Description

When there is a gold rush, sell shovels. Harmonic (HLIT) has established a dominant competitive position in the broadband equipment industry, with a long runway to take advantage of strong growth and demand for data and services. They sell what carriers are realizing is the most important component of a rapidly growing technological trend and will have entry barriers against competitors for years to come.

Bear Case: The bear cases on HLIT are that (a) it is a boring cable hardware company and thus bound by low margins and (b) that the DOCSIS 4.0 transition should end by 2030 at which point transition to fiber will favor other companies. But it turns out that the company has (a) near total monopoly position on a critical part of the next gen network and a good position on many other parts and (b) an excellent role to play in the fiber transition as well as a remarkable competitive position for the long run fiber industry. Multiples are still being set by the bear case and even at these multiples the long term stock price scenarios look good, rerating when the market realizes (a) and (b) are incorrect offer icing on top.

TL:DR: Harmonic has emerged as the unquestionable monopoly provider of the operating system for broadband cable internet providers, cOS, and is dominant in important hardware niches to the point that competitor hardware providers are forced to design their hardware around Harmonic’s software and hardware platforms just to sell as a “second source” for carriers. Harmonic’s cOS and field hardware also offers a smooth upgrade to fiber and is targeting the fiber market in the long term. Harmonic’s margins and revenue are likely to rise meaningfully for years to come. Financially, the company recently cleared out a significant convertible note and is buying back stock as management takes advantage of the current low multiples before the paradigm shift in business operations triggers a rerating.

[HLIT has been written up 3 times on VIC as far back as 2014, reading those in order is an amazing narrative of an unnoticed company that had an idea that seemed kind of out there, then seemed kind of like it maybe might be possible, then seemed like it might have some possibilities, and now has totally worked and taken over the industry.]

The Broadband Business

The transition of the cable network through DOCSIS 3.1 to DOCSIS 4.0 has certain key elements. 

The first element is the “outside equipment” which traditionally refers to amplifiers, splitters, cables and other. This is the lowest margin element with many competitors. Harmonic does participate here, specifically mostly with FDX (Full Duplex) and ESD (Extended Spectrum) amplifiers which are the higher margin and more difficult components but still have plenty of competition. For FDX amplifiers, generally considered the most difficult, Commscope and Harmonic are the two alternate sources used by Comcast, for example. The market for this element is predicted by various sources to reach about $1.3B/year by 2027 and level out until 2030 when it is expected to decrease due to shift to optical. My take is that the transition to fiber will occur later than that, but $1.3B/year in 2027 giving HLIT 15% of the market seems like a reasonable take on this market, margins won’t be great for this chunk.

The second element is the DAA or Distributed Access Architecture elements, which moves nodes containing electronics that used to be in cable data centers / head ends out closer to the customers. (The historical location of these functions and continuity of market forecasts is why this still isn’t categorized as “outside equipment” even though it is indeed physically outside.) These boxes can be thought of as servers that go on the poles, with direct interfaces to feeds that run out to customers going in the box alongside them. As time passes they get more powerful and more functions are being moved into the DAA boxes. Harmonic has had so much success with their DAA boxes that even management has been surprised. They originally had an aspirational goal of 30% market share by 2025 and found themselves blasting through that mark in 2023 and now have stopped commenting on it because they seem to be facing some pushback from customers worried about monopoly supply dynamics. Market watchers put their share up to 70%, so I’ll call it between 50 and 70%. The harmonic DAA box has a modular design with center modules that can be swapped around to handle a wide variety of network approaches and can even remain in place for the switch to optical just by changing the output module which is a quick one man job. They also interface very well with Harmonic’s vCMTS (see below), which turns out to be so critical that new competing DAA boxes from other vendors are being made compatible with Harmonic’s vCMTS system just so they can be sold. Harmonic has supported this, no doubt to reassure customers that they won’t face a monopoly on DAA boxes. So the DAA box situation is a dynamic where HLIT owns the implementation and has a multi-year lead, but allows some other vendors to participate to mollify customers and avoid accusations of monopoly. This is an often seen dynamic and it usually works out very well for the dominant player who can control their margins and market share.

