Cerebras Systems Inc (CBRS) 2026 Q2 法說會逐字稿

Good afternoon, and welcome to the Cerebras Systems second-quarter fiscal year 2026 earnings conference call.

下午好，歡迎參加 Cerebras Systems 2026 會計年度第二季財報電話會議。

(Operator Instructions)

(接線員指示)

I will now turn the call over to Sean Dorsey, head of investor relations. Please go ahead.

現在我將把電話交給投資人關係主管 Sean Dorsey。請開始。

Thank you, operator. Good afternoon, everyone, and welcome to Cerebras Systems' Q2 2026 earnings call. Earlier today, we issued our press release and posted our supplemental earnings presentation to the investor relations section of our website.

謝謝，接線員。各位下午好，歡迎參加 Cerebras Systems 2026 年第二季（Q2）財報電話會議。今天稍早，我們已發布新聞稿，並將補充財報簡報上傳至我們網站的投資人關係專區。

A replay of this webcast will also be available on our investor relations website following the call.

本次網路直播的重播也將於會後在我們的投資人關係網站提供。

Joining me today are Andrew Feldman, our Co-Founder, Chief Executive Officer and President, and Bob Komen, our Chief Financial Officer. Before we begin, I would like to remind everyone that today's discussion will include forward-looking statements under the safe harbor of the Private Securities Litigation Reform Act of 1,995.

今天與我一同出席的有共同創辦人、執行長兼總裁 Andrew Feldman，以及財務長 Bob Komen。在開始之前，我想提醒各位，今天的討論將包含依據《1995 年私人證券訴訟改革法》安全港條款所定義的前瞻性陳述。

These statements include, but are not limited to, statements regarding our future financial performance. Business strategy, market opportunity, customer demand, product roadmap, technology leadership, supply chain, operating model, and outlook for Q3 and full year 2026. Forward-looking statements are based on our current expectations and assumptions and are subject to risks and uncertainties that could cause actual results to differ materially from those expressed or implied.

這些陳述包括但不限於：我們未來的財務表現、商業策略、市場機會、客戶需求、產品藍圖、技術領先地位、供應鏈、營運模式，以及對 2026 年第三季與 2026 全年的展望。前瞻性陳述係基於我們目前的預期與假設，並受風險與不確定性影響，可能導致實際結果與明示或暗示的內容有重大差異。

These risks are described in our SEC filings. Including our final prospectus related to our IPO and our future periodic filings with the SEC. We undertake no obligation to update these forward-looking statements except as required by law.

這些風險已在我們向美國證券交易委員會（SEC）提交的文件中說明，包括與我們首次公開發行（IPO）相關的最終公開說明書，以及我們未來定期向 SEC 提交的文件。除法律要求外，我們不承擔更新這些前瞻性陳述的義務。

During today's call, we will also discuss certain non-GAAP financial measures.

在今天的電話會議中，我們也將討論若干非 GAAP 財務衡量指標。

Reconciliations between GAAP and non-GAAP results are included in today's press release and supplemental materials, which are available on the Investor Relations page of our website.

GAAP 與非 GAAP 結果之間的調節表已包含在今天的新聞稿與補充資料中，相關資料可於我們網站的投資人關係頁面取得。

With that, I'll turn the call over to Andrew.

接下來，我把電話交給 Andrew。

Thank you, Sean.

謝謝你，Sean。

Thank you all for joining us today. Q2 was a strong quarter.

感謝各位今天加入我們。第二季表現強勁。

We completed our public offering, but we did not let that distract us from execution. We delivered record core revenue and beat guidance on all metrics, core revenue, core gross margins, and core operating margin.

我們完成了公開發行，但並未因此分心而影響執行。我們交出創紀錄的核心營收，並在所有指引指標上超越預期，包括核心營收、核心毛利率與核心營業利益率。

Looking forward, we see unbound demand for fast inference.

展望未來，我們看到對高速推論的需求毫無上限。

The market is realizing that speed is not a benchmark item. Speed changes user engagement, it changes agentic performance, and it changes AI productivity.

市場正在意識到，速度不是一個基準測試項目而已。速度會改變使用者互動、改變代理式（agentic）效能，也會改變 AI 生產力。

Fast inference unlocks new applications and new markets. As we've shared with you previously, 2026 is a foundation building year for Cerebras. We've made excellent progress on multiple fronts in the past seven weeks since our last earnings call, preparing us for a massive 2027, 2028 and 2029 as we deliver on the $25 billion of RPO we currently have on our books.

高速推論將解鎖新的應用與新的市場。如同我們先前與各位分享的，2026 年是 Cerebras 打地基的一年。自上次財報電話會議以來的七週內，我們在多個面向取得極佳進展，為 2027、2028 與 2029 的大幅成長做好準備，並將兌現目前帳上 250 億美元的 RPO（剩餘履約義務）。

With the benefit of that progress, we expect to more than triple our core revenues in '27 and continue to grow at multiples in the years following.

在這些進展的基礎上，我們預期 2027 年核心營收將超過三倍成長，並在其後年度持續以倍數成長。

We think of progress in terms of capacity, capabilities, and customers. We're expanding capacity by adding new contracts for data centers around the world, expanding manufacturing capabilities and collaborating with our vendors to ensure supply and to support our extraordinary growth.

我們以產能、能力與客戶三個面向來衡量進展。我們正透過新增全球各地資料中心合約來擴充產能、擴大製造能力，並與供應商合作以確保供應，支援我們非凡的成長。

We're advancing our capabilities by inventing new technology that extends our performance and throughput and our power efficiency and we're expanding our customer base by accelerating AI productivity in existing markets like coding and agentic flows and pioneering new areas like security where speed opens up entirely new opportunities.

我們透過發明新技術來提升能力，延伸我們的效能與吞吐量以及能源效率；同時，我們也透過在既有市場（如程式撰寫與代理式流程）加速 AI 生產力來擴大客戶基礎，並在如資安等新領域開疆闢土，在那裡速度能開啟全新的機會。

On the capacity front, data center space continues to be the bottleneck for the entire industry, and we are no exception. The faster we and our customers bring on new data centers, the faster we grow. So over the last seven months, we've had an all-out push to secure and build out data centers.

在產能方面，資料中心空間仍是整個產業的瓶頸，我們也不例外。我們與客戶越快啟用新的資料中心，我們成長就越快。因此在過去七個月，我們全力推動取得並建置資料中心。

We have two advantages. First, because we are serving inference, we do not need gigawatt footprint locations like those needed for training clusters. This gives us much more flexibility to scale up. Capacity across a multitude of locations around the world.

我們有兩項優勢。第一，因為我們提供的是推論服務，我們不需要像訓練叢集那樣的吉瓦級占地據點。這讓我們在擴張時更具彈性，能在全球多個地點擴充產能。

Second, we built a repeatable process for site selection, cluster deployment, and customer activation, which is an operational muscle required to turn gigawatts into production tokens at a global scale.

第二，我們建立了一套可重複的流程，用於站點選址、叢集部署與客戶啟用；這是一種營運肌肉，能把吉瓦級電力轉化為全球規模的生產級 token。

I'm pleased to report that our push has been very successful. We now have data centers either up or under contract in Alabama, Dallas, Denver, Minneapolis, Santa Clara, Stockton, and outside the US in France, Finland, Manitoba, Montreal, Norway, Saskatchewan, and Toronto.

我很高興報告，我們的推進非常成功。目前我們在阿拉巴馬州、達拉斯、丹佛、明尼阿波利斯、聖塔克拉拉、史塔克頓等地，以及美國境外的法國、芬蘭、曼尼托巴、蒙特婁、挪威、薩斯喀徹溫與多倫多，皆已有資料中心上線或已簽約。

In total, over the last seven months, we have secured more than 600 megawatts of data center capacity that is either live now or will be delivered by the end of 2027.

總計在過去七個月，我們已取得超過 600 兆瓦的資料中心產能，這些產能目前已上線，或將於 2027 年底前交付。

And while this isn't nearly enough to meet our demand, our data center pipeline of new opportunities for expansion continues to grow and is now measured in gigawatts.

