Serve Robotics Inc (SERV) 2025 Q3 法說會逐字稿

Hello, and thank you for standing by. I would like to welcome everyone to the Serve Robotics third-quarter 2025 financial results and conference call.

您好，感謝您的耐心等待。歡迎各位參加 Serve Robotics 2025 年第三季財務業績發表會及電話會議。

Now, I would like to turn the call over to Aduke Thelwell, Head of Communications and Investor Relations. Please go ahead.

現在，我想把電話交給通訊和投資者關係主管阿杜克‧塞爾韋爾。請繼續。

Thank you, operator, and good afternoon, everyone. Welcome to Serve Robotics third-quarter 2025 earnings call. With me today are Serve's Co-Founder and CEO, Ali Kashani; and our CFO, Brian Read.

謝謝接線員，大家下午好。歡迎參加Serve Robotics 2025年第三季財報電話會議。今天與我一同出席的有 Serve 的共同創辦人兼執行長 Ali Kashani；以及我們的財務長 Brian Read。

During today's call we may present both GAAP and non-GAAP financial measures. If needed, a reconciliation of GAAP to non-GAAP measures can be found in our earnings release filed earlier today.

在今天的電話會議中，我們可能會同時提供GAAP和非GAAP財務指標。如有需要，可在我們今天稍早發布的獲利報告中查閱 GAAP 與非 GAAP 指標的調整表。

Certain statements in this call are forward-looking statements. You should not place undue reliance on forward-looking statements. Actual results may differ materially from these forward-looking statements, and we do not undertake any obligation to update any forward-looking statements we make today, except as required by law. For more information about factors that may cause actual results to differ materially from forward-looking statements, please refer to the press release we issued today, as well as the risks and uncertainties described in our most recent annual report on Form 10-K and in other filings made with the SEC.

本次電話會議中的某些陳述屬於前瞻性陳述。您不應過度依賴前瞻性陳述。實際結果可能與這些前瞻性聲明有重大差異，除法律要求外，我們不承擔更新今天所作任何前瞻性陳述的義務。有關可能導致實際結果與前瞻性聲明存在重大差異的因素的更多信息，請參閱我們今天發布的新聞稿，以及我們在最新的 10-K 表格年度報告和向美國證券交易委員會提交的其他文件中描述的風險和不確定性。

We published our quarterly financial press release and our updated corporate presentation to our Investor Relations website earlier this afternoon, and we ask you to review those documents if you haven't already.

今天下午早些時候，我們已在投資者關係網站上發布了季度財務新聞稿和更新後的公司介紹，如果您尚未查看這些文件，請盡快查看。

With that, let me hand it over to Ali.

那麼，就讓我把麥克風交給阿里吧。

Thanks, Aduke, and thank you, everyone, for joining us.

謝謝Aduke，也謝謝大家的參與。

We are at a pivotal moment for Serve. This past quarter, we crossed the threshold for 1,000 robots deployed. That's not just some round number; it's an inflection point. You can feel this in the sidewalks that we serve.

我們正處於Serve的關鍵時刻。上個季度，我們部署的機器人數量突破了 1000 台大關。這不僅僅是一個整數；這是一個轉捩點。在我們服務的街道上，您就能感受到這一點。

The future of cities is autonomous, and we are at the forefront of this, turning it into daily reality in these neighborhoods across the country. This is not just swapping humans for robots. We are unlocking new possibilities for cities. We are rewriting the operating system of how cities function, how goods move, how spaces are shared, how businesses reach residents. When the whole system upgrades like this, everything gets better: safer streets, friendly and greener cities, and more prosperous businesses and workers.

城市的未來是自治的，而我們正處於這一領域的前沿，正在全國各地的社區中將其變為日常現實。這不僅僅是用機器人取代人類。我們正在為城市開啟新的可能性。我們正在重寫城市的運作系統，改變貨物的運輸方式，改變空間的共享方式，改變企業與居民的連結。當整個系統進行這樣的升級後，一切都會變得更好：街道更安全，城市更友善、更綠色，企業和工人也更繁榮。

So why now? What's possible today that wasn't possible before? There are four forces that have really converged. First is physical AI. It's really finally caught up with our ambitions. Advances in distributed training, also the better onboard compute that is now available, it lets us ingest orders of magnitude more sensor data. And that leads to incredibly more capable AI models that can help machines really understand the world in real time.

那麼，為什麼是現在？如今有哪些事情是以前不可能的？有四種力量真正匯聚到了一起。首先是物理人工智慧。它終於趕上了我們的預期目標。分散式訓練技術的進步，以及如今更強大的車載運算能力，使我們能夠攝取數量級更大的感測器資料。這將催生出功能更強大的AI模型，幫助機器真正即時理解世界。

Our perception and planning models are improving on the streets every single day. Each mile traveled enriches our data set. Each model update expands where, when, and how quickly and how safely we can move. And this all has a compounding effect.

我們的感知和規劃模型每天都在街頭不斷改進。每行駛一英里，我們的數據集都會更加豐富。每次模型更新都會擴大我們可以移動的地點、時間、速度和安全性。這一切都會產生疊加效應。

Second, every hardware component needed to create this advanced, inexpensive, and intelligent machines has matured. This includes powerful sensors that are now at mass scale and low cost, paired with electric motors and batteries that enable new vehicle form factors. Technologies like LiDAR sensors were unaffordably expensive just a few years ago, but now, we have partners like Ouster, who are shipping thousands of sensors each quarter in record numbers. That benefits everybody in the ecosystem because of the economies of scale.

其次，製造這種先進、廉價、智慧機器所需的每一個硬體組件都已經成熟。這包括目前已實現大規模生產且成本低廉的強大感測器，以及與電動馬達和電池相結合的馬達和電池，從而實現了新的車輛外形。幾年前，像雷射雷達感測器這樣的技術價格昂貴，難以負擔，但現在，我們有了像 Ouster 這樣的合作夥伴，他們每個季度都在以創紀錄的數量出貨數千個感測器。由於規模經濟效應，這有利於生態系統中的所有參與者。

Third, consumers have adopted the convenience of online and on-demand ordering, and merchants need CapEx-light and labor-light capacity so that they can economically serve the demand. Restaurants have optimized their operations in the post-pandemic reopening, but they now need a way to unlock and realize that full potential. They want dependable, right-sized logistics that matches their demand by the hour, and that's what our fleet can deliver.

