阿特拉斯能源解决方案 (AESI) 2026年第三季度业绩电话会：数据中心电力与第四季度展望

Atlas Energy Solutions (NYSE: AESI) 在拒绝低价支撑砂竞标后公布了疲软的三季度业绩，但管理层预计 2026 年第四季度所有业务板块均将实现环比改善。业绩说明会重点讨论了两个大型数据中心电力项目、其融资结构以及合同签署前的剩余步骤。

核心要点

• 2026 年第三季度营收为 2.678 亿美元，调整后 EBITDA 为 3030 万美元，调整后 EBITDA 利润率为 11.3%，处于管理层指引区间的低端。
• 在 Atlas 拒绝低于其价格门槛的竞标后，支撑砂销量降至 540 万吨。管理层预计第四季度销量为 570 万至 610 万吨。
• 在支撑砂、物流和电力业务环比改善的支撑下，Atlas 预计第四季度调整后 EBITDA 为 4000 万至 5000 万美元。
• 公司正与一家前沿人工智能实验室就两份数据中心能源服务协议（ESA）进行后期谈判。管理层表示，在 2026 年年底前完成签署是合理的预期。
• Atlas 获得了一笔 4.24 亿美元的桥式融资，定价为 SOFR 加 250 个基点。截至季度末已提款 1.71 亿美元，用于资助由客户成本补偿协议支持的设备采购。
• Socorro 电力项目按计划推进，30 台发电机中已有 15 台交付。管理层预计建成后的设施每年将产生 5000 万至 5500 万美元的调整后自由现金流。

核心财务数据

业务与运营表现

数据中心电力

Atlas 将其数据中心战略归纳为四个步骤：锁定设备、签署能源服务协议（ESA）、安排长期融资以及交付电力。

公司已通过两项成本补偿协议为两个拟建数据中心项目锁定设备。其中一项涉及与 Atlas 的卡特彼勒（Caterpillar）框架协议挂钩的电厂辅助设备（BOP），另一项涵盖额外的 283 兆瓦发电设备。设备已分配给该客户，如果项目搬迁，设备可转移至其他场地。

拟议的能源服务协议（ESA）预计期限为 15 至 20 年。管理层将定价结构描述为主要由固定容量费组成，辅以涵盖随着利用率提高而增加的运营和维护成本的变动部分。

两个项目均计划在 2027 年底前实现首次供电，随后在 2028 年逐步提高产能，其中一个项目将持续至 2029 年。Atlas 表示，现场施工必须在 2027 年初开始才能按计划推进。

Socorro 项目

Atlas 预计永久性 120 兆瓦 Socorro 电厂将在 2027 年第一季度末投入运营，并在第二季度初开始确认营收。所有发电机基座已浇筑完毕，基础和地下电缆导管已竣工，30 台发电机中的 15 台已到货。

在建设期间，一座 26 兆瓦的移动式设施已在为现场供电。一旦永久性电厂投入运行，管理层预计每年将产生 5000 万至 5500 万美元的调整后自由现金流。

支撑砂与物流

Atlas 选择不跟随竞争对手大幅让利的竞标，这导致第三季度销量减少并推高了每吨运营成本。管理层表示，这一决策遏制了价格的进一步下滑，并为以更有利的售价承接替代业务留出了产能。

公司在 2027 年已锁定超 2300 万吨的协议销量，定价与 2026 年基本持平。管理层表示，这一数字是基于二叠纪（Permian）压裂车队活跃度基本持平的合理基线假设。

支撑砂价格保持在每吨 16 至 18 美元的区间。由于司机短缺和司机薪酬上涨，Atlas 对物流定价持更加积极的态度。扣除东北部工厂关闭成本后，第三季度物流利润率约为 11%。

搭载 Kodiak 系统的自动驾驶卡车在季度内完成了 9000 多次交付，较上一季度翻了一番。Atlas 已接收 37 辆卡车，预计到 2026 年底车队规模将接近 45 辆，到 2027 年年中达到 100 辆。在满足监管和运营里程碑的前提下，公开道路运营目标设定在 2027 年年中。

油田电力

Atlas 预计在 2026 年结束时，其油气电力业务已部署约 200 兆瓦产能，其中约一半签订了多年期协议。管理层表示，该车队产能已接近售罄，且定价正在改善，特别是微电网业务。

该业务未来的支出将主要集中在维护和翻新，而非扩张增长。

管理层指引

第四季度的大部分资本支出与通过桥式融资进行的客户支持设备采购挂钩。管理层预计，随着产能的部署，较大数据中心项目的资本支出将延续至 2028 年。

Atlas 预计大部分长期项目资金将来自公司级和项目级债务，并辅以少量的股权或股权关联融资。公司表示不打算在签署 ESA 之前筹集项目股权资金。

风险与关注领域

• 未经签署的合同：截至业绩说明会召开时，两项主要的数据中心 ESA 仍处于谈判阶段。其最终达成需要电力、租赁、天然气供应以及工程与施工等协议之间的协调。
• 德克萨斯州许可证审批：德克萨斯州暂停了数据中心项目的空气许可证审批。Atlas 在暂停前提交的一项许可证申请目前已被搁置。管理层表示，如果延迟持续，ESA 协议可提供将项目和设备搬迁至其他州的灵活性。
• 融资与股权稀释：这些项目需要大量资金。Atlas 预计债务将支付大部分成本，但也承认将需要额外的股权或股权关联融资。
• 项目成本上升：管理层表示，由于客户需要更多负荷管理设备，数据中心资本成本已超出此前讨论的每兆瓦 150 万至 250 万美元区间的下限。
• 支撑砂价格压力：竞争对手继续提交低价竞标，而 Atlas 维持价格的战略可能会减少短期销量。
• 司机可用性：司机短缺正在推高物流成本，尽管管理层同样将 Atlas 的 Dune Express 系统和自动驾驶车队视为竞争优势。

