The towers are not in America. The owners are.
North America holds 0.76% of the world's off-grid telecom towers. Africa holds 37.2%, some 49 times more. So the United States is not a market to sell into. It is where a very large share of the world's off-grid towers is owned, from two cities. This study measures the American ground carefully, finds it thin, and says what to do instead.
The number that kills the easy version of the pitch
The European tower scan found 2.88 USD/L of delivered diesel and 11,256 NPV-positive sites off the back of it. The instinct is that remote America must be worse, on bush planes and ice roads and a single barge. The instinct is wrong, because American diesel carries almost no fuel tax and the American pump price starts far below the European one.
What Alaska actually pays
- 25th percentile $3.74/gal
- Median $4.44/gal
- 75th percentile $5.04/gal
- 95th percentile $7.32/gal
- Maximum $11.85/gal
Fuel only, per kWh
- Median $0.316/kWh
- 90th percentile $0.556/kWh
- 534 GWh/yr generated from diesel
- 77 GWh/yr already non-diesel
- Population served: 94,955
Where the tail lives
- The median is unremarkable. The tail is not.
- Communities above $6/gal are the only ones where price alone argues for you.
- That is a shortlist of tens of sites, not thousands.
- Treat the US as a proof market, not a volume market.
If the American argument is cheaper fuel, the numbers do not support the trip. If the argument is that fuel arrives once a year by barge and the sun disappears for three months, the numbers support it completely. Steps Ventures · field study · August 2026
Delivery risk, not price
A remote Alaska village may receive a single bulk fuel delivery in a shipping season that only runs through the summer. The price is locked on that one day for the entire following year, whatever happens to the commodity afterwards. Villages of a few hundred people therefore store hundreds of thousands of gallons.
The 2026 season made the fragility explicit: ocean shipping rates up sharply, Bering Strait ice threatening to delay deliveries by weeks, and low water after breakup limiting barge access on the western rivers. Vendors warned of further increases on a war-driven supply crunch. The exposure is not the price of a litre. It is a missed window.
This reframes the product
Atom H2 is not competing with the cost of diesel in Alaska. It is competing with the risk of a delivery that does not arrive, and with the working capital tied up in a year of fuel bought in advance. Those are the two things a buyer there will pay to remove, and neither of them appears anywhere in the European model's NPV.
Three months with no sun, and a wind resource that peaks exactly then
This is the part of the American case that is genuinely unique, and it is the reason to come here rather than to another high-diesel market. Above the Arctic Circle the solar resource does not merely dip in winter. It goes to zero. A battery cannot bridge a season. Only a fuel can, which is why these communities burn diesel in the first place.
Can solar plus hydrogen carry a small site through the dark?
Taking a 0.5 kW continuous load, a modest repeater or monitoring site, at performance ratio 0.75 and a 35% power-to-hydrogen-to-power round trip:
| PV array | Summer surplus (kWh) | Recoverable via H₂ | Dark-season deficit | |
|---|---|---|---|---|
| 5 kW | 906 | 317 | 1,893 | short |
| 10 kW | 3,817 | 1,336 | 1,412 | short |
| 20 kW | 10,203 | 3,571 | 1,013 | closes |
| 30 kW | 16,911 | 5,919 | 937 | closes |
| 40 kW | 23,641 | 8,274 | 881 | closes |
Be honest about this, because a customer will check
A naively sized 5 kW array does not close even a 0.5 kW load. You need roughly 20 kW of panel to bank enough summer surplus to survive the winter through hydrogen. That is four times what the load suggests, on a structure that has to survive snow load and bears. And Atom H2's stated storage tops out at 242 kWh, well under the dark-season deficit, so a real installation is a stack of tanks, not one box.
The resource that is abundant precisely when solar is gone is wind: 8.15 m/s across Nov–Feb. In the Arctic the defensible architecture is wind-led generation with hydrogen as the calm-period bridge. Leading with solar here invites the one objection you cannot answer.
