TL;DR The LN will incentivize large hubs to offer managed bitcoin accounts to customers that will be dramatically cheaper and easier to use. This invites the kind of financial middlemen and mischief that Bitcoin was created to work around.
Introduction: A successful future with the Lightning Network
The Bitcoin network is currently capable of only processing a handful of transactions per second globally. This is woefully inadequate, and so there is a lot of work being done to find solutions to address this. The Lightning Network is one of the main technological developments aimed at massively increasing the number of possible Bitcoin transactions. However, if and when this particular system is deployed, it will introduce economic incentives that may spell the end for Bitcoin, or at least for most of its primary benefits including both its limited supply and its censorship resistance.
To explain how this might happen, first imagine we are a number of years into the future where the Lightning Network (LN) has been deployed and everything has gone dramatically well. Bitcoin along with the LN now underscores a large proportion of all global financial transactions. At this point, the demand for on-chain transactions would be enormous and so also extremely expensive. Even if the block size / block weight limit were 10 times higher than it is today, the mining fee to get a transaction mined could be staggering by today's standards. Being conservative, let's say it is now $100 for a typical transaction.
In this environment, the vast majority of all Bitcoin transactions would happen through the LN. People could either pay $100 each time they make a payment, or just once to open an account with a LN hub which would enable potentially limitless cheap transactions (with some conditions). Most people would also receive their income through the LN channel, which would potentially allow them to keep their payment channel (their account) with the hub open indefinitely. However, if they do not balance incoming and outgoing payments over time, they would occasionally hit the limit of their account, and need to pay (another $100 plus the balance limit) for it to be re-funded. Everyone will be strongly incentivized to keep the bulk of their spendable money in LN hub accounts, make all payments through the hub and try their best to minimize the number of times they need to close/open/refund their account.
At this point things are not too bad for the end user. They would benefit from the dramatic cost reductions from using the LN, and would remain in full control of their funds as they can always close their payment channel and settle the balance on the blockchain any time they like (for $100). There are several other benefits, including the privacy benefit of not broadcasting all transactions to the world on the main chain. However, this is not the end of the story.
Economy of scale: Hubs could get very large
Lightning Network hubs will need a fair amount of capital to get going. They will need to fund each payment channel with other LN hubs to a significant amount of bitcoin. The amount would ideally need to cover the maximum balance of all incoming and outgoing payments sent on the channel, from all payments relayed, ever. If that's not achievable, then the channels will just need to be occasionally re-funded when they hits their spend limits, along with the attached miner fees. Hubs that can afford to fund their channels with larger amounts up-front will need to refund fewer times—losing less money on miner fees—and will be able to extract significantly more income from the flow of transactions until they next need to do so. In other words, there is at least a small economy of scale here that will incentivize the growth of large hubs.
Hub-managed customer accounts
One way or another, either by economies of scale, or just by pure exuberance, very large professionalized hubs will appear. Given that they will have a lot of capacity, they will be looking for ways to onboard new customers. A significant problem all new customers face is the upfront cost of $100 for the miner fee and then the total value of whatever they would like to be able to spend on the account. One solution would be for these highly capitalized hubs to offer “managed” accounts to customers where the hub will make LN transactions on their behalf, just like how current banks can make electronic transactions on behalf of their customers.
This would not involve the creation of any new LN channels whatsoever. The hub would only need to maintain a fairly large account of liquid funds available to be spent on their open LN payment channels for their customers. These sorts of accounts could be offered with very little upfront cost, or even for free initially. The customer would even be able to start off with a zero balance, and just have money sent to it later via incoming LN transactions—from their employer for instance.
The customer would then not have direct access to the Bitcoin network at all. Instead, we are back to a situation like the present financial world where trust is fundamental. The LN hub could try to assure its customers and strengthen the level of trust by issuing tokens of some kind, rather than just leaving its customers with a promise to handle their funds appropriately. More likely though, the hub would just insure its customers' accounts against loss in the event the hub goes bust or otherwise loses the bitcoins. This latter option would mirror the current state of affairs where governments around the world are trying to phase out cash. If this happens, both with these hub-managed accounts and in a cashless society, all you will have is a bank account, a promise from the bank and their assurance: “Don't worry; it's insured”.