分析师问答环节要点

管理层表示，具有投资级信用支持的 15 至 20 年数据中心合同可以承受相当高的项目级杠杆率。同类融资市场曾支持 80% 至 90% 的杠杆率，尽管 Atlas 表示可能会采取更为保守的做法。债务通常将设在子公司层面，并由合同付款和实体资产提供支撑。

Atlas 表示，先前对卡特彼勒（Caterpillar）设备的付款已资助了首个项目预期股权需求的大部分资金。管理层继续将无杠杆回报率目标设定在 15% 至 19% 的中高区间，EBITDA 回本周期为五到六年。

关于合同签署时间，管理层表示供应链、过渡性融资及承包商均已就位。最终的技术附件和商业条款尚待敲定，但 Atlas 预计两项 ESA 都将在 2026 年年底前完成签署，以保障既定的施工进度。

对于传统业务，管理层估计支撑砂与物流、油田电力及相关公司资产每年需要约 4000 万美元的维持性资本支出。公司预计这些业务将产生自由现金流并支持资产负债表去杠杆化。

业绩说明会完整文字实录

---

完整财报电话会议逐字稿

管理层陈述

Operator

Greetings. Welcome to the Atlas Energy Solutions Third Quarter 2026 Earnings Call. [Operator Instructions] Please note, this conference is being recorded.

I will now turn the conference over to Kyle Turlington, VP of Investor Relations. Thank you, Kyle. You may begin.

Kyle Turlington

Hello, and welcome to the Atlas Energy Solutions conference call and webcast for the third quarter of 2026. With us today are John Turner, President and CEO; Blake McCarthy, CFO; Tim Ondrak, President of Power; and Bud Brigham, Executive Chairman. John, Blake and Bud will be sharing their comments on the company's operational and financial performance for the third quarter 2026, after which we will open the call for Q&A.

Before we begin our prepared remarks, I would like to remind everyone that this call will include forward-looking statements as defined under the U.S. securities laws. Such statements are based on the current information and management's expectations as of this statement and are not guarantees of future performance. Forward-looking statements involve certain risks, uncertainties and assumptions that are difficult to predict. As such, our actual outcomes and results could differ materially. You can learn more about these risks in the annual report on Form 10-K filed with the SEC on February 24, 2026, and our quarterly report on Form 10-Q for the third quarter and current reports on Form 8-K and other SEC filings. You should not place undue reliance on forward-looking statements, and we undertake no obligation to update these forward-looking statements.

We will also make reference to certain non-GAAP financial measures such as adjusted EBITDA, adjusted free cash flow, and other operating metrics and statistics. You will find the GAAP reconciliation comments and calculations in yesterday's press release.

With that said, I will turn the call over to John Turner.

John Turner

Thanks, Kyle. For the third quarter, Atlas generated revenue of $267.8 million and adjusted EBITDA of $30.3 million and adjusted EBITDA margin of 11.3%. That was the low end of our guidance range. Sales volumes were lower, and lower volumes through the plants pushed up our OpEx per ton.

Volumes were down mostly because of a choice we made. We held our pricing and let some customers go rather than match bids from competitors selling below where we were willing to sell. Choosing to hold the line cost us sales volume during the third quarter.

However, our sales team has done a good job backfilling the temporary sales gap with volumes at reasonable prices. We expect both sand volumes and profitability to improve in the fourth quarter. Blake will walk through the details.

Now to Power. I know most of your questions are here, so I'll take them head on. I think about our data center business in 4 steps: secure the equipment, sign the energy service agreements, put long-term financing in place, and deliver the power. Each step takes risk out and brings the value of these projects closer to our shareholders. Each step is done for our 2 data center projects, and Socorro is already in the delivery stage.

Let me walk through all four. First, securing the equipment. We started on this well before this quarter. In November of last year, we placed our first order for long-term power generation equipment, 240 megawatts. In March, we signed a global framework agreement with Caterpillar that secured an additional 1.4 gigawatts of natural gas generation equipment with deliveries from 2027 through 2029.

In a market where equipment is a constraint, that's what put us in the position to have these conversations. This quarter, we added to it. We signed 2 cost reimbursement agreements or CRA, with a leading frontier AI lab, along with equipment purchase agreements to secure long lead time equipment.

The first CRA covers the balance of plant equipment needed to deploy a large share of the generation equipment we purchased under our global framework agreement with Caterpillar. The second CRA covers an incremental 283 megawatts of generation equipment. Our customer committed capital to secure this equipment ahead of final agreements, and that says a lot about how serious they are. These CRAs allocate the equipment to this customer.

Today, it's slated for 2 data center projects. And if a project moves, the equipment goes with the customer. Our customers' backstop of these equipment purchases is what allowed us to put in place a $424 million bridge facility with Goldman Sachs and Barclays. We have drawn $171 million and expect to draw more this quarter as progress payments become due.

The facility is priced at SOFR plus 250 basis points. Between the CRAs and the bridge, we've taken a lot of the financing and timing risk out of these projects. Long lead time equipment won't be what holds up first power.

Second, signing the energy services agreements. As indicated by the 2 CRAs, we are in active late-stage commercial negotiations on 2 significant data center projects with the Frontier AI lab. I know some have asked questions why these haven't been executed yet. And the reason is these are large projects with multiple counterparties.

While the power purchase is a big investment for us, it's small relative to the scale of the total data center investment. The ESA, the data center lease, the gas supply agreement and the EPC contract all have to come together in concert. With that much capital at stake, everyone needs the terms to be right, and so do we.

We are negotiating 15- to 20-year terms. The CRAs in the bridge show that nobody is waiting around. Our customer wants first power as soon as possible, and so do we. We'll announce the agreements when they're signed.