Small, named, and reachable in a single trip
Against the FCC's registration of 141,644 constructed structures nationwide, Alaska holds 1,484. Of those, 439 sit within 25 km of a community whose delivered fuel price is published. 93 sit near a community above $5/gal and 37 above $6/gal.
The expensive tail
| Community | Utility | Sites | $/gal | $/kWh |
|---|---|---|---|---|
| Tuluksak | Tuluksak Traditional Power Utility | 1 | $11.85 | $0.775 |
| Saint George | City of St. George | 1 | $10.99 | $0.914 |
| Noatak | Alaska Village Electric Cooperative | 3 | $9.41 | $0.622 |
| Igiugig | Igiugig Electric Company | 1 | $9.22 | $0.694 |
| Nikolski | Umnak Power Company | 1 | $8.51 | $0.967 |
| Akiak | Akiak City Council | 2 | $7.95 | $0.660 |
| Shungnak | Alaska Village Electric Cooperative | 6 | $7.77 | $0.519 |
| Atqasuk | North Slope Borough | 2 | $7.41 | $0.554 |
| Cold Bay | G & K Inc. | 3 | $7.19 | $0.560 |
| Ruby | City of Ruby | 3 | $7.02 | $0.522 |
Who owns the estate
| Owner | Structures |
|---|---|
| GCI Communication Corp | 154 |
| AT&T Mobility Services LLC | 48 |
| TelAlaska Cellular, Inc | 24 |
| State of Alaska | 21 |
| ALASKA POWER & TELEPHONE | 17 |
One counterparty dominates. GCI also runs TERRA, a 3,300-mile fibre and microwave network reaching more than 45,000 people off the road system. That is an estate of small, remote, year-round powered relay sites that is a far better fit for a kW-class box than any village powerhouse.
What this estate is not
A village power plant runs at hundreds of kW to megawatts. Atom H2's fuel cell is 2–6 kW. Do not pitch the powerhouse. The honest targets are telecom relays, repeaters, monitoring and navigation sites, water and clinic backup. Those are the kW-class loads that sit beside the powerhouse and share its fuel problem.
It is also far smaller than the figure in circulation. There are roughly 172,000 cell towers in the entire United States and about 419,000 sites once small cells are counted, and essentially all of them are grid connected with diesel as backup. Any claim of 300,000 off-grid or diesel-dependent US tower sites is the global off-grid count mis-scaled, and it should not be repeated to a customer who can check it.
Two cities hold the decision
Alaska is a proof, not a business. The business is that a very large share of the world's off-grid towers is owned by companies headquartered a train ride and a short flight from New York. Energy runs up to 60% of tower operating cost at remote African sites, Vodacom's African energy bill was around $300M in 2025, and diesel powers the majority of Africa's roughly 500,000 towers. Nobody in that chain has to be persuaded there is a problem.
American Tower
- Sites worldwide 149,686 (31 Dec 2025)
- Africa: 23,000+ towers, 7 countries
- Latin America: 47,081 sites
- Sold India (~76,000 sites) to Brookfield, closed Sept 2024. Reports Asia-Pacific and Africa as one segment (27,857).
SBA Communications
- Sites worldwide 46,390 (30 Jun 2026)
- Africa: South Africa and Tanzania only
- Latin America: Brazil ~30% of total, roughly 13,900 sites
- 29,028 of its sites are international across 14 markets. No market outside Brazil exceeds 5% of the total.
Phoenix Tower International
- Sites worldwide 30,000+ in 24 countries
- Africa: none
- Latin America: LatAm and Caribbean, but now outweighed by Europe
- Investors are funds managed by Blackstone, Wren House, BlackRock, Grain and USS. Blackstone took 35% in 2022 and sold about 20 points to BlackRock and Grain in 2024, so it is a minority holder. France alone is about 10,000 sites.
Do not walk in fighting solar
African towercos are electrifying fast and are proud of it. Atlas Tower Kenya has 82% of sites on solar and is adding 300 more. A pitch that sounds like "replace your diesel" is heard as "replace the solar programme I just got approved". Sell the long-duration layer that sits on top of the solar they already bought, priced on site availability and litres avoided rather than round-trip efficiency.