雖然這仍遠不足以滿足需求，但我們資料中心擴張的新機會管線持續成長，目前已以吉瓦為單位衡量。

To put it in perspective, as we continue to build out our first-party cloud, it will be among the largest non-hyperscale AI clouds and whereas at the end of 2025, we're on a steep learning curve, today I'm happy to report that we're pretty good at data center buildout with a clear path. To becoming excellent.

從更宏觀的角度來看，隨著我們持續擴建第一方雲端，它將成為最大規模的非超大規模（non-hyperscale）AI 雲之一；而在 2025 年底我們仍處於陡峭的學習曲線上，如今我很高興地報告，我們在資料中心建置方面已相當熟練，並且有清晰的路徑邁向卓越。

Other key dimensions of capacity include manufacturing and supply chain.

產能的其他關鍵面向包括製造與供應鏈。

Here we've successfully increased our manufacturing capacity and are building up new factories with Flex and San Mina and expect to increase our manufacturing capacity by more than 10x in 2026 and continue that expansion in 2027. Again, preparing us for the exceptional growth expected in the years ahead.

在這方面，我們已成功提升製造產能，並與 Flex 與 San Mina 建置新工廠，預期 2026 年我們的製造產能將提升超過 10 倍，並在 2027 年持續擴張。再次強調，這是在為未來幾年預期的卓越成長做準備。

Our partnership with our supply chain vendors has also turned into a significant advantage. TSMC has once again come through, and we have the wafers needed to fuel our growth.

我們與供應鏈廠商的合作也已轉化為一項顯著優勢。台積電（TSMC）再次鼎力相助，我們已取得推動成長所需的晶圓。

Our ability to get wafer supply also benefits from the fact that we were able to deliver industry-leading performance while running on TSMC's 5 nanometer node, where wafers are less expensive and supply is less constrained.

我們取得晶圓供應的能力也受益於一項事實：我們能在台積電 5 奈米製程節點上提供業界領先的效能；在該節點下晶圓成本較低、供應也較不吃緊。

Our decades-long relationships with our supply partners reinforces our confidence that we can deliver on our growth plans going forward. These relationships are rare and valuable, particularly in times of short supply.

我們與供應夥伴長達數十年的關係，強化了我們對未來能達成成長計畫的信心。這些關係十分罕見且珍貴，尤其是在供應短缺的時期。

Finally, recall that most of the critical supply chain constraints currently faced by the industry don't apply to us. For example, we don't use HBM memory, co-op packaging, or require 3-nanometer fab capacity.

最後，請記得，目前產業面臨的大多數關鍵供應鏈限制並不適用於我們。例如，我們不使用 HBM 記憶體、CoWoS 封裝，也不需要 3 奈米晶圓廠產能。

On the capabilities front, in the second quarter, we delivered support for OpenAI's GPT-56 SOL, the largest and most capable of the frontier models. In fact, Cerebra serves 56 SOL at a speed that is 10x faster.

在能力方面，第二季我們交付了對 OpenAI 的 GPT-56 SOL 的支援，這是最大且能力最強的前沿模型之一。事實上，Cerebra 以快 10 倍的速度提供 56 SOL 的服務。

With GPT-56 SOL, this lays to rest any of the remaining concerns regarding our ability to support large frontier models.

透過 GPT-56 SOL，這也徹底消除了外界對我們支援大型前沿模型能力的任何剩餘疑慮。

Being a partner for the delivery of GPT-56 SOL and serving it to our cloud speaks to the maturity of our software stack. It takes millions of system hours of production hardening to get to the point where one can deliver hyperscale quality and reliability.

能成為 GPT-56 SOL 交付的合作夥伴，並在我們的雲端提供其服務，顯示我們軟體堆疊的成熟度。要達到能提供超大規模等級的品質與可靠性，需要數百萬小時的系統生產環境強化。

We're proud that our inference cloud can meet the requirements of the most demanding customers. Our collaboration on serving models at the Frontier has opened up new and significant strategic advantage, previously only available to NVIDIA.

我們很自豪，我們的推理雲能夠滿足最嚴苛客戶的需求。我們在 Frontier 前沿模型的服務合作，開啟了全新且重大的策略優勢，而這在過去僅 NVIDIA 才能享有。

Closed source Frontier models include a continual stream of new insights and new AI techniques.

閉源的 Frontier 前沿模型持續帶來源源不絕的新洞見與新 AI 技術。

Serving these models allows us to see into the future and to prepare for it.

提供這些模型的服務，讓我們得以窺見未來並為之做好準備。

Our roadmap from the hardware through the software stack now reflects what we're seeing. And will give us a compounding advantage in the years to come. Continuing on the theme of capabilities, let's turn to disaggregation.

我們從硬體到軟體堆疊的產品路線圖，如今已反映出我們所觀察到的趨勢。並將在未來數年為我們帶來複利式的優勢。延續能力的主題，讓我們來談談解耦（disaggregation）。

We now have disaggregated inference solutions with two of the leading chip companies, AMD with their Helios and AWS with Tranium. Disaggregation expands the market for both the GPU provider and for Cerebras.

我們目前已與兩家領先的晶片公司推出解耦式推理解決方案：AMD（其 Helios）以及 AWS（其 Tranium）。解耦同時擴大了 GPU 供應商與 Cerebras 的市場。

This aggregation enables GPUs to participate in a market currently foreclosed to them, namely fast inference.

這種聚合（aggregation）使 GPU 能夠參與一個目前對其封閉的市場，也就是高速推理。

This aggregation enables Cerebras to expand our opportunity to those customers who are more price sensitive and expands the profitability of our data centers.

這種聚合也讓 Cerebras 能將機會擴展到對價格更敏感的客戶，並提升我們資料中心的獲利能力。

Let's see how this works. As with any compute market, as inference grows and matures, opportunities for specialization emerge. Disaggregation is a form of specialization that is particularly well suited for workloads with well-known traffic patterns. In these cases, disaggregation delivers advantage by separating inference into two stages, prefill and decode, and using different processors for each stage.

讓我們看看它如何運作。如同任何運算市場，隨著推理成長並趨於成熟，專業化的機會便會浮現。解耦是一種專業化形式，特別適合具有明確流量型態的工作負載。在這些情況下，解耦透過將推理分成兩個階段——prefill 與 decode——並為每個階段使用不同處理器，來帶來優勢。

Prefill processes the input from the user or agent. It is a parallelizable workload. As a result, presill is well suited for GPUs and their HBM-based memory architectures. Decode generates the output tokens. It's the harder technical problem and is the bulk of the computational work in a disaggregated solution. It is sequential and memory bandwidth intensive and is particularly well suited for our wafer scale engine.

Prefill 會處理來自使用者或代理（agent）的輸入。它是可平行化的工作負載。因此，prefill 非常適合 GPU 及其以 HBM 為基礎的記憶體架構。Decode 會生成輸出 token。這是較困難的技術問題，且在解耦式方案中占了大部分的運算工作。它是序列式且高度依賴記憶體頻寬，特別適合我們的晶圓級引擎（wafer scale engine）。

The prefill and decode processors need to be linked to create the end-to-end solution. And this is where standards-based I/O and open engagement strategy has made integration easy and straightforward for Cerebras. A few weeks ago, we announced a partnership with AMD to build disaggregated inference solutions.

Prefill 與 decode 的處理器需要連結起來，才能形成端到端的解決方案。而這正是以標準為基礎的 I/O 與開放合作策略，讓 Cerebras 的整合變得容易且直接之處。幾週前，我們宣布與 AMD 合作打造解耦式推理解決方案。

The solutions combine their Helios racks with RCS systems. The combined solution maintains cerebras speed while increasing throughput by 5x. To understand how powerful this is, it's important to understand the difference between speed and throughput. Speed is a measure per user. It's measured in tokens per second per user. It is how fast your query is answered or how long it takes an agent to finish a task. Here it is on the x-axis. Throughput, on the other hand, is the total number of tokens the solution can produce per second. It is measured by adding up all the tokens across all the simultaneous users.

