Fascinating stuff. Curious what you think what the states could do to encourage competition between ISPs, per takeaway #4. Assuming we have the flexibility at the state level, we want to know what can be done.
Yeah this is exactly the right question. There are readers here who probably know grant program administration better than I do, but a couple random thoughts:
-- Can a state publish a version of the cost model by Census tract (or some similar level) that sets an expectation for the price they expect to pay per location? make that public so all the grant proposers have the same information and know what the state is expecting.
-- States won't, and shouldn't, always take the lowest offer. But there should be some criteria which justifies taking an offer that is not the lowest. That criteria should be clear going into the proposals.
-- Every state will have some amount of low cost locations. Some states will have a lot of them. The goal should be to get proposals that are close to $0 in BEAD funds for those. RDOF was extremely good at this.
Both your analysis and Tarana's suffer from the faulty assumption that location density is a good proxy for cost. Unfortunately, it's a lot more complex than that.
Field conditions, including topography, geology, aerial vs. underground infrastructure, condition of existing poles, special situations like railroad and water crossings, and clustering are first-order affectors. A mile of fiber in the Mississippi delta costs a lot less than a mile of fiber in the Rockies. The marginal cost of a passing is less when locations are clustered than when they are more-or-less distributed along a roadway.
Costs vary by provider. Incumbent telcos and MSOs have a huge cost and schedule advantage insofar as they have existing pole attachments, strand and frequently usable duct, as well as buildings and cabinet sites. Tier-1 and Tier-2 incumbents also enjoy volume purchasing power, master service agreements with contractors, and other economies of scale, pre-existing capital overhead (like tools, trucks and software) and will have optimized their deployment processes. To varying extents, their matching contribution can come from retained earnings rather than debt, so their cost of money is low. Some of them have publicly disclosed their cost per passing, and of those, the average is in the range of $600 to $900 -- much less than CostQuest's $5,589 (and yes, those numbers do exclude locations that they consider economically infeasible without subsidies). Of course, electric coops and munis have their own cost advantages -- especially the fact that their ROI requirements are much less -- but those typically will be overcome by the structural advantage enjoyed by incumbents.
The point here is that we really won't be able to have a reasonable estimate of cost distributions until all the bids are on the table. This doesn't help states in setting their high-cost thresholds. Some other methodology will be needed.
I agree that all of those are factors. And that it isn't really possible to capture all of them in a cost model. But as you say, we need something to set high-cost thresholds, and I would argue to set some kind of expectation on the grant proposals, even if imperfect. Something is better than nothing here. The CQA-FCC CACM is generally understood to be the starting point, and that's what's in the RDOF reserve prices. The CACM took into account some of those factors beyond density. (I'm using density here because that's how Tarana did it and if nothing else it is readable). It's unclear to me whether that's a capex model or has some kind of opex component.
I'm afraid that this is more than "the perfect is the enemy of the good". More like "garbage in, garbage out" (that's a little harsher than I intend, but captures the idea).
The objective should be for each state to spend down all of its BEAD grant while both maximizing the number of premises served by fiber AND not leaving any premise unserved or underserved. That is an optimization problem.
The high-cost threshold would be a solution IF it could be calibrated to achieve that goal. But it can't really be calibrated until the costs are on the table. That's the Catch-22. Some kind of iterative solution is needed.
If I were a state, I'd want NTIA to buy the CQA model for me (the one represented in the picture). I'd try to understand how the model was built as best as possible and how it relates to the ground situation (what kind of locations are at the low-end, middle, hockey stick). Then I'd use that to find a reasonable EHCL threshold. Perfect? no. But good enough.
I'd agree if the CostQuest model has enough input variables to capture at least the first-order affectors. Since it's a black box, we don't know that. Something-or-other about a cat.
Mike — Appreciate the piece. Your takeaways are sound. Since you asked (sort of), the math to get from $18.1k to $191B involves a material adjustment for relevant inflation, based on the BLS non-residential construction producer price index, which has had a pretty bad couple of years since the sample projects were priced, as noted in the study. Separately, your use of the phrase "settle — for wireless" in your opening suggests you are unfamiliar with recent fundamental advances on the wireless technology front. I encourage you to visit taranawireless.com to learn how our next-generation FWA platform is changing the game. Since its introduction in late 2021, over 200 ISPs in 19 countries and 40 US states have embraced our unique G1 platform and are routinely delivering fiber-class speeds with high reliability, in both licensed (CBRS) and unlicensed (5 and soon 6 GHz) spectrum. Would be happy to brief you on the details and share network performance stats if you'd like.
Thanks for the engagement and the clarification on the cost math. I've heard great things about your technology. I've read and heard a little bit about how it works and I can absolutely believe its demonstrably better than other FWA products on the market. I'm also looking forward to when you can use full power 6 GHz. I would love to talk more. If you reply to the substack email it goes to me.
Another topic we should discuss: how do you advise your customers to report their coverage areas and the throughput and latency available at those locations? I'm generalizing, but it feels like some FWA providers (not all) are filing unrealistic throughputs which, in my view, hurts the credibility of all FWA providers. I'm curious if you have recommendation/software for your customers on how to model realistic throughputs to individual locations.
I'm going to be tough to convince that with once-in-a-generation funding we shouldn't use it for the capital-intense task of laying lots and lots of fiber. But, I'm open to the conversation and I suspect I will be impressed with your technology.