Third, long-term financing. I want to address dilution directly because I know it's on your minds. As part of the capitalization of these projects, we will complement the debt with some incremental equity or equity-linked securities, and I'd rather tell you that plainly than leave you guessing.

So let me frame it. Our plan is not to raise equity for these projects until the ESAs are signed, not before. When we do, it will complement debt. The 15- to 20-year structure of these contracts supports a high level of leverage. So we expect debt to fund a substantial majority of the capital.

Part of our equity share is already funded through our payments to date on the Caterpillar equipment. So any new equity would fund contracted long-term cash flows that are highly accretive, not projects we hope to win. As we said last quarter, we expect these projects to generate EBITDA paybacks of 5 to 6 years, and we expect unlevered returns in the mid- to high teens.

With the debt these contracts are expected to support, the return to both existing and new equity shareholders would be meaningfully higher. Whether that equity comes as common equity, subsidiary level equity or convertible debt, our objective is to choose the form and timing that minimizes dilution, and we'll judge it by one measure, cash flow accretion.

Fourth, delivering the power. Socorro is a good look at how we execute. We signed the Socorro PPA on March 31. Today, all generator pads have been poured. The first gen sets are on site. Our 26-megawatt mobile facility has been powering the site during construction, allowing our customer to ramp their testing facilities until the 120-megawatt plant is online.

We expect the permanent 120-megawatt plant to be online at the end of the first quarter, about a year from signing and to begin recognizing revenue early in the second quarter. Construction is well underway. The slab for the first power hall, the battery storage pad and the power distribution center foundations are all complete and the underground electrical conduit is in. Generator deliveries have begun and 15 of our 30 generators have been delivered.

Next, we'll erect the power halls, set the generators and distribution equipment in place and move into commissioning. Once operating, this facility is projected to generate $50 million to $55 million of adjusted free cash flow a year. It's the same team, the same Caterpillar equipment and the same construction model we'll use on the larger projects.

Socorro shows what Atlas does well: build quickly, operate reliably, and put power where it's needed. It's our blueprint for on-site power, and we're just getting started.

For the 2 data center projects, both are targeting first power in late 2027 with ramps through 2028, and for one of them into 2029. To hit those dates, we need to start site work in early 2027.

That brings me to permitting in Texas. The state has paused approval of air permits for data center projects. That was unexpected. We're engaged with the relevant agencies, and we expect the state to lay out what it wants to see from these projects. Our projects were designed to bring their own power, use minimal water and mitigate noise and light, and we think that puts them in a good position.

We're not going to predict when the pause ends. If there were an extended delay in Texas, our customer has opportunities in other states with clear permitting paths that require first power in 2027, and we would work with them to move the projects and the equipment there.

These 2 projects will not be the end of our growth in this business. We're seeing growing demand for behind-the-meter power for projects starting up in late 2028 and 2029. And we are having active conversations with our equipment providers to see what can be available and when. As the build out of America's compute infrastructure accelerates, we are positioning ourselves to be right in the middle of it all.

In our oil and gas power business, we expect to exit the year with approximately 200 megawatts deployed with roughly half under multiyear agreements. The fleet is close to sold out, and we're starting to see pricing improve, particularly on microgrid work. Going forward, most of the CapEx in this business will be allocated to maintenance and refurbishment. The goal here is cash flow. So secure, sign, finance, deliver. The first step is done and each one after it takes more risk out of the value we're building.

With that, I'll turn the call over to our CFO, Blake McCarthy.

Blake McCarthy

Thanks, John. On our last conference call, I detailed how there continue to be some egregiously low-priced bids in the sand market from some of our competitors. And then a few of our customers were insisting that we meet those bids if we wanted to retain their volumes.

Instead, we chose to remove ourselves from those RFPs, as Atlas provides a level of service reliability that can frankly be taken for granted. We do our job so well that our customers forget about what life is like on the other side.

So we drew a line in the sand with respect to pricing and allowed some long time customers to see what West Texas completion activities look like without Atlas keeping the site well supplied. While the decision negatively impacted our third quarter results, it stemmed further price bleed and allowed us to make room for incremental demand from current customers and 8-tier customers we have been targeting for a very long time, setting us up for improved performance in Q4 and for a full first half of 2027.

During the quarter, our overall sales volume and Dune Express volumes were down due to the vacancies left by the rejected bids, leading to sand revenue being down 8.3% sequentially. This resulted in higher OpEx per ton, particularly at our Kermit facility with overall proppant OpEx per ton of $13.96, including royalties.

The lower demand at Kermit did enable us to perform some much needed work on our water supply and wet shed, which has resulted in meaningfully improved output and costs in October. Overall OpEx per ton is expected to improve to approximately $13 per ton in Q4 and continue into 2027.

Importantly, our commercial team has done a wonderful job of filling the volume hole left by the rejected customer volumes in Q3. Overall sales volume is expected to be up meaningfully from 5.4 million tons in the third quarter to 5.7 million to 6.1 million in the fourth quarter.

As we mentioned on our prior quarter's call, RFP season for some of our key customers got started early this year. And while sand pricing overall remains challenged, the accelerated RFP season has resulted in us having a line of sight on more than 23 million allocated tons for 2027 at pricing levels roughly in line with 2026 pricing already. We are well on our way to be fully utilized across our system for next year.

Despite the strong commodity price environment, Permian operators have yet to really hit the gas pedal on activity. So our objective is to fill up the plants and begin moving price on the incremental ton. While sand pricing remains stubbornly in the $16 to $18 range, we are significantly more bullish on the outlook for logistics pricing.

Throughout the year, we have seen steady increases in driver shortages and thus in driver rates. As Atlas' logistical efficiencies provide us significant competitive advantages in the total delivered cost of sand, we used our advantaged logistics cost structure to sweat competitors who could not find drivers for a good part of the year. While this negatively impacted our logistics margin, it enabled us to capture incremental volumes for most of the year outside of the volumes we turned away.