該方案將其 Helios 機櫃（racks）與 RCS 系統結合。整合後的方案在維持 Cerebras 速度的同時，將吞吐量提升 5 倍。要理解其威力，必須先理解速度（speed）與吞吐量（throughput）的差異。速度是以每位使用者為單位的衡量。以每位使用者每秒 token 數（tokens per second per user）來衡量。它代表你的查詢被回答得有多快，或一個代理完成任務需要多久。在這裡它位於 x 軸。另一方面，吞吐量是該解決方案每秒可產生的 token 總數。其衡量方式是把所有同時使用者的 token 數加總。

Here it is shown as it's generally done on the y-axis. Speed is critical for user experience. Throughput. Is critical for inference economics. GPU solutions can support high throughput, but only at low speeds.

在這裡它如一般慣例顯示在 y 軸。速度對使用者體驗至關重要。吞吐量。對推理的經濟性至關重要。GPU 方案可以支援高吞吐量，但只能在低速度下達成。

When configured to support even moderate speeds, GPU throughput drops precipitously. This is true not just for GPUs, but also for ASICs and all solutions that use HBM. The HBM memory architecture forces a trade-off between throughput and speed. SRAM-based architectures, like Cerebras, are the exact opposite. We support blisteringly fast tokens, but at moderate throughput. So GPUs want to get faster without giving up throughput. Cerebras wants more throughput without giving up speed.

當配置為支援即使是中等速度時，GPU 的吞吐量會急遽下降。這不僅適用於 GPU，也適用於 ASIC 以及所有使用 HBM 的方案。HBM 記憶體架構迫使吞吐量與速度之間必須取捨。以 SRAM 為基礎的架構（如 Cerebras）則恰恰相反。我們支援極高速的 token 生成，但吞吐量屬中等。因此，GPU 希望在不犧牲吞吐量的情況下變得更快。Cerebras 則希望在不犧牲速度的情況下提升吞吐量。

Herein is the strength of our disaggregated solution. It delivers Cerebras speed with 5x higher throughput. Increasing throughput by 5x while keeping our industry-leading speed has a profound impact on the economics of token generation. It means up to five times as many high-speed, high-value tokens are made by each Cerebrot system.

這正是我們解耦式方案的強項。它在提供 Cerebras 速度的同時，帶來高出 5 倍的吞吐量。在維持我們業界領先速度的同時將吞吐量提升 5 倍，會對 token 生成的經濟性產生深遠影響。這意味著每一套 Cerebrot 系統可產出最多 5 倍的高速、高價值 token。

More tokens per system at lower cost means more revenue and more gross margin. More tokens generated per CS system also means more tokens per watt, making each data center more profitable. Perhaps most important in a data center-constrained environment. The disaggregated solution allows us to serve more of the demand that we have in RPO.

每套系統以更低成本產出更多 token，代表更高營收與更高毛利。每套 CS 系統生成更多 token，也意味著每瓦 token 數更高，使每座資料中心更具獲利能力。尤其在資料中心受限的環境中，這點或許最為重要。解耦式方案讓我們能服務更多已納入 RPO 的需求。

Finally, we believe this disaggregation approach makes performance and economic sense with any GPU.

最後，我們相信這種解耦方法搭配任何 GPU 都在效能與經濟性上合理。

For operators who have already deployed large footprints of GPUs, disaggregation with Cerebras offers them an opportunity to create meaningful leverage built on their existing investments. By pairing some portion of those GPUs with Cerebras solutions, dramatically improving the value and usefulness of their data center footprint.

對於已部署大量 GPU 的營運商而言，與 Cerebras 進行解耦可讓他們在既有投資之上建立有意義的槓桿。透過將其中一部分 GPU 與 Cerebras 解決方案配對，可大幅提升其資料中心佈署的價值與實用性。

Continuing on the capabilities theme, let's turn to our roadmap.

延續能力的主題，讓我們來談談我們的路線圖。

Our engineering execution is continuing at pace.

我們的工程執行持續以既定節奏推進。

We expect to deliver new systems that double our speed each year for the next several years.

我們預期在未來數年，每年推出速度翻倍的新系統。

Remember, we're doubling our performance starting with a 15x performance advantage over everyone else in the industry.

請記得，我們是在相較產業其他所有人已有 15 倍效能優勢的基礎上，再持續將效能翻倍。

In addition, while keeping the performance crown, over the next 18 months, we plan to deliver solutions that increase throughput by more than 20x.

此外，在維持效能王冠的同時，未來 18 個月內我們計畫推出可將吞吐量提升超過 20 倍的解決方案。

Next week at our annual Supernova Conference, we will be unveiling the CS4, our fourth generation system.

下週在我們一年一度的 Supernova 大會上，我們將揭曉第四代系統 CS4。

It will be a great event with lots of product announcements, so I recommend you attend.

這將是一場有大量產品發布的精彩活動，我建議你參加。

Finally, we are currently on track to launch our CS5 in the second half of 2027. Looking even further out, our invention engine is humming. We have significant partnerships with the US government for delivery of stacked memory solutions as well as integrated wafer-scale optical solutions.

最後，我們目前按計畫在 2027 年下半年推出 CS5。再往更長遠看，我們的發明引擎正全速運轉。我們與美國政府有重要合作，涵蓋堆疊式記憶體解決方案，以及整合式晶圓級光學解決方案的交付。

In the years ahead, you can expect to see inventions from us in chip and chip architecture, as well as all elements of system design, including packaging, I/O, and power delivery. To summarize the capability section, we expect to continue to deliver pioneering advances in product and technology to drive up speed and throughput, reduce the power use per token, and slash the cost per token of our solution.

在未來幾年，你可以期待我們在晶片與晶片架構方面，以及系統設計的各個要素（包括封裝、I/O 與供電）推出發明成果。總結能力部分，我們預期將持續在產品與技術上交付開創性進展，以提升速度與吞吐量、降低每個 token 的耗電，並大幅削減我們解決方案的每 token 成本。

Now let's turn to the customer front. Fast tokens are in demand and command a premium at market.

現在讓我們轉向客戶面。高速 token 需求強勁，並在市場上享有溢價。

And fast tokens with Frontier Intelligence are only available through OpenAI Cerebra's partnership.

而具備 Frontier Intelligence 的高速 token，僅能透過 OpenAI 與 Cerebras 的合作夥伴關係取得。

Our work with AWS continues and we expect to have solutions generally available in Q1 2027 through AWS's Bedrock platform. This AWS partnership expands our market opportunity and provides us with global reach through an industry leader who is trusted by nearly every enterprise in the world.

我們與 AWS 的合作持續推進，並預期於 2027 年第一季透過 AWS 的 Bedrock 平台提供全面可用（generally available）的解決方案。這項 AWS 合作擴大了我們的市場機會，並透過一位幾乎被全球每一家企業信賴的產業領導者，為我們帶來全球觸及能力。

Our discussions with other hyperscalers are also going well. We expect to produce first revenues starting in mid-2027 and ramp through 2028 and beyond.

我們與其他超大規模雲端業者（hyperscalers）的洽談也進展順利。我們預期自 2027 年年中開始產生首筆營收，並在 2028 年及之後持續放量成長。

And with all of this progress, I think it is important to keep in mind that our $25 billion in RPO does not reflect any backlog of business from AWS or any other hyperscaler at this time. Our business outside of OpenAI and the hyperscalers continues to grow nicely.

在所有這些進展之下，我認為重要的是要記住：我們 250 億美元的 RPO 目前並未反映來自 AWS 或任何其他超大規模雲端業者的任何業務積壓（backlog）。我們在 OpenAI 與超大規模雲端業者之外的業務，也持續穩健成長。

For example, in Q2, we signed six deals north of $30 million.

例如，在第二季，我們簽下了六筆超過 3,000 萬美元的交易。

AI coding continues its rapid rate of growth. In our experience, no one says, I'm happy with slow tokens when coding. So not surprisingly, in the coding category, our footprint continues to grow. We signed new agreements with public companies such as Figma and startup leaders such as Cognition. And we extended our presence in Europe, the major win at Lovable.

AI程式碼撰寫持續以快速的速度成長。依我們的經驗，沒有人在寫程式時會說：我對慢速 token 很滿意。因此不意外地，在程式碼撰寫這個類別中，我們的版圖持續擴大。我們與 Figma 等上市公司以及 Cognition 等新創領導者簽署了新的協議。我們也將觸角延伸至歐洲，在 Lovable 取得重大勝利。

Agentic flows are growing quickly and the value of speed compounds as agentic operations rapidly evolve toward multi-step, multi-agent solutions.