The final element is the vCMTS (virtual Cable Modem Termination System), which is basically the server system that connects to all the customer cable modems and services their requests. These were hardware implementations (CMTS) until recently and multiple vendors tried to make them virtual. I have been monitoring them for years. I watched Arris invest, I watched CASA systems try to build the same thing and then try to pivot to mobile networks with what they built before going bankrupt (I know people who invested in CASA but I decided to wait for signs of success, glad I was a skeptic). Harmonic won. They currently have 70-98% market share in installed vCMTS, depending on how you count, with the remaining market shares being the remnants of some other would-be players and trials as well as a few players who tried an open source approach in Europe that doesn’t have the needed performance. The Harmonic offering is slick and built on microservices from the bottom up which means that they can move bits and pieces out to the DAA boxes as the technology matures, which becomes even more important as the fiber transition occurs. It also means that Harmonic basically sets the standard because everybody needs to be compatible with them. The margins for Harmonic vCMTS are split into hardware and software. The hardware is typical hardware margins and often doesn’t appear in HLIT results since the product (cOS) runs on commodity hardware, but the software is a very nice recurring license fee. The margins there have been steadily improving as well. The very first deal was lousy - it was a beggars deal with Comcast to get into their first data center on existing commodity intel hardware to show what they could do. Fast forward to today and they get an increasing recurring software subscription revenue that they have confirmed can be modeled as increasing with homes served.

So what do we estimate for revenue for those latter two elements or for the whole set? One approach is to look at the cable company estimates of costs per passed address for the DOCSIS 4.0 transition, for which multiple cable companies have given the aspirational number of ~$100 per address. Now that number includes hardware, software, and construction, and it’s probably an under-estimate as such things always are at this stage to try to frame negotiations with vendors but it gives us one datapoint to triangulate from. We can take that $100, apply $15% for outside the plant, 40% for DAA and 45% for vCMTS and try to guess what the relative pricing strength will be in some of those areas. Another approach is to use algebra to extract price per install vs recurring fee estimates from historic Harmonic data (making sure to correct for the Comcast and 2019 $37.5mm oddity deal), from which one can actually get multiple plausible guesses but something around $30 per address for Harmonic’s share of DOCSIS 4.0 install actually seems like a central probability that agrees very well with both approaches. Using this number gives Harmonic about $5.5 Billion in current product broadband revenue between now and whenever you want to decide the primary end of that transition is (2030? 2033?). Given the shape of the curve, we’re looking at existing product annual broadband revenue rising to maybe $700mm crest by 2030 then settling back down a bit.

Notably, this same path implies the buildup of an additional recurring annual software revenue of about $250mm from per-modem cOS fees (estimated from an extracted model of current fees). 

Updated with the latest 6/13/2024 projections from management that would be Broadband total revenue (software plus hardware) of ~$800mm around 2027, and ~$1.1B around 2030, with about $250mm of 2030 being ongoing software revenues. Gross profit ratios should climb due to the software revenues and ebitda and net income with it. After that the existing products (specifically the current gen DAA box fillers) roll off and need to be replaced by fiber product sales. That’s where Harmonic gets even more interesting.

Fiber

Harmonics DAA and vCMTS products, and their operating system for the same (formerly CableOS now branded cOS) are fiber ready. The DAA units have a modular design that can fit units for coaxial cable for conventional service or units for connecting fiber. All a cable service providers has to do to offer a fiber build-out to a business park is make a business deal with the business park, then run the fiber to the nearest DAA and replace the coaxial module (about the size of a large sandwich) there with a fiber module and just like that they’ve completed a fiber build-out for that part of their network. The total micro-service architecture of cOS system allows the Harmonic system to support the fiber transparently at the head end, and even to transfer certain microservices to the DAA box that can improve fiber functions by running locally, with more of those transfers happening as the technology grows. 