However, the market has now reached a point where we feel it's not just possible, but necessary for us to push pricing. Through the oilfield grapevine, we have heard of numerous instances of well sites dependent on competitor sand experiencing significant nonproductive time. But cheap sand only saves you money if it shows up on time. And with the shortage of drivers in West Texas, driven by continued strength in the over-the-road market and the steadily growing sucking sound coming from the data center build-out market, the number of suppliers that can still ensure that the product gets to the job on time is shrinking rapidly.

We began passing through rate increases during the third quarter, which helped to lead to margins of approximately 11%, adjusting for expenses related to the closure of our Northeast region. We were able to achieve this despite the drop in Dune Express volumes quarter-over-quarter. With the anticipated improvement in Q4 delivered and Dune Express volumes, we expect logistics margins to improve to approximately 14% to 15% during the fourth quarter and continued strength in 2027.

As of the end of the third quarter, Kodiak has delivered 37 Kodiak-powered autonomous trucks to us. Our Kodiak-powered fleet made more than 9,000 autonomous deliveries during Q3, up almost 100% sequentially from the second quarter. Our partnership with Kodiak has begun to result in significant productivity gains, which we expect to accelerate as new operational milestones are reached that will expand the operational footprint these trucks are able to reach.

We are targeting operations on public roads by the middle of next year, subject to regulatory and operational milestones. The recent leap in the operational capabilities and efficiencies of our Kodiak fleet is incentivizing us to continue to invest in growing the technology as a percentage of our overall trucking fleet, which is on course to approach 45 Kodiak trucks operating by year-end and 100 by the middle of next year. With driver shortages a key issue in West Texas today, the combination of the Dune Express and autonomous deliveries has only enhanced Atlas' reliability premium in the Permian.

Our Power business reported sequential growth of approximately 11% with improved utilization in our distributed power fleet and the ongoing operations at our new facility in El Paso. Q4 contribution from this business is expected to display continued improvement as we deploy larger portions of that fleet under long-term contracts.

Q3 adjusted cash SG&A, excluding extraordinary litigation expenses and other nonrecurring items, was $25.2 million. Adjusted cash SG&A is expected to be approximately $22 million to $24 million for the fourth quarter. Growth CapEx for the third quarter was approximately $181.9 million, of which roughly $160 million is tied to our November 2025 order and build-out of the Socorro project.

Maintenance CapEx for the quarter was $5 million. We expect fourth quarter CapEx of $315 million to $345 million, depending on the timing of project progress payments and now expect full year CapEx to approach $700 million. The majority of the Q4 spend is tied to the purchases backed by the CRAs and financed through our bridge loan facility. CapEx for the Sand & Logistics business and our Oilfield Power business will remain focused on pure maintenance activities in the fourth quarter and into 2027 as current pricing dynamics do not justify incremental growth capital, and we will instead focus on optimizing free cash flow harvest.

On the heels of our successful convertible issuance in April and the recently announced bridge financing, Atlas' existing liquidity is enough to satisfy our immediate capital needs. As John mentioned, the continuing growth of our behind-the-meter Power business will require incremental capital.

We expect the vast majority to come from corporate and project level debt with a modest equity component. And those financing efforts will follow, not precede the execution of the contracts that underwrite these investments.

Looking ahead to the fourth quarter, we expect sequential improvement in every segment of our business and for adjusted EBITDA to fall in the range of $40 million to $50 million.

I will now hand the call to our Executive Chairman, Bud Brigham, for some closing remarks before we turn the call over for some Q&A.

Ben Brigham

Thank you, Blake. Every era of progress has been built on energy. The industrial age ran on coal and steam. The digital age runs on electricity and demand is rising faster than anyone planned. Data centers are the factories of the AI economy.

A single new campus can require hundreds, if not thousands, of megawatts, and developers are running into a hard truth. The grid cannot keep pace. Interconnection queues stretch for years. New capital transmission takes a decade to permit and build. The capital is ready and the chips are ready, but the power is not. That gap is where Atlas stands.

We are building a dedicated power business, providing a fast path to reliable power on private local grids operating independently from the increasingly fragile public grids. So-called behind-the-meter power compressed project time lines provides utility class electricity without the delays and regulatory challenges of the public grid. Instead of years in a queue, customers get reliable on-site electric supply sized to their load and backed by a team that knows how to keep it running.

We have the right strategy for the times, the right equipment and critically the right people. The companies building the future need partners who move at their speed. Atlas was built in the field on large energy infrastructure projects where reliability is a daily requirement, not a slogan.

We bring that discipline to every megawatt we deliver. The next chapter of American innovation belongs to those who can power it. Atlas intends to help write it.

Thank you for joining us today for our third quarter conference call. This concludes our prepared remarks, and I will now turn it over to the operator for Q&A.

Operator

[Operator Instructions] Our first question comes from Stephen Gengaro from Stifel.

分析师问答

Stephen Gengaro

A couple for me. Can you mind giving us maybe a little more details on the financing. You talked a bit about it on the prepared remarks, but how should we think about the financing? What type of debt levels you're comfortable with on long-term contracts? And how will you kind of keep us informed as we go forward here?

John Turner

You want to take, Blake?

Blake McCarthy

Yes, sure. Yes. Thanks for the question, Stephen. So yes, we did give some color in the prepared remarks, but obviously, funding capital projects of this size, something we discuss every day. It's clearly a gigantic piece of the puzzle.

So happy to go into more detail. Only caveat being this, I don't have full control of when these contracts are executed. And since markets are fluid and can change fast, we do have to remain flexible.