第三，消費者已經接受了在線按需訂購的便利，商家需要資本支出少、勞動力成本低的產能，以便經濟高效地滿足需求。疫情後重新開業，餐廳已經優化了運營，但現在他們需要一種方法來釋放和實現其全部潛力。他們需要可靠、規模合適的物流服務，能夠按小時滿足他們的需求，而這正是我們的車隊所能提供的。

And last, but not least, it's the cities themselves. Cities are asking for quieter and cleaner and less congested streets. Smaller vehicles made for specific use cases can now replace those 2-ton vehicles that we've become so addicted to. And this is the future.

最後，但同樣重要的是，還有城市本身。城市居民希望街道更安靜、更乾淨、更暢通。專為特定用途而設計的小型車輛現在可以取代我們已經如此依賴的2噸重車輛。這就是未來。

Our robot's footprint is small, with an electric powertrain and a friendly presence. We earn the right to scale when we operate with safety and transparency and community respect and when we are working closely with the cities that we serve. And we create these new jobs, full-time employee jobs, in neighborhoods that we are serving.

我們的機器人體積小巧，採用電動動力系統，外觀友善。當我們以安全、透明和尊重社區的方式運營，並與我們服務的城市密切合作時，我們才有資格擴大規模。我們在所服務的社區創造這些新的全職員工職位。

Our third-quarter results prove that we are on the right track with all this. Our delivery reliability was nearly 100%, while our delivery volume increased 66% in a single quarter. And we continue to maintain a strong safety record. We now deliver for over 3,600 restaurants, which is an amazing 45% increase from the last quarter and more than a nine-fold increase since last year. This is all proof that autonomy can be safe, reliable, and predictable, even as we scale rapidly.

我們第三季的業績證明，我們在所有這些方面都走在正確的道路上。我們的配送可靠性接近 100%，而配送量在一個季度內成長了 66%。我們一直保持著良好的安全記錄。我們現在為超過 3600 家餐廳提供配送服務，比上一季成長了驚人的 45%，比去年同期成長了九倍多。這一切都證明，即使我們快速擴大規模，自動駕駛也可以是安全、可靠和可預測的。

We did all this while also expanding faster than anyone in our industry. In less than a year, we grew our fleet size 10x, our cities 5x, and our major platform partners 2x. Last month, we announced partnering with DoorDash, the largest delivery platform in the US and one of the largest in the world. Combined, Uber and DoorDash serve over 80% of the food delivery market in the United States, which provides us an incredible reach to consumers and merchants.

我們做到這一切的同時，擴張速度也超過了業內任何其他公司。不到一年時間，我們的車隊規模增加了 10 倍，覆蓋的城市數量增加了 5 倍，主要平台合作夥伴數量增加了 2 倍。上個月，我們宣布與 DoorDash 建立合作關係，DoorDash 是美國最大的外送平台，也是全球最大的外送平台之一。Uber 和 DoorDash 合計佔據了美國食品配送市場 80% 以上的份額，這使我們能夠觸及大量的消費者和商家。

The scale that we've achieved in the last few months really changes things. Here's how to think about it. With a few dozen robots, you're running pilots. At a few hundred, you're starting to prove repeatability. Beyond 1,000, the system tips. We run more efficiently, the economics improves, the national partners really lean in, and our learning really speeds up, all of which makes every new city launch smoother and every new robot smarter than before.

我們在過去幾個月所取得的規模確實改變了很多事情。你可以這樣思考。用幾十個機器人，你就可以進行試點了。達到幾百次，你就開始證明結果的可重複性了。超過 1000 時，系統就會失衡。我們的營運效率更高，經濟效益更好，國家合作夥伴也更積極參與，我們的學習速度也大大加快，所有這些都使得每個新城市的啟動更加順利，每個新機器人也比以前更加聰明。

As we scale with precision, we've gone from one market to five fully operational hubs, covering over 3 million population and well over 1 million households. That's nearly 70% increase in a single quarter and more than a tenfold increase in our coverage compared to the same time last year. Not only have we 10x'ed our fleet, we've 10x'ed our reach. And as importantly, each neighborhood adds to our rich data sets, with new and novel edge cases, which really accelerates our ability to learn across the network and it compresses our timeline for future city launches.

隨著我們精準地擴大規模，我們已經從一個市場發展到五個全面運營的中心，覆蓋超過 300 萬人口和超過 100 萬個家庭。這相當於一個季度成長了近 70%，與去年同期相比，我們的報道範圍增加了十倍以上。我們的機隊規模不僅擴大了 10 倍，服務範圍也擴大了 10 倍。更重要的是，每個社區都為我們豐富的數據集增添了新的、獨特的極端案例，這確實加快了我們在整個網路中學習的能力，並縮短了我們未來城市上線的時間表。

On that note, I'm excited to share with you our next three expansions. We've just been greenlit to expand into Buckhead, Georgia; Fort Lauderdale, Florida; and Alexandria, Virginia, before the end of this year. Alexandria also gives us a toehold in the Washington, D.C., area. We're building the first truly national interconnected autonomous delivery network on a common AI software platform and operations stack so that a single-partner integration would light up multiple metros and thousands of restaurants at once.

說到這裡，我很高興與大家分享我們接下來的三個擴張計劃。我們已獲准在今年底前將業務拓展到喬治亞州巴克海德、佛羅裡達州勞德代爾堡和維吉尼亞州亞歷山大。亞歷山大也為我們在華盛頓特區地區站穩了腳步。我們正在建立第一個真正意義上的全國互聯自主配送網絡，該網絡基於通用人工智慧軟體平台和營運堆棧，因此只需一個合作夥伴即可同時為多個大都市和數千家餐廳提供服務。

For example, in Q3, we announced our partnership with DoorDash. Coming in addition with our existing contract with Uber, this allows our existing robots to unlock an incredible amount of additional volume. A robot that's completing a delivery for DoorDash could do a delivery for Uber on its way back. So this would actually improve utilization levels for robots.