What a visit should actually produce
Two constraints shape everything below. The exchange is committed to Newlab in the Brooklyn Navy Yard, so Alaska is a set of sorties from that base rather than a relocation. And it runs on ESTA, which permits meetings, negotiation, conferences, site visits and market research, and prohibits productive work for a US entity, paid or unpaid. No installing, no commissioning, no running a deployment. This ends in signed intent and a funded winter trial, not in a box on a pad.
- Boston, for American Tower. 149,686 sites worldwide, 23,000+ towers, 7 countries in Africa and 47,081 sites in Latin America, about four hours from Brooklyn by train. That is around 70,000 sites in high-diesel geographies under one roof, which is more than the other two targets combined. They already sell backup power as a service using diesel, so the model is proven internally and only the fuel is in question. This is the priority.
- Boca Raton, for SBA and Phoenix Tower International in one visit. Two targets share a town, so make the trip once. SBA's prize is Brazil at roughly 30% of its estate; its African presence is only South Africa and Tanzania. Phoenix Tower gives a warm introduction through Steps' Blackstone relationship, but note Blackstone is a minority holder among five investors and cannot direct the company, and PTI's estate has shifted to Europe. Treat it as an introduction, not a shortcut.
- Ask for the person who owns the energy cost line for the African or Brazilian portfolio, not the sustainability team. Sustainability will take the meeting and cannot sign.
- Use Newlab for what it is good at. Six hundred deep-tech members, investor sessions and a steady event schedule make it a rehearsal room for a technical founder learning to carry a commercial pitch alone in English.
- One Alaska sortie around the Rural Energy Conference in Fairbanks, 27 to 29 October, with a site visit while the sun is gone and the wind is at 8.15 m/s. Come back with a reference site and a winter measurement agreement. One trip, not a residency.
- Convert one towerco into a paid multi-site pilot. A named country, five to ten instrumented sites, an agreed measurement scope and a stated expansion trigger. This is the deliverable of the whole three months.
- RE+, Las Vegas, 16–19 Nov is North America's largest clean-energy event and the towercos and their suppliers are there. Worth it only if a Boston or Boca Raton meeting has already landed and you are widening, not prospecting.
- Open the NATO Innovation Fund in parallel. European capital, no ownership friction, and Atom H2 is already a DIANA company.
- Turn conversations into paper before US corporate calendars close in the third week of December: a letter of intent, a named country and site count, a measurement scope, and who pays for what.
- Write the costed subsidiary plan properly. If a towerco pilot lands then a US contracting entity stops being theoretical, and the plan already promises the document.
- Leave before the ESTA 90 days expire. A 1 October arrival means departing by roughly 29 December. Write 90 days, never "1 Oct to 31 Dec", which is 91.
The honest read on American money
An order book
The capital event that matters is a framework with an owner of tens of thousands of off-grid sites, which funds the descent from $160,000 a unit toward the $42,000 floor. No grant on offer in the United States moves that number. Federal programmes are a distraction dressed as progress here.
NATO Innovation Fund
A €1B multi-sovereign fund investing in dual-use technology including energy resilience. Atom H2 is already a DIANA company. European capital, defence-adjacent thesis, no foreign-ownership friction. This is the most natural cheque on the board and it does not require moving the company.
US venture
Hydrogen storage remains a niche venture bet; late-stage and growth rounds took over 90% of disclosed storage capital in 2026. A US Series A on a kW-scale hydrogen thesis without field data is a hard raise. Comparable rounds, Photoncycle and PowerUP Energy, were led in Europe. Spend the three months making the raise easier, not attempting it.
Sequence, not a scattergun
Customer first, then capital. A signed multi-site pilot with a US-headquartered tower owner, plus one measured Arctic winter as the physics proof, is worth more to any investor than a season of American meetings. Raise on that, in Europe.