Secret price of a LoRa gateway


Someone mentioned this earlier. But just to reiterate the point. "Consumers" of £100-£200 gateways are tinkers and makers who understand the limitations and comfortable with diagnosing potential issues that might arise in the future.

The argument around gateway cost vs tens of nodes is valid. But realistically speaking, most enthusiasts are unlikely to have many tens of nodes. But almost all definitely need some of gateway or other.
I don't know how the situation is over in mainland Europe. But I personally find our ludicrous that we don't have much TTN coverage here in the UK (compared to say the Netherlands).
And I'm adamant to change that, even if it means funding many tens of gateways myself.
But in order to justify the costs, it will need to be lower than what's currently out there...


Everyone is a bit right :wink: . Like @jmarcelino says: companies like iMST and RisingHF aren't doing rocket science, they are mostly integrating off the shelf parts. This kind of manufacturing typically has a gross profit margin of around 7%.

On the other hand, this is a niche market where mainstream market parameters aren't that relevant. Compared to contract manufacturers, iMST and RisingHF do spend money on their own board design and that has to be paid as well. There also isn't a lot of competition. And this being a B2B market, certain services and warranty procedures are on the same level as price.

I don't see a price race to the bottom happening anytime soon. Why would you go for lower margins if the market is willing to pay Lorank8 prices for hardware worth €200 at most. Sure, maybe some tinkerer will put a lot of time in an open source RF pcb that will actually work, but I wouldn't hold my breath (remember: there already have been some attempts at this, and none have resulted in something useful). This also won't have much impact on the pricing of iMST and the likes, because companies still prefer to pay more for something that has some form of warranty attached to it.

(Jose Marcelino) #64

Indeed, we need coverage - and lots of it - to deploy actual applications and nodes people will want to use and even pay for. That's where the value is, not on charging premium pricing on gateways especially those being deployed by volunteers for all to use.

I recommend looking at the early days of the Internet circa 96, back then operators couldn't really afford the very expensive, quality, modems at full price to serve their communities. Many made do with cheap clunky ones that just made the experience bad for everyone (kind of what we have with single channel gateways now).

Modem manufacturers such as US Robotics realised this and started massively subsidising, often by over 50%, their high end modems [1] and later channel banks (essentially racks of modems) for early service providers and BBSs. They knew that by helping the community offer good connectivity the user base would be loyal - and grow by network effect - rather than being frustrated by the experience (dropped connections followed by difficult redialling were commonplace).

The end result was these communities would have more users and the modem companies sold more modems. Then you'd get an ad for the modem company subsidising the service. In the LoRa world it's even simpler - every new user is a new Semtech user.

This is why I don't believe in expensive gateways.



Right now, yes. But then you're stuck with a RasPi for 35 bucks and a i880a for 155 bucks. Throw in antennas, powersupply and stuff and you're at 200.

BTW: i actually think, Imst doesn't really make money out of their i880a boards, i think for them it's more marketing.


The SX1301 is a pretty complex, therefore expensive piece of silicon, including an FPGA. When using LoRa, compared to using WiFi for example, you're always stuck with Semtechs IP, so they're making the price and there will be no other supplier soon.

BUT: Semtech anounced a new chip, which will be less complex and cheaper.

(Jose Marcelino) #67

iMST offers 20% discount if you order 100 pieces so I'm sure there is some (healthy) margin on their side. Same with RisingHF where prices go down to $75 for the concentrator board.

The SX1301 doesn't actually include a FPGA, that's an extra component only on the AP2 reference design (and I don't know any company offering it yet in a commercial product).

It does include two microcontrollers - automatic gain control (AGC) and packet arbitrer (ARB).

The new chip offers the same functionality with only a narrower operating temperature range at a substantial discount, thus to me that shows it's pure market positioning play and not because of any technical merit.


I for one do not believe in €100 gateways, and imo you can't really compare the modem market or a company like USR to the Lorawan market or a manufacturer like iMST. The main reason being the lack of competition, on several levels.

In the days of BBS's, dial-up was mainly the only way to get online. In this defined market, USR had to compete with several other modem makers like Telebit and Hayes. All these companies could differentiate their products, e.g. USR developed X2, Lucent K52flex etc. There was real competition with different products, but all serving the same goal: getting you online through your mom's telephone line.

With LoraWan, there is no competition. Semtech holds the only IP, and manufacturers can only differentiate in trivial areas. Semtech can lower their prices all they want, if one manufacturer uses this to heighten his margin, the others will do the same. Also: the Semtech chip is only 25% of the BOM, even if they subsidize the price as much as USR did, it won't make that much difference.

When you compare Lorawan to other LPWAN techs, there isn't any competition there too, at least not concerning gateways. You can't buy a Sigfox gateway, you can't buy a NB-IOT gateway. It's reserved for carriers. And that's the big difference with modems: gateways aren't targeting the end user, it's for carriers. Nodes is where the fight will be fought, and the tech offering the lowest price per node will have the advantage. No one has any incentive to commoditise the carrier stuff.

Also: LoraWan gateways are already dirt cheap. The price of decent lightning protection (certified, not the Aliexpress stuff), site leasing, insurance, manhours of installing the gateway etc. are multitudes more expensive than the hardware.

Yes, for hobbyists not caring about all that a few 100 euro's seems very expensive, but the truth is gateway manufacturers won't shed a tear about maker communities. If they land only one contract with a mid-sized carrier, it's already worth way more than all maker sales combined.


What is that bytestream for then, which is loaded into something when starting the concentrator then?

(Jose Marcelino) #70

It's the firmware for the two small (8K RAM) microcontrollers inside. I'm not sure which kind, maybe 8051?
Perhaps @Sylvain_M here knows as I think he's listed as one of the original code maintainers.


Yeah my fault, i always understood the FPGA is inside the SX1301, never looked closely enough to the reference design.

(Mark Stanley) #72

This thread has been fascinating to read.

When I bought my first gateway I could spend £1500 or £500, so I bought the cheap one.
Thanks to the Things Network kickstarter and some great diy projects the cost of gateways has fallen towards £200-300.

£300 is not "expensive". It is phenomenally good value. It means you can deploy a wireless network across a whole city for £2-3k. That's AMAZING!

Price can be a barrier to entry, and I'm glad it is coming down so more people can get involved. I can honestly say that my gateway has brought me enormous fun and has been the best £500 I've spent on tech in decades. So I don't think you should be worrying too much about the already low price of a gateway - just focus on doing something incredible with it instead :slight_smile:


They are pretty expensive too. If they were really cheap the price of the gateways wouldn't hurt.

(Sylvain M) #74

You're right. Each SX1301 include two tiny cores (can't really be considered microcontrollers) programed in very low level C by my esteemed former colleague Matthieu. They're running fast and doing some absolutely necessary housekeeping. So, no FPGA, and no embedded firmware, everything is loaded by the driver itself.