It's important to highlight that we aren't going to fund these projects before the contracts are executed. The bridge financing we secured through our partners at Goldman and Barclays, that takes care of our near-term capital needs with the long lead time equipment, and that's all backed by the CRAs. But thinking a bit more holistically about this, the contracts that back these projects, they're different from any type of contract we've signed as a company before. We have take-or-pay sand contracts, but those are typically shorter in duration and serve more floors than actual contracted capacity.

These ESAs are hard contracts of 15 to 20 years with set monthly capacity charges where a customer has stated, I expect you to earn your return through the capacity payment. So that results in a significantly more stable cash flow stream than anything currently in the Atlas portfolio, and it opens up a whole different set of financing options.

Additionally, the contracts are typically backed by an investment-grade counterparty that guarantees the capacity payments. So behind-the-meter power in general, it's a bit new to credit lenders. So I think existing data center projects are probably the best analog is there are simply more precedent financing data points available.

On contracts of 15 years or longer with appropriate counterparties, the market has shown a willingness to underwrite up to 90% leverage on these projects. We'll likely be a bit more conservative, but assuming 80% to 90% leverage, that's not out of balance. This debt typically resides at the subsidiary level, and it's backed by the contracts and the hard assets and is supported by the credit wrap, again, from an investment-grade counterparty.

So now obviously, 80% to 90% leverage, that leaves about a 10% to 20% equity component. But as we mentioned in the prepared remarks, we have already substantially funded the equity portion of the first project through our purchase of the Caterpillar equipment, which will be contributed into the project.

The most important thing to remember on these projects and the related financings is both the debt and equity will be funding projects that are really massively accretive to the EBITDA and cash flow of the company.

To frame the size together, they represent multiples of our current EBITDA. So these are big projects at very attractive return profiles. And to put it bluntly, I think the market is going to want to fund these projects. And whatever we do, we're going to structure these projects and the financings to maximize returns for existing shareholders.

Stephen Gengaro

Great. Two other quick ones. The uncontracted portion of your global frame agreement, how much do you think these projects will absorb of that capacity?

Blake McCarthy

Yes, go ahead. So all right. Again, not going to provide too much incremental detail on the projects beyond what we've already disclosed until the contracts are executed. But you're a smart guy, Stephen. And as you probably inferred from the information we did disclose around the CRAs, these are both sizable projects, and it's not off base to assume that it would derisk a very sizable amount of the GFA.

Thinking about that a bit more, we're very pleased about how our relationship with Caterpillar has evolved. We're working very closely with them to evaluate like they have very clear end customer priorities where they want their equipment with the right people and their capacity -- and so they're managing their capacity going forward with that in mind.

We obviously don't want to just do the Socorro project and these 2 projects to be done. So we're working with them around active capacity additions to the GFA. But as you -- which basically does imply that, yes, these projects would take down a very large chunk of the GFA assets.

Operator

Our next question comes from Julien Dumoulin-Smith with Jefferies.

Patrick Coleman

It's Patrick instead of Julien actually. Back to the power contracting, I was just wondering if we could dive more into the timing of those. I think you've given us this 180-day bridge facility. You've said the negotiations are close to the end. That language has been used for a little while now. So I guess, is that kind of like a year-end expectation or before end of 1Q expectation? Or do we need to look at the extension terms on that 180-day facility?

John Turner

Yes. Look, this is John. We feel very confident about the steps we've taken and the position we currently sit in the organization and Atlas as a whole. The projects we're working on, they fit with our customers' core objective of initial compute in late 2027 with aggressive ramps from there on into '28.

So in order to hit those compute time lines in 2027, we've got to be breaking ground in Q1 and working at a pretty furious pace from there. Everything that has needed to be done, making these projects go on the front end has been done. We've derisked the supply chain. We secured interim financing and we've lined up the contractors.

The only thing left to do is execute contracts like you asked. And we expect for us to hit those time lines, we expect those contracts to be executed by year-end. We've effectively been negotiating both of those agreements and at the same time, both of the contracts that we're working on. And between those projects, so we expect pretty short order once we execute one, we'll get the other one executed.

Patrick Coleman

And a follow-up, can I just ask about the link between the customer reimbursement programs and the bridge facility. It's backing the bridge. So can you help me understand the timing link between those when the customer reimbursement facility starts paying the bridge?

Blake McCarthy

No. So effectively, what the CRAs do is they -- the customer effectively says that, hey, if we don't get to an ESA within a certain time line, we will fully reimburse the lenders for all the cash outflows. So they're effectively making their balance sheet available to backstop the entire financing.

John Turner

Through the interim.

Blake McCarthy

Yes, through the interim.

Operator

Our next question comes from Jim Rollyson with Raymond James.

James Rollyson

Blake, maybe if you could help bridge something here on CapEx. As we think about this with the project you have ongoing in Texas already and the 2 incremental projects that you've gone through the CRAs and bridge funding, and let's just say, assuming you don't add incremental capacity from CAT, maybe a little road map through next year on CapEx, so we can maybe do our own math on kind of where the funding or funding gap kind of comes in?

Blake McCarthy

Yes. I mean, I think that number changes significantly with the execution of the contracts, right? So you can -- with what we've disclosed, right, you can have like kind of the obligations of like 27 deliveries under the GFA. Those generators probably only represent 30% to 40% of the CapEx of these projects. So obviously, that number goes up pretty significantly with the execution of these contracts.

That being said, obviously, that comes with it a very clear cash flow stream on the backside of that CapEx. And so you can -- there's quite a few smart investors that have done some math based on what we disclosed on the CRA. Basically, like you can use the -- there's a couple of markers out there. I would say that the CapEx per megawatt of these projects has continued to climb as the load requirements of the load profile requirements of the end customer has continued to get more intense.

So I mean, these are big expensive capital projects, and it will be a significant amount of CapEx upon execution. But like I walk through in detail, it's a very financeable cash flows.

