With the specs complete, it's supposed to be full steam ahead for DOCSIS 4.0. But will the chipmakers that played a big role in the development of previous generations of DOCSIS technology come along for the ride?

For Broadcom, a key maker of DOCSIS silicon for modems, gateway and cable network equipment, its future role in DOCSIS products is as clear as mud.

According to multiple industry sources familiar with the situation, Broadcom has halted investment in next-gen DOCSIS technology and won't change that stance without financial support for DOCSIS 4.0 product development, as well as certain purchase assurances and commitments, from cable operators.

Broadcom, those sources say, wants the cable industry to support its DOCSIS 4.0 product development under a non-recurring engineering (NRE) model. That model would call for operators and perhaps others to help to foot those R&D costs, put more of their skin in the game and mitigate Broadcom's overall financial risk. Under the traditional model, Broadcom would bear most of those development costs and then hope that the resulting product would be bought in volume.

"There's innovation and one-time costs that go into the launch of any kind of product, and DOCSIS is no different," an industry source said, predicting that the NRE model is likely to surface more often with strategic product introduction initiatives in the cable industry. Cable operators "don't have a full appreciation of the cost to innovate and the lifecycle of the innovation and the return-on-investment required."

"They've been tossing it around," one industry source said of the NRE model. The source noted that it's also of interest to MaxLinear, the chipmaker that is now acquiring the home connectivity unit of Intel Corp., which makes Wi-Fi silicon as well as DOCSIS chips for cable modems and other types of cable CPE.

But Broadcom's pursuit of an NRE-based model "is not being well-received" by cable operators, another person familiar with the discussions said.

Once bitten, twice shy
Part of Broadcom's reluctance to invest in D4.0 without more financial assurances, another source said, stems from getting burned by its earlier investments in Full Duplex DOCSIS (FDX). FDX, originally an annex to DOCSIS 3.1 championed primarily by Comcast, is now being incorporated into the new DOCSIS 4.0 specs alongside another option called Extended Spectrum DOCSIS (ESD) that has gathered interest from most other cable operators in the US and abroad.

Multiple sources said Broadcom was part of a group of suppliers that developed FDX products for Comcast, but the operator deemed the results of those efforts too expensive. Instead, they say, Comcast ultimately decided to move in another direction, tapping Intel to run point on an FDX project that also involved an Israel-based software company called Capacicom. Comcast's near-term network priorities have shifted during the pandemic, but the MSO confirmed to Light Reading in recent weeks that the cable operator remains "fully committed" to FDX.

Still, Broadcom has "no appetite to rush into DOCSIS 4.0" and build CPE and network silicon that would support the new specs and, ultimately, assist a push toward HFC networks that would support a bandwidth upgrade to 1.8GHz, a person familiar with the situation said.

Another big question centers on what the cost of DOCSIS 4.0 silicon will be if vendors are required to support FDX and ESD on the same chip. At the same time, there might not be enough volume involved to justify the cost of spinning one chip for FDX and another for ESD.

As a result of this uncertainty, Broadcom's cable technology efforts have reached a "stagnation phase," a source said.

Broadcom has not responded to repeated requests for comment on its product roadmap for DOCSIS 4.0 and its purported interest in the NRE model.

If Broadcom ultimately decides not to pursue DOCSIS 4.0, that could leave the cable industry without an important second source of DOCSIS silicon if the bulk of the silicon piece of that next-gen business is ceded to MaxLinear. But even MaxLinear has some questions about the best path forward.

MaxLinear, which has partnered with Intel on previous generations of DOCSIS CPE silicon, is engaged on DOCSIS 4.0, Will Torgerson, VP and GM of the broadband group at MaxLinear, said in an interview soon after the company announced the Intel deal.

But he also raised questions about how to best approach the DOCSIS 4.0 market and whether it will make sense to build chips that support both ESD and FDX.

"A lot of conversations [are about] how to build the right platform that can scale across cable operators and provides the right economics for us and for them to make it successful," Torgerson said. "It's a matter of getting real, solid alignment with the operator community around what we need to execute on in order to deliver on that."

Combining FDX and ESD on one chip can be done, Torgerson said. "The bigger challenge is: are people going to be willing to pay for it? That's a lot less clear to me," he explained. "That's where we need to work through the options and how we actually get to that vision … There are smart ways to do it and there are brute-force ways to do it."

There's an "interesting chicken and egg situation" occurring with DOCSIS 4.0 due to the split of interest in FDX and ESD and the acknowledged costs for a silicon spin that would support both technologies, said Jeff Heynen, senior director, broadband access and home networking at Dell'Oro Group.

He noted that activity centered around a new distributed access architecture (DAA) is also "up in the air" during the pandemic as cable operators focus on adding capacity on their existing networks, including upstream expansions that don't necessarily require DAA.

"It shows that operators … are taking a cautious approach, especially this year," Heynen said, noting that the situation could cause DOCSIS 4.0 product development to be pushed out or delayed.

Still time to get the bands back together
While the possibility of Broadcom being absent from the DOCSIS 4.0 party would raise risks, the truth is that there's still time to repair the situation even as this apparent game of chicken between Broadcom and the cable industry plays out.

Cable operators don't need to upgrade to DOCSIS 4.0 tomorrow. Plus, DOCSIS 3.1 silicon already exists today that would enable an expansion to 1.2GHz (up from today's networks that support up to 1GHz of spectrum) and provide enough capacity runway for a few years as cable operators look at bulking up the upstream through a "mid-split" that would raise the upstream ceiling to 85MHz or a "high-split" that would push it to 204MHz.

According to some industry insider estimates, a move to 1.2GHz could buy cable another six or seven years, while a pivot to 1.8GHz could extend that runway another 15 years. Multiple cable engineers expressed concerns that timing for a push to 1.8GHz could be disrupted or put in jeopardy without firm DOCSIS 4.0 commitments from the chipmakers.

"The cable industry is squeezing the silicon vendors, and when you squeeze too hard, things break," a person with close and longstanding ties to the DOCSIS tech ecosystem said.

There's also a possibility that other silicon companies could be convinced to pursue the DOCSIS 4.0 market and bring some new blood to the market.

Speaking on a DOCSIS 4.0-focused webinar in March, Doug Jones, principal architect at CableLabs, said ten chipmakers were already involved in the new specs. "We don't know their roadmaps yet, but they are working on their product plans," Jones said.

Related posts:

— Jeff Baumgartner, Senior Editor, Light Reading