例如，在第三季度，我們宣布與 DoorDash 建立合作關係。加上我們與 Uber 的現有合約，這將使我們現有的機器人能夠釋放出驚人的額外運力。正在為 DoorDash 送餐的機器人，在回程途中還可以為 Uber 送餐。因此，這實際上會提高機器人的利用率。

And we're not done yet. In addition to our existing partnerships with Shake Shack and Little Caesars, we've also started delivering for Jersey Mike's Subs, the famed sub sandwich chain with over 3,000 locations nationwide. And while it's too soon to give details, we also expect to add another well-known national QSR brand to the lineup. So the partnership platform is growing nicely.

我們還沒完。除了我們與 Shake Shack 和 Little Caesars 的現有合作關係外，我們還開始為 Jersey Mike's Subs 提供外賣服務，Jersey Mike's Subs 是一家著名的潛艇三明治連鎖店，在全國擁有 3000 多家分店。雖然現在透露細節還為時過早，但我們預計將另一個知名的全國性快餐品牌加入產品陣容。所以，合作平台發展勢頭良好。

And concurrently with that, we are also developing an unparalleled map of cities, the curb cuts and slopes and potholes and obstacles and patterns, a living atlas that becomes a valuable asset and an operational advantage. And while we do that, we are also deepening our community bonds at a very hyper-local level. With merchants and landlords and HOAs and precincts, we need to integrate respectfully into these neighborhoods as we grow.

同時，我們也正在繪製一幅無與倫比的城市地圖，包括路緣開口、坡度、坑洞、障礙物和模式，這是一幅生動的地圖集，將成為寶貴的資產和營運優勢。同時，我們也在非常在地化的層面上加深與社區的連結。隨著我們的發展，我們需要與商家、房東、業主協會和社區保持尊重，並融入這些社區。

Now, let's take a step back. Serve is pioneering a robotics and autonomy-as-a-service platform that packages the full power of our autonomy stack, our hardware, and our urban robotics operation playbook. With the last quarter's acquisition of Vayu Robotics, our platform is now increasingly reinforced by AI foundation models and a scalable simulation-powered data engine.

現在，讓我們退後一步。Serve 正在開創一個機器人和自主即服務平台，該平台整合了我們自主技術堆疊、硬體和城市機器人操作手冊的全部功能。隨著上個季度對 Vayu Robotics 的收購，我們的平台現在越來越多地得到了人工智慧基礎模型和可擴展的模擬驅動資料引擎的加強。

Under the hood of all this is the physical AI flywheel that powers everything. Our third-generation fleet leverages the best-in-class sensors which creates this proprietary urban data sets and that, in turn, leads to better AI models, which then creates more efficient and more autonomous fleets. A better fleet expands our TAM and increases our operating domain and verticals that we can do. And that, in turn, creates more pool in the market for more robots. So more miles, it's more robots, it's more data, it's better AI, and the cycle repeats itself.

這一切的背後是驅動一切的實體人工智慧飛輪。我們的第三代車隊利用一流的感測器創建了專有的城市資料集，進而產生了更好的人工智慧模型，從而創建了更有效率、更自主的車隊。更強大的機隊可以擴大我們的潛在市場規模，並增加我們的營運範圍和我們可以涉足的垂直領域。而這反過來又為更多機器人創造了更大的市場空間。因此，行駛里程越多，機器人就越多，數據就越多，人工智慧就越好，如此循環。

The integration of Vayu will actually accelerate this loop because it turns data into better models faster, and that leads to tangible gains in our delivery speed and autonomy and market reach. Our library of long-tail edge cases is expanding faster than ever, and the latest data that we gather on weather and obstacles and parades and detours, it all provides the learnings that are applied network-wide. So every robot is learning from every robot.

Vayu 的整合實際上會加速這個循環，因為它能更快地將數據轉化為更好的模型，從而在交付速度、自主性和市場覆蓋範圍方面帶來實際的收益。我們的長尾極端案例庫正以前所未有的速度增長，我們收集的有關天氣、障礙物、遊行和繞行路線的最新數據，都為我們提供了在整個網絡中應用的經驗教訓。所以每個機器人都在向其他機器人學習。

And this AI flywheel that we are building, the flywheel that's now accelerating, in turn, attracts exceptional talent. Our rare data scale and this real-world fleet presence that we have is pulling in elite builders. And they, in turn, ship better systems, and better systems attract even more elite builders. So the talent flywheel is also compounding.

我們正在建造的這個人工智慧飛輪，這個正在加速運轉的飛輪，反過來又會吸引傑出的人才。我們罕見的數據規模和我們在現實世界中的艦隊實力正在吸引頂尖的建造商。而他們又會製造出更好的系統，更好的系統又會吸引更多頂尖的建造者。因此，人才的飛輪效應也不斷增強。

All the forces that I mentioned are helping us execute our vision. I'm really proud of our team for reaching the 1,000-robot milestone last quarter. In September alone, we shipped over 380 robots. That is in a single month. That means we launched more robots in a single month than the prior quarters. This was a pivotal milestone. We promised to ship 2,000 robots by the end of the year during our IPO, and we are on track to do it, with robot number 2,000 planned to deploy in Miami in mid-December.

我提到的所有因素都在幫助我們實現願景。我為我們團隊在上個季度達到1000台機器人的里程碑感到非常自豪。光是九月份，我們就交付了超過 380 台機器人。那隻是一個月內發生的事。這意味著我們一個月內推出的機器人數量比前幾個季度都多。這是一個關鍵的里程碑。我們在 IPO 期間承諾到年底交付 2000 台機器人，目前進展順利，第 2000 台機器人計劃於 12 月中旬在邁阿密部署。

But we are not going to stop there. We envision a future where Serve's fleet reaches 1 million robots deployed across cities globally. They will travel billions of miles annually. They become embedded into the core fabric of a modern city, and they unlock new possibilities.

但我們不會就此止步。我們設想，未來 Serve 的機器人數量將達到 100 萬台，部署在全球各個城市。他們每年將飛行數十億英里。它們融入現代城市的核心結構，並開啟新的可能性。

Our conviction is simple. We are entering the age where things will move at our will but on their own. Autonomy will become this essential infrastructure in our lives. It's rarely noticed because it just works, but it's sorely missed when it's not available. And we want to be the company that you will trust to run this.