James Rollyson

Understood. And maybe -- go ahead.

Tim Ondrak

Yes. I think the other thing to note on that from a CapEx perspective is CapEx will be deployed as the assets are deployed. And so some of these contracts ramp into 2028, and therefore, some of that CapEx will also bleed into 2028.

James Rollyson

Yes. No, that makes sense. And maybe a bigger picture question. Early on in the development of this business, as it was kind of pivoting towards the data center opportunity as it kind of ramped up in your face, I think there was some view that this was mostly going to be bridge power that leads to ultimately grid connections. And when Bud was talking earlier, just made me think, obviously, 15- to 20-year deals seem more like a permanent solution.

I'd love to hear what kind of context your customers or potential customers that you're talking with, how they're viewing that now given all the changes that have happened in the market since this all started?

John Turner

Yes. I think, first off, as I think you can look at the agreements that we're talking about here with these 15-, 20-year contracts. I mean, customers are looking at this as permanent power. There has been a change, obviously, over the past 6 to 9 months about the grid and its capacity to fund it, then you've had some political backlash.

But I think it's been pretty clear there's been a turn that if a data center is going to come, it's going to need to bring its own power and it's not going to be able to rely on the grid and they won't be able to rely on the grid.

So that's really where you've seen the turn, and I'll let others chime in on that. But the discussions we're having are really the ones that are about providing primary power.

Ben Brigham

Yes. This is Bud. I just want to amplify one of the comments that John made. That has become obviously increasingly apparent that the grids are less stable and a bit more fragile and particularly with the increasing demand on the grid, but also the intermittent energy that's been subsidized into the grid.

My view is, and I think it's pretty obvious that the development of these private grids, private power that is not pulling from the grid and that can actually be supplemental power -- provide supplemental power to the grid is very healthy. And particularly politically, it will be very beneficial, less vulnerable to the regulatory problems associated with the grid and in fact, again, beneficial to those challenges that the grid faces.

So we think the playing field has tilted our direction for providing this long-term power to power these AI factories that are going to drive significant economic growth for our country.

Operator

Our next question comes from Alexa Breno with Goldman Sachs.

Alexa Breno

Just wanted to ask on the prospective contracts. Can you talk a little bit about what that structure might look like? Is it more of a fixed capacity payments? Or is it minimum take-or-pay commitments? Just how do we think about it as the data -- potential data center customers are then ramping up their loads?

Blake McCarthy

Yes. I mean, it's a pretty simple pricing structure. The majority of the payment comes through just a fixed capacity charge where it's just like this capacity is available. Like I said, you can think about the base return.

In addition to that, there's typically a variable component that covers the O&M fees -- sorry, operating and maintenance fees that fluctuates with utilization. That is more for like potential cost overruns and expenses and stuff like that. I mean, these are -- like we've talked about, both the loads, the load profiles are pretty harsh on the systems. And so the more they use it, obviously, the more the maintenance spending goes up and say that functions to capture that incremental expense. But it's -- they're really relatively simple contracts in terms of the pricing structure.

Alexa Breno

Okay. That's helpful. And then on the base business, could you just talk a little bit about what that free cash flow could look like in a normalized or a through-cycle environment? How do we think about the CapEx and cash associated with that part?

Blake McCarthy

Yes, that's a great question. And I think that -- when you think about just the base business, we've really entered this more like mature phase where from -- really think about from the genesis of Atlas until just very recently, we've been in a pretty steady growth phase with pretty heavy capital investment.

And it's very much -- Sand & Logistics is very much like leans towards the infrastructure side of the OFS universe, right, where it's heavy CapEx on the front end and then low capital intensity and maintenance on the back side. As you could see in our Q3 results, our maintenance CapEx has continued to decline. It was only around $5 million last quarter.

As we think ahead, the combination of the legacy chain logistics business with the Oilfield Power business and then some cats and dogs in terms of the corporate side and IT CapEx, you're probably looking at this business being the base business, only really requiring somewhere around like $40 million of maintenance CapEx per annum.

And I think that we've continued to see improving operations from our plants. Our logistics -- we talked about there's positive momentum on our logistics margins moving forward. And then the Power team has done a really good job of contracting that base legacy oilfield fleet.

And so that we're really thinking about all 3 businesses being very steady cash cows going forward and delivering positive free cash flow to help contribute to delevering the balance sheet as we move forward.

Operator

Our next question comes from Derek Podhaizer with Piper Sandler.

Derek Podhaizer

I guess just on Power, I think you mentioned in your prepared remarks with the CRA agreements that if the project moves with the AI Frontier lab, the equipment moves with it. So maybe you could just talk a little more about whether this equipment is specifically designed for these 2 data center projects with this customer or if there's something else that we should be thinking about with that comment? Just some more color on that would be helpful.

Blake McCarthy

Yes. That's a good question. And I think you hit the nail on the head where the equipment is dedicated to the customer, right? They're the ones making their balance sheet available to backstop the CRAs and it belongs to them, not necessarily to the project. That being said, they continue to prioritize these 2 projects. I think everybody is very optimistic about moving forward with that.

I think one thing you touched on that's a pretty interesting trend we're seeing from the larger players is they're starting to standardize their data center designs, the actual data halls. I think, obviously, that creates more efficiencies as they look to aggressively ramp their own internal compute profiles. And so that makes it much easier for these systems to move from location, right?

Like there's not going to be a big delta in the design of one data center versus the other. I think that they're going to -- that trend is going to continue to gain steam as we move forward.

John Turner

Yes. I also think it's going to impact the construction of it as well. Standardized, it's going to make it much smoother. They're going to be able to deploy it more quickly so that they can meet their first power needs and then ramp pretty aggressively through that.

So I don't think it's specifically designed. This equipment is not specifically designed for -- our power equipment is not designed for a specific site. It's although it's being applied to a specific site, it can be used in multiple places.

