American domestic AI buildout.
The import-export paradox
America’s developing AI trade policy also shows a striking inconsistency.
Washington is considering restricting Chinese optical components—even though American AI data centers depend heavily upon Chinese manufacturing and alternative suppliers cannot replace all that capacity overnight.
That could raise costs or slow parts of the U.S. AI buildout.
Meanwhile, the United States has authorized limited, licensed exports of Nvidia’s powerful H200 AI chips to approved Chinese customers. Small shipments had begun by July, although the exports remain restricted and closely monitored.
Those exports could help approved Chinese companies expand their AI capabilities.
Taken together, the policies create an import-export paradox: restricting some of the optical plumbing required by American AI infrastructure while permitting China limited access to substantially more powerful American computing technology.
There may or may not be national-security arguments for both decisions. But their combined industrial effect deserves scrutiny.
Huawei’s 7.2T engine sits directly inside that contradiction.
It is a technical achievement, an emerging standards play and a reminder that excluding Chinese technology from the United States does not remove it from the global market.
Huawei already leads that market. It already operates at a scale its competitors cannot easily reproduce. And it already reaches customers across most of the planet.
The immediate question is whether Huawei’s 7.2T engine can interoperate with another supplier’s implementation. But the larger question is whether the industry is heading toward a future in which that matters less — because Huawei has sufficient scale to build a global optical ecosystem of its own — and whether the U.S. is inadvertently creating a regulatory environment that helps it to do so.