我們的信念很簡單。我們正進入一個萬物將按照我們的意願自主運作的時代。自主性將成為我們生活中不可或缺的基礎。它很少被注意到，因為它一直運作良好，但當它無法使用時，人們會非常想念它。我們希望成為您信賴的、能夠承接這項業務的公司。

On the path to 1 million robots, we are still early. But with 1,000 robots operating from coast to coast, we've really just crossed that chasm where the technology and the market all say go. From here, every additional robot, every additional hour, every additional block makes the whole Serve ecosystem more essential and more valuable to the entire network. We are building something durable, and we are just getting started.

距離擁有100萬台機器人的目標還有很長的路要走。但是，隨著 1000 台機器人從東海岸到西海岸投入運營，我們真正跨越了技術和市場都認可的鴻溝。從這裡開始，每一個額外的機器人、每一個額外的小時、每一個額外的模組，都會使整個 Serve 生態系統對整個網路變得更加重要和更有價值。我們正在打造一個經久耐用的事業，而這只是個開始。

With that, I'll turn it over to Brian to cover our Q3 results in more detail.

接下來，我將把麥克風交給布萊恩，讓他更詳細地介紹我們第三季的業績。

Thank you, Ali, and it's great to be with you all today.

謝謝你，阿里，今天能和大家在一起真是太好了。

This quarter marked another step change for Serve, one defined not only by scale, but by strategic execution. We advanced in every meaningful area, expanding our fleet, strengthening our technology base, and executing with greater precision across operations, engineering, and finance.

本季對 Serve 而言又是一次飛躍式的變革，其特點不僅在於規模，更在於策略執行。我們在每個重要領域都取得了進步，擴大了我們的艦隊規模，加強了我們的技術基礎，並在營運、工程和財務方面執行得更精準。

We've also been extremely opportunistic. During the quarter, we've integrated two key acquisitions that deepen our competitive moat. Acquiring Vayu, a pioneer in urban robot navigation using large-scale AI models, we're expanding our physical AI capabilities by accelerating our roadmap. As we further integrate Vayu into our autonomy stack, we expect it to create opportunities to enhance our leadership and autonomous delivery, as well as to reduce data infrastructure costs and improve operational metrics over time. These integrations allow us to convert more of our operational data into faster model improvements and richer monetization layers, all while reinforcing Serve's position as the category's innovation leader.

我們也一直非常善於抓住機會。本季度，我們完成了兩項關鍵收購，進一步鞏固了我們的競爭優勢。透過收購 Vayu（一家利用大規模 AI 模型進行城市機器人導航的先驅企業），我們正在加快推進我們的路線圖，從而擴展我們的實體 AI 能力。隨著我們進一步將 Vayu 整合到我們的自主技術堆疊中，我們期望它能創造機會，增強我們的領導地位和自主交付能力，並隨著時間的推移降低資料基礎設施成本和改善營運指標。這些整合使我們能夠將更多營運數據轉化為更快的模型改進和更豐富的貨幣化層，同時鞏固 Serve 作為該領域創新領導者的地位。

Our focus remains clear: to scale efficiently, deploy capital strategically, and translate our growing operational advantage into sustainable financial performance, all in service of building an enduring business in this new age of autonomy and physical AI. As Ali described, the Serve flywheel is accelerated: more robots, richer data, smarter AI, and stronger economics.

我們的目標仍然明確：高效擴大規模，策略性地部署資本，並將我們不斷增長的營運優勢轉化為可持續的財務業績，所有這些都是為了在這個自主和物理人工智慧的新時代打造一家永續經營的企業。正如阿里所描述的那樣，Serve 的飛輪效應正在加速：更多的機器人、更豐富的數據、更聰明的人工智慧和更強大的經濟效益。

Let's dig into our Q3 results, showcasing how these effects translate into measurable financial impact and expanding leverage across our business. Total revenue for Q3 2025 was $687,000, an increase of 210% versus last year and in line with our guidance provided for the quarter. Fleet revenue was $433,000. Significantly, this quarter, we saw branding revenue jump 120% sequentially over Q2. As we've mentioned previously, the growth of our robot fleet into the thousands unlocks a pipeline of large-scale branding opportunities, and we delivered on that in Q3.

讓我們深入分析第三季業績，展示這些影響如何轉化為可衡量的財務影響，並擴大我們業務的槓桿作用。2025 年第三季總營收為 687,000 美元，比去年同期成長 210%，符合我們對該季度的預期。船隊收入為 433,000 美元。值得注意的是，本季品牌推廣收入較第二季環比成長了 120%。正如我們之前提到的，我們的機器人數量成長到數千台，開啟了大規模品牌推廣機會的管道，而我們在第三季實現了這一目標。

Software revenues continued to transition from one-time to recurring and were $254,000 in the quarter. We delivered exactly what we said we would. Fleet revenue is becoming the predictable growth engine we've envisioned, and we're now meaningfully stacking platform and data services on those same routes.

軟體收入繼續從一次性收入轉向經常性收入轉變，本季達到 254,000 美元。我們完全兌現了承諾。機隊收入正成為我們所設想的可預測的成長引擎，我們現在正在這些航線上有效地疊加平台和數據服務。

Gross margin performance this quarter reflected the balance between rapid fleet expansion and deliberate investment in our long-term efficiency infrastructure. As planned, we continue to build capacity ahead of 2026 scale, expanding our operations footprint, onboarding new cities, and integrating the systems and teams from our recent acquisitions. These near-term investments are already yielding returns in the form of measurable operational gains across reliability and autonomy.

本季毛利率表現反映了快速擴張船隊和對長期效率基礎設施進行審慎投資之間的平衡。按照計劃，我們將繼續提升產能，為 2026 年的規模目標做好準備，擴大我們的營運範圍，拓展新的城市，並整合我們最近收購的系統和團隊。這些近期投資已經產生了回報，在可靠性和自主性方面帶來了可衡量的營運效益。

Average daily operating hours per robot increased another 12.5% sequentially from Q2, driven by the growing mix of Gen 3 hardware across our fleet. This is a strong leading indicator that each unit is capable of contributing more value.

受我們車隊中第三代硬體組合不斷增長的推動，每台機器人的平均每日運行時間較第二季度環比增長了 12.5%。這是一個強有力的領先指標，顯示每個單位都有能力創造更大的價值。

Robot intervention rates saw meaningful reduction through the quarter, and further, our best-in-class sidewalk autonomy is getting more and more capable. We saw a consequential increase in the proportion of miles driven in autonomous mode in the last week of Q3 compared to the first week of the quarter, indicating a return on continued R&D investment.