When I interviewed people in the cable industry about Harmonic it was this ability, even beyond the cOS ubiquity, that leapt out as the thing that impressed the cable company engineers. They all knew to install Harmonic DAA boxes and cOS specifically because it meant their business teams could go around upselling fiber services to business parks and apartment buildings and they could upgrade the service in however long it took to pull the fiber to the nearest DAA plus about an hour for a box swap and a couple clicks in the cOS control panel. (One engineer also told a story about how he accidentally configured the network to route about 10 times as much traffic through one set of Harmonic DAA boxes as he should have and the whole thing autobalanced, handled it flawlessly, and put up a helpful little notice that perhaps this should be reconfigured when convenient. Nobody even had any service issues.) With recently released abilities to daisy chain together fiber runs between Harmonic DAA boxes to extend optical networks, Harmonic has provided the easy default upgrade path for legacy cable companies.

Due to this fundamental architectural advantage, Harmonic’s cOS monopoly puts them in the driver’s seat for the fiber transition and they are already building and selling new fiber products. So far it’s a few tens of millions of dollars and they haven’t really been breaking out the details, but the products are good and most importantly all of them fit into and work with Harmonic DAA, vCMTS, and cOS systems that the operators are already installing. This means that Harmonic will continue to collect their ongoing software licensing fee while selling new hardware for the fiber transition from an excellent competitive position of being the company who provides all the incumbent infrastructure. This puts Harmonic in an even better competitive position post 2030 than they are in now.

Additionally, it’s worth mentioning that Harmonic has been getting some wins since late 2022 with non-cable fiber accounts including traditional telecom operators and this is one of the places they are going after for scaling up their go to market. To this end Harmonic has introduced vBNG (the servicing that legacy optical networks used by telecoms uses) running in cOS, although the only one using it deeply that I know of is still a cable operator. It is still early days for Harmonic moving into the telecom industry but they have a solid monopoly to build off of and weak competition (see below). None of this optionality is built into the financial estimates or predictions here.

Competition 

As of right now there is no serious competition to cOS. None of the other players have the right micro-services architecture and/or the fiber flexibility for future proofing, and this explains the rapid growth of cOS. Now that cOS has reached 70%+ market share (98% according to some counts) and other vendors have begun designing their DAA units specifically for cOS compatibility, continued dominance seems unavoidable. The barriers to entry for any newcomer would be huge as the risks for any service provider to adopt an unproven alternate solution at this point would be immense and the savings very small compared to overall network costs. Meanwhile there are no significant scale customers available for a competitor to survive off of.

Some of the declared competition for those who want to look into it include Commscope/Arris who has a product called “vCore” that hasn’t seen much traction. Very recently, Commscope won the bidding for the cable products of bankrupt CASA systems. They have declared they want to be the viable “second source” for cable companies. Never a great market position but a market dynamic we have seen many times. It may be that we see Commscope in the long run as a viable second source with 10-20% of the market allowed and encouraged to remain compatible by Harmonic to avoid monopoly regulation. 

Vecima Networks (TSX: VCM) offers the “Entra” vCMTS product which seems to be at about the technical maturity Harmonic was at in 2021 but is now facing an entrenched Harmonic product. Vecima’s main angle is that they have a good market share in remote MACPHY termination. The Vecima product requires far more setup, is less integrated, and appears to be a point solution for heavy Vecima hardware users who aren’t ready for a full vCMTS solution yet. It is in limited deployment. I see it purely as a placeholder for companies that aren’t ready for Harmonic yet.

This could all shake out with Commscope possibly stronger in DSL/fiber vBNG and Harmonic stronger in cable vCMTS and each competing across to the other, Harmonic seems to have a stronger head start crossing that chasm than vice versa. Fundamentally, Harmonic and cOS are where Microsoft was in the 1980s, set for complete ownership of the platform of the broadband delivery industry.

The Video Business

Harmonic is doing well here. They streamed the superbowl and have a lot of good customers in a market where everyone realizes they need a streaming solution. For now I value this business merely as an ongoing Hardware/SaaS business. The margins are improving as they move to SaaS but I don’t see Harmonic as especially privileged in their technology or position so I give them no special future predictions for this business unit. Don’t get me wrong, this would be a great little company, but it doesn’t have the barriers to entry and monopoly advantage that HLIT has on the cable operations software side.