代理式流程（agentic flows）正在快速成長，而隨著代理式作業迅速演進至多步驟、多代理的解決方案，速度的價值會持續複利放大。

Companies as diverse as Block, AlphaSense, and GSK signed new agreements during the second quarter with Cerebras to leverage FastInference to provide their custom-made agents.

在第二季，Block、AlphaSense 與 GSK 等多元類型的公司與 Cerebras 簽署了新的協議，運用 FastInference 來提供其量身打造的代理（agents）。

Fast AI also opens up new markets, extending the CAM for Cerebras. Security is one such example.

快速 AI 也開啟了新市場，擴大 Cerebras 的 CAM。資安就是其中一個例子。

Our recent win with CrowdStrike is an application that only exists if AI is fast. Fast AI enables AI-based security devices to sit in line with enterprise traffic and use LLMs to secure traffic so quickly that nobody notices. Fast AI enables an LLM to provide security that is invisible to users.

我們近期與 CrowdStrike 的勝利，是一項只有在 AI 足夠快時才存在的應用。快速 AI 讓以 AI 為基礎的資安設備能夠串接在企業流量的資料路徑中，並使用 LLM 以快到讓人無感的速度保護流量。快速 AI 使 LLM 能提供對使用者而言「看不見」的安全防護。

The AI provides the security. The speed creates the invisibility that enables the security to avoid delay and disruption. We expect this type of security to become the norm given the rapidly evolving threat landscape. Enterprises will soon expect vast swaths of their traffic to be inspected in this way, creating massive new opportunities made possible exclusively through fast AI.

AI 提供安全防護。速度創造了「無形性」，使安全防護得以避免延遲與中斷。鑑於威脅態勢快速演變，我們預期這類安全防護將成為常態。企業很快就會期待其大量流量以這種方式被檢查，進而創造龐大的新機會，而這些機會唯有透過快速 AI 才能實現。

Frontier Labs, hyperscalers, leading chip makers, the fastest growing startups, and massive enterprises are all now customers and partners of Cerebras and benefit from our blazing fast inference.

Frontier Labs、超大規模雲端業者（hyperscalers）、領先的晶片製造商、成長最快的新創，以及大型企業，如今都是 Cerebras 的客戶與合作夥伴，並受益於我們極致快速的推論能力。

To summarize, overall a strong quarter. We went public in a successful IPO. We beat on all metrics, core revenue, core margins, and core operating margins. We made progress in each of our key domains, capacity, capability, and customers.

總結來說，整體而言是強勁的一季。我們成功完成 IPO 並上市。我們在所有指標上都優於預期：核心營收、核心毛利率與核心營業利潤率。我們在各個關鍵領域都取得進展：產能、能力與客戶。

These are the foundations on which we will achieve our goals of massive growth in 2027 and 2028 and continue this exceptional rate of growth in '29 and beyond.

這些是我們在 2027 與 2028 年實現大幅成長目標的基礎，並在 2029 年及之後延續這種卓越的成長速度。

And with that, I'll turn things over to Bob. Bob?

接下來，我把時間交給 Bob。Bob？

Thank you, Andrew, and good afternoon, everyone.

謝謝你，Andrew，各位下午好。

We made tremendous progress in the first half of 2026. As we described, 2026 is the foundation for multiples of growth over the years ahead. We entered the year having won one of the largest technology deals ever, creating RPO of more than $25 billion.

我們在 2026 年上半年取得了巨大的進展。如同我們所述，2026 年是未來多年倍數成長的基礎。我們在年初就贏得了史上規模最大的科技交易之一，創造超過 250 億美元的 RPO。

This required us to immediately work on major increases in three critical components of capacity.

這要求我們立即著手在產能的三個關鍵組成部分上大幅提升。

First, we needed to increase our wafer supply. As Andrew described. Due to our strong relationship and support from TSMC, we did that and are now well positioned, not just for the remainder of this year, but for the next year as well.

第一，我們需要增加晶圓供應。如 Andrew 所描述。由於我們與台積電（TSMC）的強健關係及其支持，我們已做到這點，並且現在不僅對今年剩餘期間、也對明年都處於良好位置。

Second, we needed to scale our manufacturing capacity. We are already four times above where we were in the first half of 2025, and we will have increased the manufacturing capacity more than 10x in 2026. So we're making great progress here.

第二，我們需要擴大製造產能。我們目前已是 2025 年上半年水準的四倍，而在 2026 年我們將把製造產能提升超過 10 倍。因此我們在這方面進展非常順利。

And third, we need to substantially increase our data center capacity. We've made significant progress with over 600 megawatts now up or under contract expected for delivery by the end of 2027, plus a pipeline in gigawatts. So the foundation is in place to support a tripling or better in our core revenue in 2027.

第三，我們需要大幅增加資料中心產能。我們已取得顯著進展，目前已有超過 600 兆瓦的容量已上線或已簽約，預計於 2027 年底前交付，另有以 GW 計的管線。因此，支撐 2027 年核心營收成長至三倍或更高的基礎已經到位。

And additional multiples in future years. This growth is also setting us up for significant margin expansion in 2027 and beyond.

以及未來年度的更多倍數成長。這樣的成長也讓我們在 2027 年及之後具備顯著的利潤率擴張條件。

Turning to Q2 financial results. We had another quarter of strong results beating expectations across each element of our guidance. We delivered record core revenue. We beat on core gross margin and on core operating margin.

接著談第二季財務結果。我們又交出一季強勁成績，在指引的各個項目上都優於預期。我們創下核心營收新高。我們在核心毛利率與核心營業利潤率上也都優於預期。

I will be using the same core business framework introduced last quarter to describe our progress. The definition of our core business metrics and reconciliations of all of them to GAAP are included in today's earnings release and on our website.

我將沿用上季提出的相同核心業務架構來說明我們的進展。核心業務指標的定義，以及所有指標與 GAAP 的調節表，均收錄於今日的財報新聞稿與我們的網站。

Core revenue was $209.9 million, up 103% year over year. Our private cloud business is growing at an extraordinary pace. Core cloud and other services revenue was $127.7 million, up 287% year-over-year.

核心營收為 2.099 億美元，年增 103%。我們的私有雲業務正以驚人的速度成長。核心雲端與其他服務營收為 1.277 億美元，年增 287%。

This nearly four-fold increase reflects the tremendous demand we have for Cerebras' fast inference service. Core hardware revenue was $82.1 million in the quarter, up 17% compared to last year.

這接近四倍的成長反映出市場對 Cerebras 快速推論服務的強勁需求。本季核心硬體營收為 8,210 萬美元，較去年成長 17%。

We focus on total core revenue, not the mix between the two, which can vary significantly quarter to quarter due to the timing of large new cloud capacity additions and hardware shipments.

我們關注的是核心營收總額，而非兩者之間的組合占比；由於大型新增雲端產能與硬體出貨的時點不同，該占比可能在各季之間大幅波動。

In Q2. Most of the total core revenue was attributable to increases in our core cloud offering, reflecting the ramp in our OpenAI deployment, increases in our other cloud customers' usage, and finally, hardware customers who are also wrestling with the timing of new data center capacity.

在第二季。核心營收總額的大部分來自核心雲端產品的增加，反映出我們 OpenAI 部署的爬坡、其他雲端客戶使用量的提升，以及最後，硬體客戶也正面臨新資料中心產能時程的挑戰。

The demand for fast inference continues to be strong with several late-stage hardware deals representing hundreds of millions of dollars in the pipeline from new customers as well as significant new cloud deals for 2027.

對快速推論的需求仍然強勁；我們的管線中有數筆後期硬體交易，來自新客戶、金額合計達數億美元，此外也有面向 2027 年的重要新雲端交易。

Existing fast inference markets are growing and new ones are getting started.

既有的快速推論市場正在成長，新的市場也正在啟動。

We see disaggregation as an important unlock to drive new use cases since it dramatically improves the economics of inference and of data center ownership.

我們認為解耦（disaggregation）是推動新使用情境的重要解鎖因素，因為它能大幅改善推論與資料中心持有成本的經濟性。

Today, this means that up to 5x more tokens are produced per CS system, so power and cost are significantly reduced per token.