本季機器人幹預率顯著下降，而且，我們一流的人行道自主系統也變得越來越強大。與第三季第一週相比，第三季最後一週自動駕駛模式下的行駛里程比例顯著增加，這表明持續的研發投資獲得了回報。

Taken together, these factors drove higher autonomous run times which, in turn, drive improvements in our average speed. This leads to compounding gains. Even a small increase in the average speed corresponds to an increase in our potential delivery volumes. These efficiency improvements are compounding. Each additional robot, each additional mile, and each new market contributes data that sharpens our models and reduces human touchpoints across the network.

綜合來看，這些因素提高了自動駕駛時間，進而提高了我們的平均速度。這將帶來複利效應。即使平均速度略有提高，我們的潛在配送量也會隨之增加。這些效率提升正在不斷累積。每增加一台機器人，每增加一英里，每開拓一個新市場，都會貢獻數據，從而改進我們的模型，並減少整個網路中的人工接觸點。

On the expense side, we remain disciplined, investing in the capabilities that drive our competitive advantage. GAAP operating expenses for Q3 were $30.4 million, increasing from Q2 from deliberate investments in new market launches, M&A integrations, and expanded operational capabilities to support our national scale. On a non-GAAP basis, operating expenses were $21.8 million.

在支出方面，我們保持自律，投資於能夠提升我們競爭優勢的能力。第三季 GAAP 營運費用為 3,040 萬美元，較第二季有所增加，原因是我們在新市場推出、併購整合以及擴大營運能力以支持我們的全國規模方面進行了有意識的投資。以非GAAP準則計算，營運費用為2,180萬美元。

R&D remained our largest investment area, totaling $13.4 million on a GAAP basis or $10.7 million on a non-GAAP basis, primarily directed towards advancing our autonomy stack, expanding our AI foundation models, and integrating new data and hardware capabilities from our recent acquisitions. These initiatives are accelerating our pace of innovation while positioning us for long-term cost structure.

研發仍然是我們最大的投資領域，按 GAAP 計算總計 1340 萬美元，按非 GAAP 計算總計 1070 萬美元，主要用於推進我們的自主技術棧，擴展我們的 AI 基礎模型，以及整合我們最近收購的新數據和硬體功能。這些措施正在加快我們的創新步伐，同時為我們實現長期成本結構奠定基礎。

G&A and go-to-market spending remain disciplined and aligned with our city expansion cadence. We're executing with leverage, adding cities, partners, and robots without literally increasing headcount or our overhead. Our approach remains consistent: invest where we have clear line of sight to efficiency, differentiation, and scale advantage while maintaining financial discipline and measured growth.

一般及行政費用和市場推廣支出保持嚴格控制，並與我們的城市擴張節奏保持一致。我們正在利用槓桿效應，在不實際增加員工人數或營運成本的情況下，增加城市、合作夥伴和機器人。我們的方法始終如一：在能夠帶來效率、差異化和規模優勢的領域進行投資，同時保持財務紀律和穩健成長。

On the balance sheet, we ended the quarter with $211 million in cash and marketable securities. In October, we executed a stock sale that generated approximately $100 million, which will be used to fund working capital and expansion activities. CapEx for the quarter was $11 million, tied to robot production, market launch, and expansion infrastructure. Our strong liquidity and debt-free balance sheet remain a competitive advantage, providing us flexibility to scale responsibly and invest opportunistically. Adjusted EBITDA was negative $24.9 million, driven by operational expansion in the quarter, expected to accelerate efficiency through 2026.

資產負債表顯示，本季末我們持有現金及有價證券2.11億美元。10月份，我們進行了股票出售，籌集了約1億美元，這些資金將用於補充營運資金和進行擴張活動。本季資本支出為1,100萬美元，主要用於機器人生產、市場推廣及基礎建設。我們強勁的流動性和零負債的資產負債表仍然是我們的競爭優勢，使我們能夠靈活地進行負責任的規模擴張和掌握投資機會。經調整後的 EBITDA 為負 2,490 萬美元，主要受本季營運擴張的影響，預計到 2026 年將進一步提高營運效率。

And now to our outlook. Once again, we delivered results at the high end of our Q3 guidance range. Building on this momentum, we now expect to generate more than $2.5 million in revenue for the full-year 2025. Our underlying recurring fleet revenues, which exclude non-recurring software, is projected to grow 3x year over year, from roughly $0.6 million in 2024 to roughly $2.1 million in 2025.

現在談談我們的展望。我們再次實現了第三季業績預期範圍的高端目標。憑藉這一勢頭，我們現在預計 2025 年全年收入將超過 250 萬美元。（不包括非經常性軟體收入）我們的基本經常性車隊收入預計將逐年增長 3 倍，從 2024 年的約 60 萬美元增長到 2025 年的約 210 萬美元。

2025 was a pivotal year focused on establishing our national footprint, deploying 2,000 robots, expanding into new markets, and deepening our partnership portfolio. With this groundwork in place, we remain confident in our ability to generate annualized revenue run rate of $60 million to $80 million. We intend to update 2026 full-year guidance early next year. Initial indications show our expansion and operational plan positions Serve to deliver roughly 10x inflection in revenue during 2026.

2025 年是關鍵的一年，重點是建立我們的全國影響力，部署 2000 台機器人，拓展新市場，並深化我們的合作關係。有了這些基礎，我們仍然有信心實現每年 6,000 萬至 8,000 萬美元的營業收入。我們計劃於明年初更新 2026 年全年業績預期。初步跡象表明，我們的擴張和營運計劃定位將在 2026 年實現收入約 10 倍的成長。

Q3 marked another step forward in both scale and precision. We're executing with discipline, expanding intelligently, and translating operational progress into tangible financial results. Each quarter, our fleet becomes more capable, models more refined, and economics more efficient. The foundation we've built across technology, partnerships, operational excellence position us for sustainable growth through 2026.