Fascinating stuff. Curious what you think what the states could do to encourage competition between ISPs, per takeaway #4. Assuming we have the flexibility at the state level, we want to know what can be done.
Yeah this is exactly the right question. There are readers here who probably know grant program administration better than I do, but a couple random thoughts:
-- Can a state publish a version of the cost model by Census tract (or some similar level) that sets an expectation for the price they expect to pay per location? make that public so all the grant proposers have the same information and know what the state is expecting.
-- States won't, and shouldn't, always take the lowest offer. But there should be some criteria which justifies taking an offer that is not the lowest. That criteria should be clear going into the proposals.
-- Every state will have some amount of low cost locations. Some states will have a lot of them. The goal should be to get proposals that are close to $0 in BEAD funds for those. RDOF was extremely good at this.
Mike,
Great work!
Both your analysis and Tarana's suffer from the faulty assumption that location density is a good proxy for cost. Unfortunately, it's a lot more complex than that.
Field conditions, including topography, geology, aerial vs. underground infrastructure, condition of existing poles, special situations like railroad and water crossings, and clustering are first-order affectors. A mile of fiber in the Mississippi delta costs a lot less than a mile of fiber in the Rockies. The marginal cost of a passing is less when locations are clustered than when they are more-or-less distributed along a roadway.
Costs vary by provider. Incumbent telcos and MSOs have a huge cost and schedule advantage insofar as they have existing pole attachments, strand and frequently usable duct, as well as buildings and cabinet sites. Tier-1 and Tier-2 incumbents also enjoy volume purchasing power, master service agreements with contractors, and other economies of scale, pre-existing capital overhead (like tools, trucks and software) and will have optimized their deployment processes. To varying extents, their matching contribution can come from retained earnings rather than debt, so their cost of money is low. Some of them have publicly disclosed their cost per passing, and of those, the average is in the range of $600 to $900 -- much less than CostQuest's $5,589 (and yes, those numbers do exclude locations that they consider economically infeasible without subsidies). Of course, electric coops and munis have their own cost advantages -- especially the fact that their ROI requirements are much less -- but those typically will be overcome by the structural advantage enjoyed by incumbents.
The point here is that we really won't be able to have a reasonable estimate of cost distributions until all the bids are on the table. This doesn't help states in setting their high-cost thresholds. Some other methodology will be needed.
I agree that all of those are factors. And that it isn't really possible to capture all of them in a cost model. But as you say, we need something to set high-cost thresholds, and I would argue to set some kind of expectation on the grant proposals, even if imperfect. Something is better than nothing here. The CQA-FCC CACM is generally understood to be the starting point, and that's what's in the RDOF reserve prices. The CACM took into account some of those factors beyond density. (I'm using density here because that's how Tarana did it and if nothing else it is readable). It's unclear to me whether that's a capex model or has some kind of opex component.
I'm afraid that this is more than "the perfect is the enemy of the good". More like "garbage in, garbage out" (that's a little harsher than I intend, but captures the idea).
The objective should be for each state to spend down all of its BEAD grant while both maximizing the number of premises served by fiber AND not leaving any premise unserved or underserved. That is an optimization problem.
The high-cost threshold would be a solution IF it could be calibrated to achieve that goal. But it can't really be calibrated until the costs are on the table. That's the Catch-22. Some kind of iterative solution is needed.
If I were a state, I'd want NTIA to buy the CQA model for me (the one represented in the picture). I'd try to understand how the model was built as best as possible and how it relates to the ground situation (what kind of locations are at the low-end, middle, hockey stick). Then I'd use that to find a reasonable EHCL threshold. Perfect? no. But good enough.
I'd agree if the CostQuest model has enough input variables to capture at least the first-order affectors. Since it's a black box, we don't know that. Something-or-other about a cat.
Mike — Appreciate the piece. Your takeaways are sound. Since you asked (sort of), the math to get from $18.1k to $191B involves a material adjustment for relevant inflation, based on the BLS non-residential construction producer price index, which has had a pretty bad couple of years since the sample projects were priced, as noted in the study. Separately, your use of the phrase "settle — for wireless" in your opening suggests you are unfamiliar with recent fundamental advances on the wireless technology front. I encourage you to visit taranawireless.com to learn how our next-generation FWA platform is changing the game. Since its introduction in late 2021, over 200 ISPs in 19 countries and 40 US states have embraced our unique G1 platform and are routinely delivering fiber-class speeds with high reliability, in both licensed (CBRS) and unlicensed (5 and soon 6 GHz) spectrum. Would be happy to brief you on the details and share network performance stats if you'd like.
— Steven Glapa
Hi Steven,
Thanks for the engagement and the clarification on the cost math. I've heard great things about your technology. I've read and heard a little bit about how it works and I can absolutely believe its demonstrably better than other FWA products on the market. I'm also looking forward to when you can use full power 6 GHz. I would love to talk more. If you reply to the substack email it goes to me.
Another topic we should discuss: how do you advise your customers to report their coverage areas and the throughput and latency available at those locations? I'm generalizing, but it feels like some FWA providers (not all) are filing unrealistic throughputs which, in my view, hurts the credibility of all FWA providers. I'm curious if you have recommendation/software for your customers on how to model realistic throughputs to individual locations.
I'm going to be tough to convince that with once-in-a-generation funding we shouldn't use it for the capital-intense task of laying lots and lots of fiber. But, I'm open to the conversation and I suspect I will be impressed with your technology.
I hope we can connect,
Mike