在今日，這代表每套 CS 系統可產出最多 5 倍的 token，因此每個 token 的耗電與成本都顯著降低。

By continuing to invest heavily in R&D and our product roadmap. Cerebras will quadruple our current industry-leading speed and increase throughput by more than 20x through the end of 2027, drastically improving our performance and the economics of inference.

透過持續在研發與產品路線圖上大力投資。Cerebras 將在 2027 年底前把目前業界領先的速度再提升四倍，並將吞吐量提高超過 20 倍，從而大幅改善我們的效能與推論的經濟性。

Turning now to gross margin. Year-over-year, core gross margins improved substantially. Q2 core gross margin was 40.6%. Approximately 940 basis points higher than Q2 twenty-five.

接著談毛利率。年對年而言，核心毛利率大幅改善。第二季核心毛利率為 40.6%。較 2025 年第二季高約 940 個基點。

The increase in value of fast inference by the market, our continuous stream of product improvements and additional economies of scale. Breaking the total core gross margin into its components, core cloud and other services gross margin was 41.8%, 1,600 basis points better than Q2 '25Core hardware gross margin was 38.8%, 510 basis points higher than a year ago.

市場對快速推論價值的提升、我們持續不斷的產品改良，以及額外的規模經濟。將核心毛利率拆分為各組成部分：核心雲端與其他服務毛利率為 41.8%，較 2025 年第二季改善 1,600 個基點；核心硬體毛利率為 38.8%，較一年前提高 510 個基點。

As we described last quarter, we are meeting some of the overwhelming demand for our fast inference service by temporarily renting some of our own systems back from our cloud customers and making it available through the Cerebras cloud.

如同我們上季所述，為了滿足對我們快速推論服務的部分龐大需求，我們暫時向部分雲端客戶回租我們自有的部分系統，並透過 Cerebras 雲端提供使用。

Serving this inference demand sooner strengthens our ability to meet the needs of our cloud customers. And to grow with them overtime. We believe this will create additional long-term value for Cerebras and its shareholders.

更早滿足這些推論需求，能強化我們滿足雲端客戶需求的能力。並隨著時間與他們共同成長。我們相信這將為 Cerebras 及其股東創造額外的長期價值。

In the short-term, it reduces gross margin as we have a higher cost for this rented capacity. As a result, sequentially, core gross margin was 40.6% versus 46.5% in Q1 '26. Had we not had higher costs due to increasing our private cloud capacity by renting back more of our systems, core gross margins would have been approximately 500 basis points higher and more similar to last quarter.

短期而言，由於這些租用產能的成本較高，會拉低毛利率。因此，按季比較，核心毛利率為 40.6%，而 2026 年第一季為 46.5%。若非因我們透過回租更多自有系統來擴增私有雲容量而導致成本上升，核心毛利率約可高出 500 個基點，並更接近上一季水準。

Looking forward. We expect Q3 to be the low point for core gross margin before improving significantly in Q4 '26 as we bring on more data centers filled with lower-cost CerebroZone systems.

展望未來。我們預期 2026 年第三季將是核心毛利率的低點，之後在 2026 年第四季，隨著我們啟用更多資料中心並部署成本更低的 CerebroZone 系統，核心毛利率將顯著改善。

This will cause core cloud gross margin to step back up.

這將使核心雲端毛利率回升。

Core gross margin will also continue to improve in 2027 and trend towards our target of 60% plus for several reasons. The market has recognized that fast tokens are more valuable tokens. This supports higher pricing, which is reflected in hardware and cloud deals that will be recognized over the next several quarters.

核心毛利率也將在 2027 年持續改善，並因多項原因朝我們 60% 以上的目標邁進。市場已認知到快速 token 的價值更高。這支撐了更高的定價，並反映在未來數季將認列的硬體與雲端合約中。

Over the next few quarters, we will roll off higher-cost rented systems and replace them with lower-cost owned systems in our private cloud. Our product roadmap has us increasing throughput by 20x over the next 18 months. This reduces the cost to produce tokens per system and per unit of power.

在接下來幾季，我們將逐步汰換成本較高的租用系統，並以私有雲中成本較低的自有系統取代。依照我們的產品路線圖，未來 18 個月我們將把吞吐量提升 20 倍。這將降低每套系統以及每單位電力所產生 token 的成本。

As our scale grows, our bill of material costs in our supply chain will improve more. By being on the five-nanometer node, our wafer costs are lower than others who need to be on the three or two nanometer node. We're also purchasing wafers in much higher volumes.

隨著規模擴大，我們供應鏈中的物料清單（BOM）成本將進一步改善。由於採用 5 奈米製程節點，我們的晶圓成本低於那些必須使用 3 奈米或 2 奈米節點的業者。我們也以更高的採購量購買晶圓。

Finally, we do not rely on HBM, which pressures those who use it to either raise prices or lose margin points. We are not exposed to that risk, which we believe will improve our value proposition and pricing flexibility.

最後，我們不依賴 HBM，使用 HBM 的業者會因此承受壓力：不是提高價格，就是犧牲毛利率。我們不暴露於該風險之下，這也將提升我們的價值主張與定價彈性。

Turning to operating margin.

接著談營業利潤率。

Core operating loss was $33.6 million. Core operating margin was negative 16% compared to negative 42% a year ago. An improvement of approximately 2,600 basis points year over year. Our ability to deliver this significant improvement in core operating margin while more than doubling revenues and stepping up our investments in all areas demonstrates the strong operating leverage inherent in our business model.

核心營業虧損為 3,360 萬美元。核心營業利潤率為負 16%，相較於一年前的負 42%。年增改善約 2,600 個基點。在營收增加逾一倍、且我們在各領域加大投資的同時，仍能實現核心營業利潤率的顯著改善，顯示我們商業模式內含強勁的營運槓桿。

Today, we are investing in world-class people, manufacturing and data center capacity and company infrastructure to support the significant increase in scale we expect to deliver over the next several years.

目前，我們正投資於世界級人才、製造與資料中心產能，以及公司基礎建設，以支援我們預期未來數年將實現的顯著規模擴張。

Remaining performance obligations at June 30, 2026, are $25.4 billion. This backlog provides us visibility to have high confidence in future revenue growth and to invest as needed ahead of it.

截至 2026 年 6 月 30 日的剩餘履約義務（RPO）為 254 億美元。這些在手訂單為我們提供能見度，使我們對未來營收成長具高度信心，並可在需要時提前投資以支應成長。

Our existing large strategic customers provide validation, contractual visibility, and the economic support required to build new capacity at scale. At the same time, we're having success expanding our addressable market and customer base.

我們既有的大型策略客戶提供了驗證、合約能見度，以及以規模建置新產能所需的經濟支持。同時，我們也成功擴大可服務市場與客戶基礎。

OpenAI provides, among other things, scale and frontier insight. AWS provides global enterprise reach. Recent collaboration with AMD expands the market opportunity to include disaggregated inference and fast inference is cracking open more new markets like security.

OpenAI 提供的其中幾項價值包括規模與前沿洞察。AWS 提供全球企業客戶觸及能力。近期與 AMD 的合作擴大了市場機會，涵蓋解耦式推論（disaggregated inference）；而快速推論也正在開啟更多新市場，例如資安。

We ended Q2 with more than $8.6 billion in cash equivalents, restricted cash and marketable securities. We also have a revolving credit facility of up to $850 million that has been unused to date. Our liquidity and balance sheet position is strong and was enhanced by our IPO in Q2. It is a significant advantage that provides us with flexibility to invest and adjust opportunities in these very dynamic and high-growth market conditions.

我們在第二季末持有超過 86 億美元的約當現金、受限制現金與有價證券。我們另有最高 8.5 億美元的循環信用額度，迄今尚未動用。我們的流動性與資產負債表狀況強健，並因第二季 IPO 而進一步提升。這是一項重大優勢，使我們能在高度動態且高成長的市場環境中，彈性投資並調整機會。

In addition, we have the advantage of much lower net capital expenditures per megawatt than the vast majority of AI cloud providers for two key reasons. We primarily incur Capex for the deployment of our own hardware in our data centers at much lower BOM costs that does not include the high profit margins many others must pay.