第三季在規模和精度方面都取得了另一個進步。我們以嚴謹的紀律執行，以明智的方式擴張，並將營運進展轉化為實際的財務成果。每個季度，我們的機隊性能都會提升，機型會更加精良，經濟效益也會更高。我們在技術、合作夥伴關係和卓越營運方面建立的基礎，使我們能夠在 2026 年實現永續成長。

We're proud of what the team has accomplished this quarter and even more excited about the opportunities ahead. Serve is defining this category, and we're confident in our ability to lead it for years to come.

我們為團隊本季所取得的成就感到自豪，也對未來的機會感到更加興奮。Serve 正在定義這個類別，我們有信心在未來幾年繼續引領這個類別。

With that, I'll hand it back to Aduke for Q&A.

接下來，我將把問題交還給阿杜克，由他來回答問題。

Thank you, Ali and Brian. We will now move into the Q&A session. First, I'd like to say a big thank you to all the investors and analysts who submitted questions via email. We really appreciate your engagement.

謝謝阿里和布萊恩。接下來進入問答環節。首先，我要衷心感謝所有透過電子郵件提交問題的投資人和分析師。我們非常感謝您的參與。

First question, I think this might be for Ali. Do you expect to add more robots in 2026? If so, what would be timing and magnitude of the addition? Ali?

第一個問題，我覺得這可能是問阿里的。你們預計2026年增加更多機器人嗎？如果屬實，新增規模和時間安排會是怎麼樣的？阿里？

Thank you, Aduke. Yeah, I can take this one. Good question.

謝謝你，阿杜克。好的，我可以應付。問得好。

We aren't going to share the specific numbers right now. Hopefully, we have more to share early next year. But I do want to explain how we are thinking about growth. As we are looking to get towards our 1 million robots goal, what we want to do is make sure we grow quickly but also with precision and discipline.

我們現在不打算透露具體數字。希望明年年初我們能與大家分享更多。但我確實想解釋一下我們是如何看待成長的。為了實現100萬台機器人的目標，我們希望確保快速發展，同時也要保持精準和嚴謹。

We've been really laser-focused in getting the fleet really efficient and effective every day and driving utilization, while at the same time layering new partners, going to new geographies. All of this makes that scale-up easier. So in a way, being efficient and growth kind of line up together. So in that sense, it's the same type of effort that it takes to get there. So we are definitely going to push on growth, but we want to do it responsibly.

我們一直非常專注於提高車隊的日常營運效率和利用率，同時不斷拓展新的合作夥伴，進軍新的地區。所有這些都使得規模化生產更容易。所以從某種意義上來說，效率和成長是相輔相成的。從這個意義上講，達到那個目標所需的努力是相同的。所以我們肯定會推進成長，但我們希望以負責任的方式進行成長。

All right. Thank you. Next question is about robot design. Could you provide details on robot design simplification and cost reduction beyond economies of scale? Ali, do you want to take this one?

好的。謝謝。下一個問題是關於機器人設計的。除了規模經濟之外，能否詳細介紹如何簡化機器人設計並降低成本？阿里，你想接這個嗎？

Yeah, I'll take this one too. I think there's a few different factors here. First of all, there is a ton of progress that we've made when it comes to the robot design. We've made it a lot more modular, easier to manufacture, fewer custom assemblies. We've also really strengthened our supply chain to get better parts at lower prices. So this both cuts down the cost of the material, but also cost of assembly.

嗯，這個我也要。我認為這裡有幾個不同的因素。首先，我們在機器人設計方面取得了巨大的進步。我們使其模組化程度更高，更易於製造，並減少了客製化組件。我們也大力加強了供應鏈，以便以更低的價格獲得更好的零件。這樣既降低了材料成本，也降低了組裝成本。

At the same time, as we improve our design, we've also benefited from our scaled manufacturing, which obviously helps bring the cost down as well. And while all of that is happening, the broader kind of ecosystem of suppliers, they're also getting more mature.

同時，隨著我們設計的改進，我們也受益於規模化生產，這顯然也有助於降低成本。同時，更廣泛的供應商生態系統也變得更加成熟。

I think a really good example that I'm excited about is Ouster. They have done a phenomenal job bringing these advanced LiDAR sensors to market at scale. They're shipping record number of sensors right now -- I think thousands per quarter -- and we are directly benefiting from that. And I think a lot of folks in the autonomy space would benefit from more affordable LiDAR sensors that just didn't seem within that realm just a few years ago.

我覺得 Ouster 就是一個很好的例子，我很興奮。他們出色地完成了將這些先進的光達感測器大規模推向市場的工作。他們現在正在出貨創紀錄數量的傳感器——我認為每季出貨數千個——我們正從中直接受益。我認為，自動駕駛領域的許多人都會受益於價格更親民的光達感測器，而幾年前，這種感測器似乎還不屬於這個領域。

So combining our improvements to the design and our improved supply chain and our scaled manufacturing and the maturity of the ecosystem, that per unit cost of the robots is definitely coming down substantially to the point that, as we've shared in the past, our Gen 3 robots are third the cost of our Gen 2 robots. And we are going to keep pushing these improvements forward.

因此，結合我們對設計的改進、我們改進的供應鏈、我們規模化的製造以及生態系統的成熟，機器人的單位成本肯定會大幅下降，正如我們過去所分享的那樣，我們的第三代機器人的成本只有第二代機器人的三分之一。我們將繼續推動這些改進措施。

Okay. Thank you for that. Next question, what are the next steps in your DoorDash relationship? How do you see that helping the business? Ali?

好的。謝謝。下一個問題，您與 DoorDash 的合作關係下一步是什麼？您認為這對企業有何幫助？阿里？

Yeah. We are working very close with our partner at DoorDash. First and foremost, it's about integrating the robots into the fleet in a thoughtful way and planning the market rollouts over time. DoorDash obviously unlocks an enormous network of restaurants and consumers for us. We have over 1,000 robots right now and soon 2,000 that can deliver for those restaurants and customers. So the timing is perfect. And I expect that in the next few months, we will start to really grow the volume under the new channel with DoorDash, basically.

是的。我們與合作夥伴 DoorDash 的合作非常緊密。首先，最重要的是以周全的方式將機器人融入車隊，並隨著時間的推移規劃市場推廣。DoorDash 顯然為我們打開了一個龐大的餐廳和消費者網絡。我們現在有超過 1000 台機器人，很快就會達到 2000 台，可以為這些餐廳和顧客提供服務。所以時機剛剛好。我預計在接下來的幾個月裡，我們將透過與 DoorDash 合作的新管道真正開始擴大業務量。

I do want to emphasize this is a really important milestone for us because we've always envisioned this multi-platform approach. I think a single robot being able to alternate between deliveries from each platform, from DoorDash and Uber, it is really, really important that we are able to do that. And I think we are now proving that we can. And this kind of interoperability actually increases our utilization which, in turn, lowers the cost per delivery. And that actually benefits all of our partners as well.