The re-emergence of fractional reserve lending
At this stage, hubs would be financially incentivized and able to hold fractional reserves to start loaning out depositors' bitcoins to borrowers. As with modern banking, most of the general public would not be opposed to this. It is quite a familiar practice and arguably good for the economy. It would also allow the LN hubs to offer free accounts indefinitely, perhaps ultimately paying interest to depositors out of the revenue generated from the loans.
Now, hubs that implement all of these features would probably be very large. Smaller hubs wouldn't have quite the same ability or inclination to get into this kind of risky business (and stay in business). Naturally, these larger hubs would end up under the spotlight of government oversight and would move to preemptively register for all recommended licenses. To the general public, these hubs will become the best known and most attractive of the available options. They would have a good strong established reputation, be fully insured, have close government supervision, government approval, and provide a good free service.
Once hubs of this size and sophistication have developed, it would be next to impossible for anyone to compete, much as it is now essentially impossible for anyone to start a bank. As the industry becomes professionalized, all hubs would become subject to government regulation and it would become illegal to start one without first getting the relevant licenses. It might be possible for people in their basements to start black market hubs anyway. And even though they would be horribly expensive to set up and wouldn't make anything close to the profit of the regulated hubs, they might nevertheless continue to exist to help reduce costs for a small number of idealists passionate about privacy and true freedom. The network of black market hubs might end up somewhat similar to the Bitcoin network as it is today, or smaller; essentially a little-known irrelevance to the vast majority of people, but an option to escape the system for those who choose to do so.
Embrace, extend and extinguish
So at this point, the lightning network would effectively either merge with or morph into a simile of the current banking system. Just as gold used to be physically ferried around in an expensive and inefficient exercise to conduct settlements between banks, bitcoins will instead be used to settle between hubs. Assuming that this system takes over completely from the current global banking system, then the mining fees for these settlement transactions might be truly astronomical, say $1,000 per transaction. So in the end, Bitcoin would be priced out of reach of everyone except financial institutions. Once it becomes particularly awkward and expensive for people to take delivery, due to these sorts of fees, Bitcoin's relevance will be gradually de-emphasized and eventually detached from the financial system by government decree—exactly as happened with gold in recent decades.
What went wrong
So how did this happen? The primary incentive for the growth of fractional reserve Bitcoin banks will come from any economies of scale in the costs of issuing transactions. So for instance, this applies where a service provider with a lot of capital can make transactions on behalf of its customers much more cheaply than the customers could themselves. Currently—or at least until very recently—no such economy of scale exists, primarily because there is currently no (well established) mechanism for aggregating multiple arbitrary payments into fewer or smaller blockchain transactions. With the Lightning Network however, heavily capitalized hubs will be able to make transactions at negligible marginal cost, while end users will need to spend a far larger relative chunk of their money opening and closing channels. As a result, there is an inescapable and huge economic incentive for hubs to act on behalf of their customers to issue transactions for them. This is fertile ground for hubs to then become banks, start up fractional reserve practices, dilute the money supply, gradually divert attention away from the underlying asset (Bitcoin) and ultimately detach it entirely from the financial system.
What can we do?
There are no doubt many clever ways to avert this possible future. One way is to ensure that the base layer is friction free. That is, we should aim to minimize the time and cost of on-chain transactions. This will leave no room for the growth of professional financial middlemen to re-emerge and reintroduce their bureaucracy, enforced mediation, censorship, monetary meddling, confiscation, counterfeiting, bailouts, bail-ins and wildly disproportionate influence and control they wield over the direction of the development of civilization.
Mining and node centralization has been explored in great detail. This certainly could become a problem too and we should try to guard against it. However we must not ignore the risks of centralization in higher levels, such as those within the Lightning Network.