此外，基於兩個關鍵原因，相較於絕大多數 AI 雲端供應商，我們每兆瓦的淨資本支出（Capex）明顯更低。我們的資本支出主要用於在自有資料中心部署自家硬體；由於 BOM 成本更低，且不包含許多其他業者必須支付的高額利潤加成，因此成本更具優勢。

Second, we are reimbursed for a meaningful portion of the remaining Capex for data center fit-out as data center pass-through cost reimbursement from our largest customer.

第二，我們最大的客戶會以資料中心轉嫁成本（pass-through）方式，補償資料中心裝修（fit-out）剩餘資本支出中的相當一部分。

Now turning to our outlook. For Q3 2026, we expect. Core revenue to be in the range of $214million to $216 million, core gross margin in the range of 38% to 40%, and core operating margin in the range of minus 25% to minus 23%. For the full year 2026, we're raising core revenue to the range of $880 million to $890 million.

接著談我們的展望。對於 2026 年第三季，我們預期：核心營收介於 2.14 億至 2.16 億美元，核心毛利率介於 38% 至 40%，核心營業利潤率介於負 25% 至負 23%。對於 2026 全年，我們將核心營收上調至 8.8 億至 8.9 億美元。

We're raising core gross margin to the range of 41% to 43% and we're raising core operating margin to the range of negative 19% to negative 17%. In closing, Q2 was a very strong quarter of continued execution and growth for Cerebras.

我們將核心毛利率上調至 41% 至 43%，並將核心營業利潤率上調至負 19% 至負 17%。最後總結，第二季對 Cerebras 而言是持續執行與成長非常強勁的一季。

We delivered record core revenue, cloud and services revenue nearly quadrupled and gross margin and operating margin were also significantly better than our guidance. We improved our guidance for each of these items for the full year. We've made great progress building our capabilities, capacity and customers and ended the quarter with more than $8.6 billion in cash and cash equivalents and investments to continue to execute our growth plans.

我們創下核心營收新高，雲端與服務營收接近成長四倍，毛利率與營業利潤率也明顯優於我們的指引。我們也針對全年上述各項指標上調指引。我們在能力、產能與客戶拓展方面取得重大進展，並在季末持有超過 86 億美元的現金、約當現金與投資，以持續推動我們的成長計畫。

We're well positioned to grow revenue by more than 3 times in 2027 and for tremendous additional growth in the following years.

我們已具備良好條件在 2027 年將營收成長超過 3 倍，並在其後年度實現更可觀的額外成長。

While also significantly expanding gross and operating margins towards our targets.

同時也將毛利率與營業利潤率大幅提升，朝我們的目標邁進。

I'll now turn this over to Andrew for final thoughts.

接下來我把時間交給 Andrew 做最後補充。

Thank you, Bob.

謝謝你，Bob。

More than 10 years ago, we started Cerebras with the belief that we could build a better processor for AI and the belief that to deliver the processor, we would need to build a full accelerator system and racks.

十多年前，我們創立 Cerebras，基於一個信念：我們能為 AI 打造更好的處理器；並且我們也相信，為了交付這顆處理器，我們需要打造完整的加速器系統與機櫃（racks）。

Today, Cerebras is one of only four companies, Google, Amazon, NVIDIA, and Cerebras, to build processors, systems, data centers, and deliver AI-based cloud services to customers.

如今，Cerebras 是僅有的四家公司之一——Google、Amazon、NVIDIA 與 Cerebras——能夠打造處理器、系統、資料中心，並向客戶提供以 AI 為基礎的雲端服務。

Thank you for listening to our prepared remarks.

感謝各位聆聽我們事先準備的發言。

And with that, I ask the operator to please open the line for questions.

接下來，我請接線員開放提問。

(Operator Instructions)

（接線員指示）

Our first question comes from Timothy Arcuri with UBS. Your line is open.

我們的第一個問題來自 UBS 的 Timothy Arcuri。請發言。

Thanks a lot. Andrew, I wanted to ask about customer concentration. So you did say that revenue would be up more than 3x next year and obviously we know that OpenAI is ramping right now, so that's a big piece of your incremental revenue today. I would think that AWS could be $1 billion next year, something like that, maybe more. So how do you think about customer concentration when you look at next year, like is it going to be.

非常感謝。Andrew，我想問一下客戶集中度。你確實提到明年營收會成長超過 3 倍，而我們也知道 OpenAI 目前正在快速放量，所以這是你們當前增量營收中的一大部分。我認為 AWS 明年可能會達到 10 億美元左右，或許更多。那你如何看待明年的客戶集中度？例如，會不會是……

Two-thirds of your revenue is like those two customers? And then can you also speak to your talks with some of the other folks, Google and Microsoft and folks like that?

你們營收的三分之二都來自這兩個客戶？另外，你也能談談你們與其他業者（例如 Google、Microsoft 等）洽談的情況嗎？

Thanks a lot.

非常感謝。

Sure. I think it's a good question and I think some historical perspective might be worthwhile, right? In 2021.

當然。我認為這是個好問題，也許回顧一些歷史背景會有幫助，對吧？在 2021 年。

People complained that we only had government customers. And then when we won a sovereign cloud at $1 billion, there were concerns we only had a sovereign cloud. And then we won the largest lab, Frontier Lab. And then there were concerns that we didn't have a hyperscaler, and then we won AWS.

人們抱怨我們只有政府客戶。接著當我們拿下價值 10 億美元的主權雲案子時，又有人擔心我們只有主權雲。然後我們贏得了最大的實驗室——Frontier Lab。接著又有人擔心我們沒有超大規模雲端業者（hyperscaler），然後我們就拿下了 AWS。

And so I think in each of those cases, we were able to use the momentum that the previous step gave us to expand our business. I think OpenAI is an enormous customer, and they're an enormous part of not just our business, but of everybody's business in the sector.

所以我認為在上述每一種情況下，我們都能利用前一步帶來的動能來擴展業務。我認為 OpenAI 是一個非常大的客戶，而且他們不只是我們業務的重要部分，也是這個產業裡每個人業務的重要部分。

And I think they'll say a big part next year. But you're absolutely right that AWS and others. Whether they're rapidly growing coding companies or some of the use cases around security, they will be a larger portion and OpenAI will shrink as a percentage of our revenue over time.

而且我想他們明年會占很大一部分。但你說得完全正確，AWS 和其他客戶——不論是快速成長的程式開發公司，或是一些與資安相關的使用案例——它們將占更大的比重，而 OpenAI 隨著時間推移，作為我們營收占比會逐步下降。

But I think you can expect for next year them to still be a meaningful portion of our revenue.

但我認為你可以預期明年他們仍會是我們營收中相當有意義的一部分。

Great. And then just as a quick follow-up. So I know, Bob, you said that capacity, I think you said it's going up 10x this year over year.

很好。接著快速追問一下。所以我知道 Bob，你說產能——我記得你說今年年增大概是 10 倍。

Is there any sense of how much it's going to grow next year? I know that Andrew said that revenue is going to grow 3x, but is there any sense in terms of how much your manufacturing capacity will actually grow next year over year?

那明年大概會成長多少，有沒有一個概念？我知道 Andrew 說營收會成長 3 倍，但就製造產能而言，明年相較今年實際會成長多少，有沒有任何概念？

Yeah, we're going to end the year well over 10x our manufacturing capacity.

是的，我們今年年底的製造產能會遠遠超過 10 倍。

And we already have contracted facilities three or four times more for growth in 2027, and we still have some time to contract for more. So we're looking at enormous growth over a several year period.

而且我們已經簽約鎖定了 2027 年成長所需的產能，規模是目前的三到四倍，並且我們仍有一些時間可以再簽更多產能。所以我們預期在未來數年會有非常巨大的成長。

Thank you. Our next question comes from Joshua Buckhalter with TD Cowen. Your line is open.

謝謝。下一個問題來自 TD Cowen 的 Joshua Buckhalter。請發問。

Hey, guys, thank you for taking my question.

嗨，各位，謝謝讓我提問。

Maybe following up on Kim's previous one, can you maybe just walk us through how the economics of the Amazon deal are going to work? Is the plan that it'll be available next year and offered in AWS Cloud? And then we'll basically see how much demand is and so it's difficult to forecast right now.