我想強調，這對我們來說是一個非常重要的里程碑，因為我們一直以來都設想採用這種多平台方法。我認為，讓一台機器人能夠交替地從 DoorDash 和 Uber 這兩個平台進行配送，這點真的非常重要。我認為我們現在正在證明我們能夠做到這一點。這種互通性實際上提高了我們的利用率，進而降低了每次交付的成本。這其實也有利於我們所有的合作夥伴。

Perfect. Thank you. We have a question on acquisition. Can you quantify the autonomy effect from Vayu? For example, would average speed increase or would the ratio of robots to operators improve?

完美的。謝謝。我們有一個關於收購的問題。你能量化Vayu帶來的自主效應嗎？例如，平均速度會提高嗎？或者機器人與操作員的比例會提高嗎？

Brian, do you want to take this one?

布萊恩，你想接手這個任務嗎？

Yeah. So yeah, good question. And I mean, I think the simple answer to start here is we're very early in this integration process to dive into those results exactly. And this is the type of integration that can take months. But we're actually doing the call here today from Vayu offices, and the excitement from the teams to hit the ground running as soon as the merger was completed was tremendous. And there's just a lot of excitement on both sides to go faster and deeper into that roadmap to bring those new capabilities into the fleet.

是的。是的，問得好。我的意思是，我認為最簡單的答案是，我們目前還處於整合過程的早期階段，還無法深入分析這些結果。這種整合可能需要數月時間。但實際上，我們今天是在Vayu辦公室進行這次電話會議的，合併完成後，各團隊都很興奮地想要立即投入工作。雙方都非常興奮，希望加快步伐，深入推進該路線圖，將這些新功能引入艦隊。

I think we think about it as part of a flywheel where, over time, that integration will allow our robots to be faster and smarter while maintaining the safety and reliability that we focus on daily. And that, in turn, then drives efficiency and utilization, ultimately landing in unit economics and overall the benefit for these acquisitions.

我認為我們可以把它看作是飛輪的一部分，隨著時間的推移，這種整合將使我們的機器人更快、更智能，同時保持我們每天關注的安全性和可靠性。而這反過來又會提高效率和利用率，最終體現在單位經濟效益和這些收購的整體收益上。

Okay. Thank you. Our next question, what are some differences between deployments in different cities? What have you learned from new deployments and expansions that will help you scale further?

好的。謝謝。我們的下一個問題是，不同城市的部署有哪些差異？從新的部署和擴展中，您學到了哪些有助於您進一步擴展規模的經驗？

Ali, can you take this one?

阿里，你能接這個嗎？

Absolutely, yeah. Each city has its own distinct personality. It's like they're different in ways that's actually very helpful for us and honestly quite fun for our team as we've been expanding.

當然，沒錯。每個城市都有其獨特的個性。它們在某些方面有所不同，這實際上對我們非常有幫助，而且說實話，隨著我們團隊的擴張，這也讓我們的團隊感到非常有趣。

Seeing, for example, in Atlanta and Miami, learning about humidity and the different kinds of pedestrian intersections and city design compared to what we had in Los Angeles before. They have different width in their sidewalks, different nuances about how the best routes for traversing would actually look like. Or our new market in Chicago, it's an incredible place for getting data on really dense urban environments with cold weather where you have to look at battery efficiency and snow detection and traction. So a lot of things that we have tried to test in advance now are being put to test in real life, and we are learning a lot from that.

例如，在亞特蘭大和邁阿密，我們了解到當地的濕度、不同類型的行人交叉路口以及城市設計，並與我們之前在洛杉磯的情況進行了比較。他們的人行道寬度不同，最佳通行路線的實際形狀也各有不同。或者我們在芝加哥的新市場，這是一個獲取真正密集的城市環境數據的絕佳場所，那裡氣候寒冷，你必須考慮電池效率、雪檢測和牽引力。所以，我們之前嘗試測試的很多東西現在都在現實生活中接受檢驗，我們從中也學到了很多。

What's really powerful, I think, is that as we go to these new cities, it really enriches the models, and the data in these new environments actually helps the model across the board for the entire platform. And that actually means that every subsequent city launch, as I mentioned earlier, gets more reliable and better. In fact, we saw this in Chicago when we first launched. It was the fastest market for us to get to our SLAs that were comparable to our more mature markets. So it kind of proves that the playbook is working well and the robots are getting smarter.

我認為真正強大的地方在於，當我們進入這些新的城市時，模型會得到極大的豐富，這些新環境中的數據實際上對整個平台的模型都有幫助。正如我之前提到的，這意味著後續每個城市的推出都會變得更加可靠和完善。事實上，我們在芝加哥首次推出產品時就看到了這一點。這是讓我們最快達到與更成熟市場相當的服務等級協定 (SLA) 的市場。這在某種程度上證明了這套方案行之有效，機器人也變得越來越聰明。

And just to finish that, from a financial standpoint, I think, too, is we're using these learnings. We're translating them directly into efficiency through our operations teams. So we're seeing shorter payback periods with the expansion. We're seeing the higher utilization as we deploy into new markets and neighborhoods to continue the expansion.

最後，從財務角度來看，我認為我們也在運用這些經驗教訓。我們正透過營運團隊將這些目標直接轉化為效率。因此，我們看到隨著擴張，投資回收期縮短了。隨著我們向新市場和社區部署以繼續擴張，我們看到了更高的利用率。

And that's exactly what we're building towards. So we've been talking internally about describing this as being sharper with our scale. And we believe all of these technology improvements are going to compound, which will show up in our financial results. So that is really what's positioning us to expand in a disciplined, capital-efficient way in 2026.

而這正是我們努力的方向。所以我們內部一直在討論如何用更清晰的尺度來描述這一點。我們相信所有這些技術進步都將產生疊加效應，這將在我們的財務表現中反映出來。因此，這確實使我們能夠在 2026 年以有紀律、有資本效率的方式擴張。

Okay. Thank you. Next question, what can you share about the pipeline for software and data sales? How are you looking to accelerate software revenues in 2026 and beyond?