Derek Podhaizer

Got it. Okay. That's helpful. And then switching over to oil and gas and the proppant side. So you talked about more than 23 million tons allocated for next year, 2027. Can you maybe break that apart as far as what that implies for Permian activity from a frac fleet or frac activity perspective?

And then what really needs to happen through RFPs? Where do you see those volumes go? Could we get closer to the mid-20s if we add to more frac fleets that we're not thinking about? Just maybe some color around the different drivers and levers when we think about that more than $23 million and what it means for Permian activity.

John Turner

Yes. I mean, we feel really -- I mean, we're pretty confident in our position right now as far as it relates to volumes for next year. I know we've got a line of sight on more than enough tons to completely fill up our plants.

Obviously, still -- and I think the RFP season, I'll let Brian comment on this. I think the RFP season is -- we're through a big part of it. We do have line of sight. And I do think any increase in frac activity, meaning with the addition of frac crews out there could put upward pressure on sand pricing. But do you want to go ahead and give more comments?

Brian Leveille

Yes. I think John covered it well. Most of the large operators have already completed their RFP process and made awards for those programs. There's always going to be spot work and programs that add crews depending on what the market looks like throughout the course of the year. And we typically find a lot of success in capturing that business. That's really, I think, where that incremental volume could come from at this point given the fact that most of those large operators have already awarded their volumes.

Blake McCarthy

Yes. And then, Derek, to the heart of your question, though, that more than $23 million of allocated tons is based around effectively flat pressure pumping fleets in the Permian. Like as we talked about on the call, it's been a really attractive commodity price environment, but the operators have certainly not hit the gas pedal in terms of adding incremental crews.

And I think part of that is due to just a shortage of actual available pressure pumping equipment capacity. And so it's been -- after initial -- like right after the beginning of the Iran conflict, you saw a few incremental crews get added. It's been pretty steady state. And so all of the allocated tons that we've been able to suck up for next year, that's really been market share capture.

Operator

Our next question comes from Eddie Kim with Barclays.

Edward Kim

Just on the Power side, I know you mentioned that a lot needs to happen before an ESA actually gets over the finish line. But as you see things right now, do you think we could get to a signed ESA on those 2 CRA agreements before year-end? Or does it look like it might be more of an early 2027 event?

Blake McCarthy

Yes. We already answered that earlier in the call. I think that we said that we've already done most of the steps on the front end of this project, right? We derisked the supply chain, we secured interim financing, we've lined up the contractors.

The only thing really left is to probably update some technical exhibits in the ESAs and hash out some final terms. But I think it's, yes, very reasonable to expect both ESAs by year-end.

John Turner

I mean, in order for us to hit the time lines that we laid out, we're going to have to -- we expect to hit those -- like I said, hit those time lines and execute the ESAs by year-end.

Edward Kim

Got it. Understood. And then just shifting over to Fin, just on that same line of questioning, you mentioned line of sight to more than 20 million -- 23 million allocated tons for next year. Just looking at the past 2 years, it looks like -- I mean, your final kind of year-end volumes is trending more or less in line with the allocated tons that you quoted at the beginning of both of those years.

So for next year, I mean, should our kind of baseline assumption or should we take, I guess, 23 million tons to be sort of preliminary guidance in terms of volumes at this point? And to the extent that unallocated tons come through that would represent upside, of course. Just curious kind of your confidence level on the -- tons coming.

Blake McCarthy

That more than 23 million tons probably serves as a good baseline assumption.

John Turner

Yes.

Operator

Our next question comes from Keith MacKey with RBC Capital Markets.

Keith MacKey

Just first question, you mentioned the air permit moratorium. Do these 2 projects that you're working on now currently already have their air permits? And if not, is an air permit a prerequisite to signing the ESA? Just some clarity there would be helpful.

Tim Ondrak

So we have one air permit that's been submitted. It was submitted pre-moratorium. And so that's in hold right now. But our air permits are standard air permits. And so it's a permit by rule. So we're obviously watching the governor's office very closely for any more guidance coming out of there.

Those are going to be in any contract that a power developer signs, those are going to be a conditions precedent. And we expect to have that cleared up and to move forward. We would sign ESAs with that being in there and allowing the flexibility to permit the project somewhere else should that moratorium continue in Texas.

Keith MacKey

Okay. And just on the CapEx per megawatt, you mentioned that it's been climbing. Can you just maybe give us some help as to where you think things are now?

Blake McCarthy

Yes, it continues to climb. I think the one thing -- the key thing to remember on CapEx per megawatt climbing is that it doesn't change the economics of the contract.

As John reiterated on the call, these projects it's a pretty easy back solve where we expect them to generate unlevered returns in the mid- to high teens with EBITDA paybacks of 5 to 6 years.

I think that going -- if you rewind 12 to 18 months on the -- as people started first using distributed power assets to power data centers, the CapEx per megawatt was much lower. And people quickly found that the systems as they were designed, were not able to handle the load swings.

And so there's significantly more load management equipment that's been added to those systems, and that's expensive equipment. And so you -- I think that previously, we've talked about it being anywhere from $1.5 million to $2.5 million per CapEx per megawatt. That's probably trended at a low level above $2 million, just the low end of that range now starts at the $2 million, and it continues to climb. But obviously, that passes through into incremental price per megawatt.

Operator

Our next question comes from Josh Jayne with Daniel Energy Partners.

Joshua Jayne

Most of might have been answered on the Power side. So I'm going to focus on the Sand & Logistics business. The first one is you talked about walking away from some lower-priced sand business in the third quarter. I'm just curious, we have a pretty different crude price environment today than we had a year ago.

So when we think about the market in 2027, you alluded to what you've ultimately contracted, but are you seeing a different sense of urgency from some of the larger operators today heading into next year than you did a year ago.