也許延續 Kim 前一個問題，你們能否帶我們梳理一下與 Amazon 這筆交易的經濟模式會如何運作？計畫是明年就能上線並在 AWS Cloud 提供嗎？然後我們基本上就看需求有多大，因此目前很難預測？

Thank you.

謝謝。

Sure, it will be available. It is deployed in Amazon data centers.

當然，會提供。它會部署在 Amazon 的資料中心。

It will be delivered through Amazon's API service Bedrock.

它將透過 Amazon 的 API 服務 Bedrock 交付。

And we are in the process right now of organizing deployments. So I think that's sort of the way to think about it. We expect the service to be live in Q1.

而我們目前正在安排部署事宜。所以我認為可以用這種方式來理解。我們預期該服務會在第一季上線。

Got it.

了解。

Thank you for that, Andrew. And then maybe with the AMD engagement, any more call you can give on the go to market as you connect with the Helios rack and timeline you would expect to revenue? And then regarding the AMD engagement, they made an acquisition of an inferencing hardware company recently.

謝謝你，Andrew。接著關於與 AMD 的合作，你能否再多談一些你們與 Helios 機櫃（rack）對接後的上市策略（go-to-market）以及你們預期開始貢獻營收的時間表？另外，關於與 AMD 的合作，他們最近收購了一家推論（inferencing）硬體公司。

Could you maybe speak to how that fits in with what you guys are offering as we think about their broader suite?

你能否談談這如何與你們所提供的方案相契合，並讓我們從他們更廣泛的產品組合角度來理解？

Thank you.

謝謝。

Sure. I think a couple of things. I think that.

當然。我想有幾點。我想——

We'll be announcing additional parts of our arrangement with AMD over time.

我們會隨時間推進，陸續宣布與 AMD 合作安排的更多細節。

But I think the joint solution of Helios racks in front of Cerebra systems, the Helios racks doing prefill and Cerebra is doing decode, is an extremely strong offering, right? An offering in which we deliver. Vastly faster speed than Helios can deliver and vastly more throughput than Cerebras can deliver alone and that solution is enormously compelling and we have buyers for it already.

但我認為，Helios 機櫃與 Cerebra 系統前後搭配的聯合解決方案——由 Helios 機櫃負責 prefill、Cerebra 負責 decode——是一個非常強的產品組合，對吧？一個我們能交付的方案。其速度遠遠快於 Helios 單獨能提供的速度，吞吐量也遠遠高於 Cerebras 單獨能提供的吞吐量；而且這個方案非常有吸引力，我們已經有買家了。

The second question is of recent acquisition by AMD.

第二個問題是關於 AMD 最近的收購。

Look, I think the company they acquired was interesting and innovative and no one's more excited about. Innovative hardware than we are. I think buying hardware startups, there's a lot of time between when you buy them and when they deliver. I think we were impressed by what those guys were working on and think there are many applications in AMD's portfolio for them.

你看，我認為他們收購的那家公司很有意思、也很有創新性，沒有人會比我們更對創新硬體感到興奮。對創新硬體的熱情不會有人比我們更高。我認為收購硬體新創後，從收購到真正交付之間通常還有很長的時間。我們對那些人正在做的事情印象深刻，也認為在 AMD 的產品組合中有很多應用場景。

I don't see the first application there being data center inference.

我不認為它的第一個應用會是資料中心推論。

Thank you. Our next question comes from Tom O'Malley with Barclays. Your line is open.

謝謝。下一個問題來自 Barclays 的 Tom O'Malley。請發問。

Hey guys, this is Kyle Bluestein on for Tom O'Malley.

嗨，各位，我是代 Tom O'Malley 提問的 Kyle Bluestein。

Thank you for taking our question. I wanted to go back to Josh's question on the economics with the AMD deal. Is the way this kind of works, you buy an AMD Helios rack, install it in your cloud, and then all the revenue that comes from customers renting out the disaggregated inference solution goes to you, or is there some sort of revenue sharing agreement that happens here?

謝謝讓我們提問。我想回到 Josh 關於 AMD 交易經濟模式的問題。這個運作方式是否是：你購買一套 AMD Helios 機櫃，安裝在你們的雲端裡，然後客戶租用這個解耦式推論（disaggregated inference）解決方案所產生的所有營收都歸你們所有？還是這裡會有某種營收分成協議？

Yes to the first part of the question.

第一部分的答案是肯定的。

Okay, thank you. And then for my follow-up, the AWS deal is getting installed in their clouds first. Do you see an eventual path to you hosting Trainium and the CS3 together in your cloud or in other hyperscale clouds, just trying to think about how disaggregated inference can evolve in terms of future deployments.

好的，謝謝。那我的追問是，AWS 這筆合作會先安裝在他們的雲端。你們是否看到一條可能的路徑：未來你們在自己的雲端或其他超大規模雲端中，同時託管 Trainium 與 CS3？我只是想理解解耦式推論在未來部署上可能如何演進。

I think we are very interested in that approach. I think that.

我認為我們對那種做法非常有興趣。我認為——

As you've seen with Google, there is an opportunity for hyperscalers with their own parts to seek to deploy those parts outside of the boundaries of their own data centers. That's something we'd be interested in, not just with AWS, but with others.

正如你在 Google 的案例中所看到的，超大規模雲端業者若擁有自家零組件，確實有機會尋求把這些零組件部署到自家資料中心邊界之外。這是我們會有興趣的方向，不僅是 AWS，也包括其他業者。

And so I think that's very much on the table for.

所以我認為這件事非常可能——

For the future with AWS.

在未來與 AWS 的合作中納入考量。

Thank you. Our next question comes from Quinn Bolton with Needham & Co. Your line is open.

謝謝。下一個問題來自 Needham & Co. 的 Quinn Bolton。請發問。

Just wanted to just a quick clarification on the AMD deal. Andrew, if you purchase and stand up the Helios racks in your Cerebus cloud, but that service is.

我想就 AMD 這筆合作做個快速釐清。Andrew，如果你們在 Cerebus 雲端中採購並架設 Helios 機櫃，但該服務是——

Delivered to OpenAI under your contract, does that represent an additional revenue opportunity or how should we think about the potential for revenue in that instance? And then I've got a follow-up.

依照你們與 OpenAI 的合約交付給 OpenAI，這是否代表額外的營收機會？或在那種情況下，我們應該如何看待潛在營收？然後我還有一個追問。

I think that whenever you increase throughput while keeping your performance the same, you increase your opportunity for revenue, right? Throughput is.

我認為只要你在維持同樣效能的同時提升吞吐量，你就提升了營收機會，對吧？吞吐量是——

The number of customers that you can simultaneously support and if you can do that without giving up speed, you've got more revenue per system.

你能同時支援的客戶數量；如果你能在不犧牲速度的情況下做到這點，那每套系統就能帶來更多營收。

Now, I don't want to go into the specifics of our relationship with OpenAI, but one of the things that makes us so excited about this partnership is that you keep our speed and you increase throughput, which makes each system more profitable.

現在，我不想深入我們與 OpenAI 關係的細節，但讓我們對這個合作如此興奮的原因之一，是你能維持我們的速度並提升吞吐量，這會讓每套系統更有利潤。

Each system is generating more tokens. That means tokens cost less. Tokens use less power. So not only do we make more on top-line, but our margins improve.

每套系統會產生更多 tokens。這代表每個 token 的成本更低。tokens 耗用更少電力。所以我們不僅營收（top-line）增加，毛利率也會改善。

Not only do our margins improve and our top-line improve, but it makes each data center investment more valuable because data center is a power envelope and if you can get more tokens out of that power envelope, that converts to more dollars.

不僅我們的利潤率提升、營收也提升，而且也讓每一筆資料中心投資更有價值，因為資料中心受限於電力上限；如果你能在同樣的電力上限下產出更多 token，就能轉換成更多美元收入。

So it's an enormously powerful thing and with AWS doing this aggregation with us and with AMD doing this aggregation with us. We've tied up about half the leading chip makers.

所以這是一件極其強大的事情，而且 AWS 與我們一起做這種聚合，AMD 也與我們一起做這種聚合。我們已經把大約一半的領先晶片製造商都綁在一起了。

And so it's very powerful story.

因此這是一個非常有力的故事。

Excellent. And then the following question is just you mentioned in the script a couple of times that you'll increase your throughput of the wafer scale engine by a factor of 20 by the end of 2027.