Valuation

Keeping it super simple, here is a valuation model with the broadband model as discussed above out to about 2030 (maintaining some agnosticism about fiber). The 2024 revenue adds up to their guidance, but I have divided things into Broadband hardware and software using my own model, so I can expand out into the future that way. If you want, you can just look at the top two lines together as “broadband revenue” and be in line with current reporting. The 2026 values agree with management's projections from their 6/13/2024 investor day, as they have historically been fairly good in their predictions (as good as management can be). The EBITDA margins are in line with history and management guides with slight rises after 2026 in cable due to monopoly impacts.

In this first model I used a terminal value of only 3x revenue (which is HLIT’s cable sector median value),assuming people still see the company as a boring cable hardware company. I also present lines for year by year valuation based on sales and EBITDA using sector medians, suggesting that the current sales projections produce good stock returns even with lousy industry standard multiples. 

As an additional probability, it seems highly likely to me that there should be a rerating at some point when people realize that this company is not only in a monopolistic position, but also positioned well for the long term transition to fiber.

Trading

Last summer there was a slowdown in shipments of cable units. Management focused on explaining this via inventory overbuild when customers overbought after the logistics slowdown whiplash after COVID, and this is likely true. At the time, however, the industry was also seeing hesitance at moving to DOCSIS 4.0 and DOCSIS 3.1 (“Turbo”) was just reaching test availability. My sources have indicated there was also an element of people wanting to go for 3.1 instead of 4.0. The availability of 3.1 from HLIT coincided with resumed sales growth. The nice thing about DOCSIS 3.1 (some people avoid the “Turbo” term, others relish it, I’ll skip it for brevity) is that moving to 3.1 is non-disruptive but at the same time the 3.1 to 4.0 step up is very easy with Harmonic gear and can be done in local islands for individual customers willing to pay for the difference. Thus 3.1 has been a huge hit and is much easier to justify to management while being a stealth path to 4.0. Selling 3.1, together with removing the inventory logjam, undid the stall in sales last summer. It is also worth noting that management saw the sales slump coming and warned about it on calls and saw the recovery coming and predicted that on calls as well, they have generally been very accurate in their predictions.

Some Related Thoughts

Trying to avoid going overly long here but to address some things people might wonder about.

What about Wireless Broadband?

Keep in mind that wireless broadband is limited to a few homes per neighborhood then the providers stop offering it due to saturation. Considering the growth rate in data demand for home broadband, and how much higher that growth rate is than mobile, combined with the higher infrastructure cost for wireless broadband, I think wireless broadband is a mistake in the long term. I wouldn’t be surprised if we're all reading articles in a decade about what a silly business decision wireless broadband was. Craig Moffet has a good analysis during this interview: https://www.youtube.com/watch?v=_KOVfQ_mpF4

What about Starlink?

It’s beginning to look like Starlink will improve Cable industry margins. Cable has to offer generally similar prices to all their customers and to offer service to everyone within a broad geographic/political area as part of the deal for their semi-monopoly role. That means that they are stuck offering a price that gives them a margin on the average but some customers cost 10x or even 100x more to deliver service to. Those are the same customers Starlink services well (typically the more geographically remote customers). Some rural service government programs help to offset some of these costs but in the long run cable would much rather simply have those few percent rural customers off the cable network and consuming Starlink so cable can focus on the customers that it is much cheaper to service. Higher margins for the cable industry means more money within the ecosystem for upgrades and service level improvements. Meanwhile Starlink has definite saturation levels (that they haven’t reached yet) and will never be a viable provider for dense metropolitan areas which are exactly where cable and fiber are most profitable. The two are excellent complements.

Catalyst

Rerating on software license margin clarity

EBIT and Sales rising at dramatically above GDP for the foreseeable future

Realization by industry watchers that HLIT is well positioned for fiber