Tim Ondrak

Yes. I mean, a different sense of urgency on operations.

Blake McCarthy

Well, a different sense of urgency just in terms of securing their stand.

Tim Ondrak

Yes. And I think for sure...

Blake McCarthy

Josh, I think that was 100% evidenced by the fact that they pulled forward their RFP. Like typically, we have this call like our Q3 call, we're like, all right, as we enter RFP season, we'll update you all on our allocated tons.

Well, we're having this call already, and Brian and his team have already kind of filled up the cup. And so I think that the procurement teams of our customer base, they're sharp people. And they knew that like, hey, prices like commodity price is moving, let's take risk off the table.

And there are still quite a few zombie mines out there, right? They just -- they're not really producing -- they might be running 20%, 30% utilization, but they're still like serving there in bids, like getting there in the RFPs and trying to bring down -- like just fill up their volume for fixed cost absorption.

So savvy procurement people use that to try to bring down the price of sand. But we effectively said like, hey, we're not going below this. And we also know with the dynamics you've got in the trucking market right now, like that's a big differentiator, right?

You can buy dirt cheap sand -- but you're not going to save money when it doesn't show up on the well site, and it causes a lot of NPT, right? Like having all that equipment out there not working because the sand hasn't shown up is a lot more expensive than paying an extra $3, $4 a ton of sand. And really, that's the position we've taken.

And with our larger customers, they understand that. They understand that the incremental efficiencies that Atlas provides their operations. And we've been very satisfied with how we've moved through the RFP process.

Now one last thing I'd say is, though, is that now that we've got that cut filled for next year, I would say that like Brian mentioned, hey, we've always been really successful when spot opportunities come up.

Well, when you have that marginal ton is really where we would look to raise price, right? Like that's the -- hey, like when I got to move things around. I got to potentially like add risk to my production system. Like I'm not going to do that without there actually being incremental price coming through. And so that's really what we're going to look to do.

And in general, I think that our Q4 guide points to a stronger Q3. And really, it's like you get under the hood, the entire Q4 progression is sequential improvement into year-end, where based on our customer schedules, December is going to be a really strong month, which kind of runs counterintuitive to typical seasonality you see, except in strong commodity environments where customers want to hit the next year -- early in the next year, they want to hit running. And so we're certainly seeing that. And then -- so I think things look tighter as we get into Q1.

John Turner

I mean, I think next year, I mean, I think we're probably starting -- I think we're going to start seeing double-digit increases in logistics pricing. There's going to be pressure all across the basin to meet needs of the operators. I think they're seeing that because all of a sudden, there's opportunity outside of the Permian where a trucking company can go make 3x their profit a day by going and hauling for data centers versus hauling for hauling sand and doing logistics in the oilfield.

So I think there's going to be significant pressure on the logistics piece, especially going into the first quarter. And I think that's a clear reflection of what we're seeing with the operators and the RFPs right now.

Ben Brigham

Which is where we're advantaged.

Blake McCarthy

Exactly.

Joshua Jayne

And we agree with you on the trucking side as we tracked over the course of this year. My follow-up, sort of building on that point is just on automation and what you're seeing with Kodiak. So you gave some pretty helpful guide points, start your 37 trucks today, 45 at the end of this year, 100 by the end of next year potentially.

And when you talk about your [Audio Gap] is there more room to run on the current fleet that you have on those 37 trucks from a utilization standpoint? I'm just trying to unpack how much further like on a load per truck you could go today relative to where you're operating? And maybe just expand on that a little bit more and the importance of automation in that relationship moving forward. And then I'll turn it back.

John Turner

Yes. I'll start. I think that the relationship with Kodiak is, I mean, is very important to Atlas. I mean, obviously, we got into the autonomous business. We've been doing autonomous deliveries out there for a couple of years.

Kodiak continues to improve their efficiencies. We've done a significant number of loads, and we continue to increase that as we move through and further expand that relationship. I mean, ultimately, we want to be on the road and Kodiak and Atlas are working towards that. And then we will continue -- that's when you will continue to see a lot of the efficiencies continue to accrue to Atlas.

I think from a standpoint of efficiencies and costs, I think we have a long way to go on the Kodiak relationship. I think there's obviously things that we've been doing to see the improved profitability on our side. But the one thing I will say is that going into next year and once we get to over the road, we're going to see real pressure on driver shortages and the relationship and what we do with Kodiak is really going to separate and give Atlas a clear competitive advantage in the Permian because we're the only ones that can run those Kodiak trucks on the road and use those in our deliveries.

Operator

We have reached the end of our question-and-answer session. I would now like to turn the floor back over to John Turner, for closing comments.

John Turner

Yes. Thanks, everyone, for your time and your questions. Here's where we are. I'd say in Power, we've secured the equipment for our 2 data center projects, and we're working to get those agreements signed. We'll make those announcements when we do.

Our plan is the same one that we laid out last quarter and laid out this quarter, long-term financing followed by no equity until signed contracts and permanent financing is put in place. Obviously, very excited about where our Power business is in the organization. I mean, this has been transformational for Atlas, and we're continuing to execute. And I feel very confident that we're going to get this done. And we're also going to continue to grow this business.

Obviously, in the Sand & Logistics business, excited. I mean, we've been in a trough over the past year or 2 years. And I do think 2027 is going to be a good year, not only for sand and for the logistics side, also if we have continued higher oil prices, we're going to continue -- you're going to see sand prices go up because you're going to have frac crews added. And when you do that, I mean, there's nobody better positioned than Atlas to take advantage of that with our cost structure.

And I just want to thank everybody for joining us, and we look forward to reporting to you our fourth quarter results in next year. Thanks.

Operator

This concludes today's teleconference. You may disconnect your lines at this time. Thank you for your participation.