好的。謝謝。下一個問題，可以分享一下軟體和數據銷售管道的情況嗎？您計劃如何加速2026年及以後的軟體收入成長？

Yeah, this is a good question. The revenue pipeline for these other opportunities, like the delivery platform, the software that's powering the robots, as well as the data that's generated by the robots, it's been a really strong pipeline. We are in substantial discussions with multiple partners that want to basically use the platform or the data that we are creating.

是的，這是一個很好的問題。其他機會的收入來源，例如配送平台、驅動機器人的軟體以及機器人產生的數據，都是一個非常強勁的收入來源。我們正在與多家合作夥伴進行深入洽談，他們希望使用我們正在創建的平台或數據。

And I think we are trying to be smart and selective in terms of who we engage with and apply some filters there to pick the right partners. But the amount of inbound interest we're getting really reinforces that what we have is quite differentiated. And I'm hoping that as we move some of these conversations forward and have more updates to share, we'll actually tell you more about those relationships as well.

我認為我們正在努力做到精明和有選擇性，在選擇合作夥伴方面會運用一些篩選條件，以挑選合適的合作夥伴。但我們收到的諮詢量確實印證了我們所擁有的產品和服務確實與眾不同。我希望隨著我們推進這些對話，並有更多最新進展可以分享，我們也能向大家詳細介紹這些關係。

And if I can, to also wrap the financial aspect into this question, as the fleet scales, these data and AI insights are going to become more valuable to all of the people Ali just mentioned that we're talking to and the substantive discussions we're having. So that's going to enable our team to look at opportunities for adding more recurring software as we go throughout 2026 and focus on that robotics and autonomy-as-a-service offering.

如果可以的話，我還想把財務方面也納入這個問題。隨著艦隊規模的擴大，這些數據和人工智慧見解對於阿里剛才提到的所有我們正在交談的人以及我們正在進行的實質討論來說，將變得更加有價值。這樣一來，我們的團隊就可以在 2026 年全年尋找機會，增加更多循環軟體，並專注於機器人和自主即服務產品。

So it's really a long-term vision. This is a balanced model. We're focused on diversifying revenue, and fleet revenue is that foundation, with software and data as that real high-margin accelerant that we're focused on as we enter 2026.

所以這確實是一個長遠的願景。這是一個平衡的模型。我們專注於實現收入多元化，車隊收入是基礎，而軟體和數據則是我們進入 2026 年時重點關注的真正高利潤加速器。

Okay. And our last question, you mentioned the $60 million to $80 million run rate. When do you expect to reach that run rate?

好的。最後一個問題，您提到了 6000 萬至 8000 萬美元的年化收入。你預計何時能達到這個運行率？

Brian, can you take this one?

布萊恩，你能接這個嗎？

Yeah. Let me give a little bit more color. I know in the script, we did mention the outlook for 2026. So we'll point everybody back to that commentary. But obviously, this is a good question to end on here as we think about this Q3 update.

是的。讓我再添一點細節。我知道在劇本中，我們確實提到了 2026 年的展望。所以我們會引導大家回顧那篇評論。但很顯然，在我們思考第三季業績更新時，這是一個很好的結尾問題。

So the path to hitting $60 million to $80 million is underway, and that's really the final step. When we think about the ambitions we laid out a few years ago to deliver 2,000 robots, $60 million to $80 million is the end point. But along the way, we've exceeded a lot of expectations, especially as we near the end of 2025. And that's been a testament to the team and how we've delivered.

因此，實現 6,000 萬至 8,000 萬美元的目標已經走上正軌，而這真的是最後一步了。當我們回想幾年前我們制定的交付 2000 個機器人的宏偉目標時，最終需要 6000 萬至 8000 萬美元的資金。但一路走來，我們超越了許多預期，尤其是在 2025 年即將結束之際。這充分證明了團隊的實力和我們的工作成果。

I mean, to summarize what we've talked about on a lot of the earnings calls, we are on track to deliver the 2,000 robots. We've expanded into multiple markets with more coming. We talked about new partnerships and adding top-of-the-funnel orders into our pipeline. But last, but not least, we have the acquisition. So across all of these verticals, we are firing and we're really exceeding what we set out to achieve in 2025.

我的意思是，總結一下我們在許多財報電話會議上討論的內容，我們正按計劃交付 2000 台機器人。我們已拓展至多個市場，未來也將拓展更多市場。我們討論了新的合作關係，以及如何將漏斗頂端的訂單加入我們的銷售管道中。但最後，也是非常重要的一點，我們還有收購。因此，在所有這些垂直領域，我們都取得了巨大成功，並且遠遠超出了我們在 2025 年設定的目標。

I'd like to remind investors and anybody that wants to understand our story that that's all great. But critically, we're maintaining the safety and the reliability throughout that network as we're building it. And so the final boss, as Ali likes to say, is achieving that financial milestone is continuing to improve the utilization across the fleet. And so we have that momentum through 2025 and accelerating into 2026 to approach that $60 million run rate target.

我想提醒投資者和所有想了解我們故事的人，這一切都很棒。但至關重要的是，我們在建設過程中始終保持整個網路的安全性和可靠性。因此，正如阿里喜歡說的那樣，最終目標就是實現財務里程碑，並持續提高整個機隊的使用率。因此，我們將保持這一勢頭直至 2025 年，並​​在 2026 年加速，以接近 6,000 萬美元的年化收入目標。

To be clear, I think we're still more than 12 months out. And we'll certainly -- as we indicated, we'll have more to say on this in the next call early next year.

說實話，我覺得距離目標日期還有超過12個月的時間。我們當然會——正如我們之前所說——在明年年初的下次電話會議上就此發表更多看法。

Okay. Thanks so much. That's all the time we have for today. And that concludes our session. Thank you for your thoughtful questions and participation. And with that, I hand it over to the operator.

好的。非常感謝。今天的時間就到這裡了。我們的會議到此結束。感謝您提出的寶貴問題和積極參與。現在，我把麥克風交給接線生。

That concludes today's call. You may now disconnect.

今天的電話會議到此結束。您現在可以斷開連線了。