很好。接下來的問題是，你在講稿中提到好幾次，你們將在 2027 年底前把晶圓級引擎（wafer scale engine）的吞吐量提升 20 倍。

Does that remove the need for some of this disaggregated compute or heterogeneous inferencing that you're talking about? Or does that just make the entire throughput of the heterogeneous solutions just that much?

這是否會消除你所談到的某些解耦式運算或異質推論的需求？還是說這只會讓異質解決方案的整體吞吐量變得更高？

Faster or higher throughput?

更快或更高吞吐？

I think we're exploring all sorts of ways to drive throughput up, right? If your throughput increases 20x and your costs stay the same, you're in pretty darn good shape, right? So our systems are improving throughput. We're looking for ways to improve the throughput of disaggregated solutions. We're looking at all sorts of different inventions, technologies, partnerships that continue our sort of pattern of industry-leading performance and vastly increasing throughput.

我想我們正在探索各種方式來提升吞吐量，對吧？如果你的吞吐量提升 20 倍而成本維持不變，那你的狀況就相當不錯，對吧？所以我們的系統正在提升吞吐量。我們也在尋找提升解耦式解決方案吞吐量的方法。我們在看各式各樣的發明、技術與合作夥伴關係，以延續我們一貫的產業領先效能並大幅提升吞吐量。

That's a really good question. I mean, that is what we're thinking about in our roadmap, that exact point.

這是個非常好的問題。我的意思是，這正是我們在路線圖中思考的事情，就是你提到的那個點。

Thank you. Our next question comes from Joe Moore with Morgan Stanley.

謝謝。下一個問題來自摩根士丹利（Morgan Stanley）的 Joe Moore。

Your line is open.

您的線路已開通。

Yeah.

是的。

Thank you. On the lines of what you were just talking about, when you talk about disaggregated decode, where are you in terms of commercialization of that? Like is there we know you can do fast inference at scale. You've done it.

謝謝。延續你剛才談到的內容，當你談到解耦式解碼（disaggregated decode）時，你們在商業化方面進展到哪一步了？例如，我們知道你們可以在大規模下做快速推論。你們已經做到了。

When it comes to disaggregation, is that ready to deploy now? And what work needs to be done over the next kind of year to get to the types of improvements that you guys are talking about?

談到解耦，現在已經可以部署了嗎？而在接下來大約一年內，為了達到你們所說的那些改善，還需要完成哪些工作？

We have disaggregated inference with GPUs running in our labs right now.

我們目前在實驗室裡已經有以 GPU 運行的解耦式推論（disaggregated inference）。

I think it will be deployed and available in Q4.

我認為它會在第四季部署並提供使用。

Okay, thank you. And then you talked in your script about the ability to work with the installed base of GPUs.

好的，謝謝。接著你在講稿中談到能與既有安裝基礎（installed base）的 GPU 協作。

Is there when you work closely with Amazon, work closely with AMD, is that stuff going to work better than kind of what you would be able to do with like NVIDIA installed base GPUs that are out there?

當你們與 Amazon 密切合作、與 AMD 密切合作時，那些東西是否會比你們能在市面上既有的 NVIDIA 安裝基礎 GPU 上做到的效果更好？

I think it's fair to say, though we haven't done it yet, I think it's certainly fair to say that Helios racks will give us bigger.

我認為可以這麼說，雖然我們還沒做過，但我覺得可以合理地說，Helios 機櫃（racks）會帶給我們更大的。

And better solution than if we were to use the 355s, right?

以及更好的解決方案，相較於如果我們使用 355s，對吧？

And Trainium 3s will give us better solution than if we were to use Trainium 2s.

而 Trainium 3 也會帶給我們比使用 Trainium 2 更好的解決方案。

And if we were to use other GPUs, the current generation, the top of tree generation, will give us better performance than the top of tree minus one generation.

而如果我們使用其他 GPU，現世代、最頂級的那一代，會比最頂級的前一代（top of tree minus one）帶來更好的效能。

But I think it's also fair to say that the minus one generation will be vastly better in a disaggregated solution than not in a disaggregated solution.

但我也認為可以合理地說，前一代在解耦式解決方案中，會比不採用解耦式解決方案時好得多。

All right. And in an environment where everyone's trying to extend the life of their hardware and continue to keep it delivering valuable tokens, this is an important option. Does that make sense?

好的。而在一個每個人都在嘗試延長硬體壽命、並持續讓它產出有價值 token 的環境中，這是一個重要的選項。這樣說有道理嗎？

Thank you. Our next question comes from Vijay Rakesh with Mizuho. Your line is open.

謝謝。下一個問題來自瑞穗（Mizuho）的 Vijay Rakesh。您的線路已開通。

Just a couple of quick questions. On the you mentioned the 600 megawatt signed capacity power and then 10x increase in capacity by the end of the year.

我有幾個很快的問題。關於你提到已簽約的 600 兆瓦電力容量，以及到今年年底容量提升 10 倍。

Do you think that should help you accelerate some of the ramps in 2027?

你認為這是否能幫助你們在 2027 年加速某些爬坡（ramp）？

Yeah, of course. I think that we are pursuing data center capacity around the world every day.

是的，當然。我想我們每天都在全球各地追求資料中心容量。

And we're doing it because we have tremendous demand for fast inference and the faster we can deploy, the faster revenue grows and that's true not just for our cloud business, but it turns out to be true for our customers on-prem business, that the faster they can get data centers, the faster we can ship them hardware and so it is top of mind. It is something I spend an enormous amount of time on.

我們這麼做是因為對快速推論有極大的需求；我們部署得越快，營收成長就越快。這不僅適用於我們的雲端業務，也同樣適用於我們客戶的地端（on-prem）業務：他們越快拿到資料中心，我們就能越快把硬體出貨給他們，所以這是我們最優先關注的事情。我在這上面花了非常多時間。

We have a whole team now. We're pretty darn good at chasing down data centers around the world and once you've signed them, your job isn't done, as you well know.

我們現在有一整個團隊。我們非常擅長在全球各地追蹤並拿下資料中心，而且一旦簽下來，你的工作還沒結束，這點你很清楚。

We have people on site every day. We are engaged with the developer and the construction firms at every stage to do our best to keep them on track.

我們每天都有人在現場。我們在每個階段都與開發商與營造公司保持合作，盡最大努力讓進度維持在軌道上。

And so the faster we can do that, I think the faster we can ramp our revenue.

因此我認為我們做得越快，營收爬坡就越快。

And then as you look at partnering, I know you mentioned hyperscalers, but there's a whole emerging.

另外，談到合作夥伴，我知道你提到了超大規模業者（hyperscalers），但還有一整個新興的。

Neo Cloud group that's coming up, they're getting financing. There's a lot of financing structures being developed across Wall Street, I guess.

新雲（Neo Cloud）族群正在崛起，他們正在取得融資。我想華爾街正在開發很多融資結構。

How is that pipeline developing for you?

你們在這方面的管線（pipeline）發展得如何？

Thanks.

謝謝。

Sure. I think early on, the Neo Clouds were very focused on NVIDIA.

當然。我認為在早期，Neo Clouds 非常聚焦於 NVIDIA。

I think as the business has become clearer to investors. There are neo-clouds that are diversifying and find themselves less dependent on one hardware vendor.

我認為隨著投資人對這門生意的理解更清晰。有些 neo-cloud 正在多元化，並發現自己對單一硬體供應商的依賴降低了。

And so the opportunities for us in that category are large. There are neo-clouds that are multi-vendor. There are neo-clouds that are AMD only. There are neo-clouds that are coming up out of people who have power assets.

因此我們在這個類別的機會很大。有些 neo-cloud 是多供應商（multi-vendor）。有些 neo-cloud 只用 AMD。也有一些 neo-cloud 是由擁有電力資產的人所建立起來的。

And I think in 2027, that'll be an important part of our business.

我認為在 2027 年，這將成為我們業務的重要部分。

Thank you. I'm showing no further questions at this time. This concludes today's conference call.

謝謝。目前顯示沒有其他問題。今天的電話會議到此結束。

Thank you for participating. You may now disconnect.

感謝各位參與。您現在可以掛線。